Start with what you can see

If you look only at the big number, nothing happened this week. The S&P 500 hit its twelve-month high in the week ending August 16, at 7,785.76. Seven days later, the weekly close was 7,674.37. That's 1.4% lower. Over twelve months it's still up 18.7% — a figure that in any other year would have made headlines.

A 1.4% pullback is nothing. It happens several times a year and almost never means anything. But the index is an average. And an average, by definition, hides what its components are doing.

The move the index doesn't show

Here's the part worth understanding, because the concept has a technical name — sector rotation — and underneath it is simple.

Picture the market split into two families. On one side, cyclicals: companies that earn a lot when the economy is strong and suffer when it cools — construction, autos, travel, industrials, banks. On the other, defensives: companies that sell the same thing whatever happens — power, water, staple food, pharma, telecoms. Nobody stops paying the electricity bill because there's a recession.

When investors are comfortable, they buy cyclicals. When they start not to be, they move to defensives. And here's the detail: if money moves from one to the other inside the same index, the index barely notices. What leaves one place enters another. The average stays almost flat.

That's why the ratio between the two families is watched. It's a thermometer of appetite, not of level. Our cyclical-versus-defensive ratio has fallen around 10% in two weeks. A 10% move in the ratio, with the index giving up 1.4%.

Read it again. The index moved 1.4%. Risk preference, seven times more.

And sentiment, which turned in two weeks

It's not an isolated reading. The survey says it, the flows say it, and our composite says it.

Our weekly sentiment reading — which combines several sources into a single number — has spent two weeks in negative territory, each worse than the last, dragging the six-week average to the edge of zero. We don't publish the numbers, for the same reason we don't publish the recipe: it shows in the dial. What we can tell you is what's underneath, because that is public.

• Professional advisors. Investors Intelligence bulls fell from 57.4% to 54.7% in a week; bears rose from 14.8% to 15.1%. Not a reversal — a give.

• Retail. AAII went from 37.0% to 34.7%.

• The bull/bear ratio crossed below one. From 1.08 to 0.75. That cross matters more than the level: for the first time in weeks, more people expect falls than rises.

• And money, which doesn't lie. Weekly flows into US equity funds went from +$1,329M coming in to −$853M going out.

Four different measures, made by different people, with different methods. All four pointing to the same place in the same fortnight.

The detail that doesn't fit

And yet something doesn't add up, and I'd rather say it than paper over it.

If people were truly scared, they'd buy protection. The most common way to hedge equities is to buy put options — insurance that costs a premium and caps the loss if the market falls. When fear rises, demand for puts rises, and the put/call ratio rises with it.

This week it did the opposite. The put/call ratio has fallen three weeks running: 0.652, then 0.638, then 0.636, now 0.609.

So: declared sentiment turned over, money rotated to defensives, and nobody is paying to hedge. Let's not overstate it. That 0.609 isn't a twelve-month low — 0.51 was seen in May, 0.52 in January — so this isn't extreme complacency. But the mismatch is there, and of the three things that happened this week it's the one that fits me least.

Meanwhile, the hard data is at multi-year highs

This is where the week gets genuinely interesting.

Our Warning Signal — the indicator that flags when a vicious circle is forming between the macro, corporates and finance, the pattern that preceded 2000, 2008, 2020 and 2022 — just posted its best reading since October 2018. Almost eight years. It has risen in eight of the last nine weeks.

And here you have to state clearly what it means, because it's easy to read backwards: in this indicator, the dangerous ground is negative territory. In the spring of 2020 it collapsed to levels not seen since 2008. A high reading is not an alarm. It's the opposite: it says there is no vicious circle today. What it does not say — and this is worth underlining — is anything about what price will do next week.

At the same time, our leading indicator of the real economy posted, in its raw reading, its highest level since June 2022. Four years. That the two coincide is rare. That they coincide just as sentiment breaks, more so.

But. The leading indicator has been stuck in the same place for two months. The raw number keeps rising; the smoothed index no longer gains slope. It arrived high and stopped. And the new quarter's GDP comes in below the last: 2.10 versus 2.69.

Underneath all that, something orderly is worth a mention. The lagging indicator — the one that reflects what already happened — has been falling for three months while the leading one rises. Leading up, lagging down: that's the typical sequence of an early phase of the cycle, not a late one. In Europe, the economic sentiment index also improves for a third month (95.4 in June, 96.9 in July).

The precedent worth looking at

The last time the Warning was where it is today was October 2018. It's worth remembering what happened next, because it's one of the six episodes we document in the 33-year backtest. The market fell hard in Q4 of that year. And the system gave no alarm — it classified that as a correction within the regime, not a regime change. It was right: the market recovered it all in the following months.

Same level of reading. Different context. It's not a prediction, and I'm not using it as one: it's the only comparable precedent we have, and it seems more honest to show it than to hide it.

Hypothetical results on partially reconstructed data; process validation, not a recommendation.

The read across our four layers

Layer 1 · Warning — 🟢 and strengthening. Its best reading since October 2018, up in eight of the last nine weeks. No vicious circle between macro, corporates and finance. The soundest part of the picture.

Layer 2 · The Crowd — 🟡 and this is the week’s change. Sentiment crosses into negative territory in its weekly reading, surveys ease, flows leave and money rotates to defensives. With the anomaly already noted: nobody is buying protection.

Layer 3 · The Money — 🟡, unchanged. No Fed meeting. The July 29 picture still stands: rates at 3.50–3.75% and a 9-to-3 vote, with Hammack, Kashkari and Logan calling for a quarter-point hike. The minutes published on August 19 added a nuance that doesn't change the colour but does change the weight: “several participants favored an increase of 25 basis points” — more than the three who ultimately dissented argued for it in the room. The lid stays on.

Layer 4 · The Macro — 🟢 with an asterisk. Leading indicator at a four-year high in the raw number, lagging falling, coincident no longer deteriorating. The asterisk is the slope: the leading indicator has stopped accelerating right at the top.

Same regime, board in motion

That the regime doesn't change doesn't mean nothing is happening — it means not enough has happened yet. The regime is still Compressed Spring, and this week it fits better than it did a month ago.

The idea is simple: fundamentals pushing from below, monetary policy pressing from above, and a mood that doesn't go along. The more distance between what the data says and what people feel, the more tension the system stores. Through July and much of August, optimism had blurred that shape. This week it fits again — and with hard data higher than in years.

That's not a forecast of a rise or a fall. It's a description of how much energy is stored. A regime changes when two or more layers turn at once, and this week one turned.

We'll tell you the day it changes. That's the job.

In 30 seconds

• The S&P closed at 7,674, 1.4% below its August 16 high. Over twelve months, +18.7%.

• Inside, the cyclical/defensive ratio shed close to 10% in two weeks. Money rotated even as the index barely moved.

• Sentiment turned — our weekly reading strings together two weeks in the red: advisor bulls 57.4% → 54.7%, AAII 37.0% → 34.7%, bull/bear ratio 1.08 → 0.75, flows from +$1,329M to −$853M.

• The anomaly: put/call falls three weeks running (to 0.609). Nobody is buying protection.

• Our Warning posts its best reading since October 2018, and the leading macro indicator its best since June 2022. Both at once.

• The but: the leading indicator has gained no slope for two months, and the new quarter's GDP comes in at 2.10 versus 2.69.

• The “so what”: when sentiment withdraws and hard data doesn't follow, the tension doesn't resolve — it accumulates. That doesn't anticipate a direction; it describes a situation. And this week's situation is that the market got nervous inside without the index telling it outside.

What to watch over the next ten days

Three dates, and Thursday's carries the most signal of the year.

• August 27–29 · Jackson Hole. Kevin Warsh's first symposium as Fed Chair. He arrived promising a change in how monetary policy is decided and set up five working groups whose conclusions aren't published until year-end. What to listen for is not where rates go: it's whether he talks framework or conjuncture, and whether he pins anything down on the balance sheet.

• August 28 · BLS preliminary benchmark revision. It falls inside the symposium. It corrects the published payrolls against real tax records. If the revision is large and to the downside, the strong-labor-market story weakens the same day the Fed speaks. We promised this in the comments two weeks ago and we'll cover it here.

• September 15–16 · FOMC with dot plot. Where whatever is hinted in Wyoming gets made concrete.

Let’s keep the conversation going

One question I'll leave open, and that goes beyond this week: if the index can stay almost still while money changes place inside, what is it really measuring when someone says “the market is calm”?

Open the subscriber chat

For investment committees, family offices and company boards: we produce the Institutional Macro & Markets Report on demand, subject to qualification — the same four-layer process, with the full proprietary series and the watchlist. Write to us stating your role and mandate.

Request the institutional report →

About the author

Francisco Salvador. Macro analyst since 1993. FGA Research & Advisory · Est. 2006 · 33 years of study. The 33-year backtest is open, with a permanent DOI, on the Evidence page: macro.fgaresearch.com/p/evidencia-la-prueba-esta-en-abierto

Sources: FGA proprietary series (Warning Signal, Sentiment, leading and coincident macro) as of August 21–23, 2026 · Investors Intelligence, AAII, CBOE and weekly fund flows, same dates · FOMC minutes of July 28–29, published August 19 (federalreserve.gov) · European Commission (ESI).

Notice. Informational and educational content: cycle analysis and reading of public documents, not a regulated investment service. It is not advice or a recommendation — general or personalised — nor a buy/sell signal. The readings and regimes describe the phase of the cycle; they do not anticipate tops or bottoms nor promise protection against falls. Historical results cited are hypothetical, not audited, and constitute process validation, not a recommendation. Past results do not guarantee future results; investing involves risk, including total loss of capital.

Independence. Capital. Conviction. · FGA Research & Advisory · Est. 2006 · 33 years of study

In one sentence. The July employment report does not say companies are letting people go: it says they have stopped hiring. The two look alike in a headline and behave very differently in a cycle. It touches Layer 4, the Macro, and does not move it: the labour market own leading indicator sits at its best level since 2022. The regime remains Compressed Spring.

Start with the contradiction, because it resolves itself

On Friday 7 August, the US Bureau of Labor Statistics published two sentences that look incompatible.

The first: the economy lost 23,000 jobs in July. The second: the unemployment rate fell, from 4.2% to 4.1%.

How do you lose jobs and see unemployment fall at the same time?

Because these are two different surveys measuring different things. And because, above all, the unemployment rate only counts people who are actively looking for work. If someone stops looking, they leave both the numerator and the denominator: unemployment falls without anyone finding a job.

That is exactly what happened. The participation rate — the share of the population working or looking for work — fell to 61.4%. It is down 0.7 points since January. The employment-to-population ratio also fell, to 58.9%.

In plain terms: unemployment did not fall because jobs were created. It fell because fewer people are looking for them.

The three numbers that actually matter

Beyond the headline, three concrete things.

One: the revisions. May and June were revised down. May went from +129,000 to +63,000. June, from +57,000 to +20,000. Between them, 103,000 jobs that had been counted and are no longer there. The average over the prior twelve months was +34,000 a month; July came in at −23,000.

Two: where the losses are. Local government education (−50,000), retail trade (−19,000) and financial activities (−14,000, now 121,000 below their May 2025 peak). Health care is still adding jobs (+22,000), but more slowly than its twelve-month average (+36,000).

Three: temporary layoffs. They rose by 153,000, to 921,000. Permanent job losers barely moved. That is the difference between sending people home for a while and closing the position.

The same labour market, measured at three different points of the cycle. Source: BLS, US Department of Labor and The Conference Board · FGA Research illustration.

In 30 seconds

Regime: Compressed Spring — no change. This report does not change the colour of any layer.

• Layer 1 · Warning — 🟢↑ no direct link. A labour market cooling through hiring rather than firing does not close any vicious circle.

• Layer 2 · Crowd — 🟡 US retail investors raised their bullishness to 37.0% in the AAII survey of 6 August, which closed the day before the data. Sentiment ran ahead and has not digested it yet.

• Layer 3 · Money (the Fed) — 🟡 here there is a consequence: a labour market that cools without breaking is the argument of the wing that does not want to raise rates. The next real date is Jackson Hole, 27–29 August — Kevin Warsh first as Fed Chair.

• Layer 4 · FGA Macro — 🟢↑ the layer this touches, and it does not move: the leading indicator of employment is at its best level since October 2022. It is the coincident one that has stalled.

The so-what: employment is the worst place to look for the future and the best place to confirm the past. When the leading measure says no layoffs and the coincident one says no hiring, what you have is an economy cooling without breaking. That is not a recession starting; it is a cycle losing momentum. Telling the two apart is half the job.

This is how we read the cycle every week: four layers, made clear. Get it free in your inbox.

Subscribe for free

Why employment always arrives late

This is the part that is rarely explained, and it is what gives the data its value.

Put yourself in the shoes of someone running a company and it becomes clear in a minute. Sales start to soften. What do you do in month one? Nothing. You wait. Month two? You stop replacing leavers: you do not fire anyone, you simply do not backfill. Month three? You cut overtime. And only when there is no alternative left do you make redundancies.

Quarters can pass between an economy cooling and that cooling showing up in the unemployment data. Which is why employment, as a cycle thermometer, always arrives late.

And this is not our opinion: it is in the official accounting of the cycle. The Conference Board, which has built the reference indices for decades, files employment variables into three separate drawers:

• Leading — weekly initial jobless claims. They measure layoffs in close to real time. Four-week average: 198,750. Below 200,000 for the first time since October 2022.

• Coincident — nonfarm payrolls. They measure the jobs that already exist. July: −23,000, with 103,000 of downward revisions across the two prior months.

• Lagging — average duration of unemployment. It measures the scar of the previous cycle. 1.8 million people have been out of work for 27 weeks or more: 25.5% of all unemployed.

Read it top to bottom and the report stops being contradictory: nobody is firing (leading indicator at its best in almost four years), hiring has stopped (coincident stalled), and those already out find it hard to get back in (lagging still tight).

That is the natural order of a cycle decelerating. Not of one breaking. When it breaks, the first drawer is the one that jumps — and that is precisely the one in the best shape today.

The detail almost nobody has reported: the surveys say the opposite

Four days before the employment data, another number came out, pointing the other way.

The ISM manufacturing index — the monthly survey of industrial purchasing managers, one of the longest-running leading indicators there is — printed 55.6 in July, its highest since May 2022. Above 50 means expansion. Its employment component rose to 52.8, entering growth territory for the first time in 33 months. Sixty percent of respondents said their company is hiring.

And the ISM services index, published two days later, said the exact opposite in its employment component: 47.4, back into contraction after a single month in positive territory.

Three measures of the same month. One says industry is hiring again. Another says services have stopped. The third — actual payrolls — says the net balance is negative, and that the hole is in local government and retail.

This is not a data failure. It is a two-speed economy: the part that manufactures is recovering; the part that employs the most people is stalling.

Reading of the four layers as of 13 August 2026. The report adds context to Layer 4; it does not move it. Source: FGA Research.

And the rule everyone quotes: today it says nothing

Every time a weak employment print lands, the Sahm rule appears. It is worth understanding, because it is simple and because today it settles the debate.

The rule says: if the three-month average of the unemployment rate rises half a point or more above its low of the previous year, historically a recession had already begun. It does not predict: it dates. It confirms that something already underway was indeed underway.

With the July data, the reading is −0.03. The threshold is +0.50. It is not close: it is on the other side, and falling.

Here honesty about the caveat matters, because it is what separates a useful reading from a reassuring headline: the rule looks at the unemployment rate, and the unemployment rate fell this month for the wrong reason — people who stopped looking, not people who found work. Any indicator built on that rate inherits that blind spot. Which is why we never read a layer off a single number.

How we read it

Our Layer 4 systematically compares leading against coincident. That comparison is what produces the phase signature, and it works in both directions:

• Leading up + coincident down → an early turn, the economy is changing feet.

• Leading down + coincident up → the end of an expansion: the good number of today is the momentum of yesterday.

The July report fits the first. And it fits what we have been reading for two weeks: our own coincident measure posted its largest jump in a year in the raw monthly data, while the lagging measure sank to its low for the year. Local government and retail subtract; real activity adds.

The obligatory caution, stated just as plainly: the classic leading index does not fully agree. The Conference Board index fell 0.2% in June, to 99.1, unwinding part of the prior two months of gains. Its July reading is published on 20 August. If it falls again with the ISM at four-year highs, we will have a divergence that needs explaining — and we will explain it.

Layer 4 stays green today. The report does not change the colour of any layer. It changes where we look inside it.

What does NOT change

The regime remains Compressed Spring, and it will remain so until two or more layers turn at once.

And there is a discipline here that is hard to hold. A headline reading US economy destroys jobs triggers a physical urge to do something. The method exists precisely for that: to separate what is already happening from what would have to happen for the diagnosis to change.

What would have to happen here is concrete and watchable: weekly jobless claims would have to start rising in earnest. That is the leading drawer. Today it is at its best level since 2022. While it stays there, a month of negative payrolls is a cycle cooling, not one breaking.

What to watch — three specific dates

This piece does not close today. It closes over the next three readings, and all three have a date:

1. 19 August — FOMC minutes. On 29 July the Fed held rates at 3.50–3.75% with three votes to raise them, the largest dissent in a single direction since 2016. The minutes will show whether that hawkish wing was already factoring in a cooling labour market.

2. 20 August — Conference Board Leading Economic Index. If it falls again while the ISM rises, the divergence between surveys and hard data stops being anecdotal.

3. 28 August — the BLS benchmark revision. Once a year, payrolls are checked against the actual tax records of nearly every employer. That day brings the preliminary estimate of how much had been over- or under-counted. In recent years that correction has been large. It is by far the most important employment number of the month — and almost nobody has it in the calendar.

All three fall in the same two-week block. And the third lands on the opening day of Jackson Hole.

A report gives you the photograph; the follow-up gives you the film. In a labour market cooling through the hiring channel, the only thing that really matters is whether the cooling moves from the coincident drawer to the leading one. You cannot see that in one data point: you see it in the sequence.

And you — which layer are you watching this week: the Warning, the Crowd, the Money or the Macro? Drop it in the comments; I read them all.

Editorial and educational content, not investment advice. No buy or sell signals and no price targets. Past results — real or simulated — do not guarantee future results; investing involves risk, including the total loss of capital.

Why we read employment the way we do

On the cycle, we publish the record. The four-layer framework is documented in a working paper with a DOI (WP-2026-01, DOI 10.5281/zenodo.20709770), and the results are stated in full: 6 of 6 episodes ahead of the break (exact binomial p ≈ 0.016) — defensive before 2000, 2008, 2020 and 2022, with no false exit in 1998 or in the fourth quarter of 2018. In production since 2007–2008, with no reoptimisation over the episodes being judged. And a pre-registered public forward test on OSF: every regime change is timestamped live, before the outcome is known.

We say it for one specific reason: anyone can claim an employment print changes nothing. That claim is worth exactly what the track record of the person making it is worth. Ours is open, dated and checkable — failures included.

Reading this from an investment committee or from the top of a company?

We produce the Institutional Macro and Markets Report on request and subject to qualification: the same four-layer process, with the full proprietary series and the watchlist. Write to us stating your role and your mandate; we reply within 24 hours.

Request the institutional report →

Information, not advice. Independence. Capital. Conviction.

Independence. Capital. Conviction. · FGA Research and Advisory · Est. 2006 · 33 years of study

Sources

• U.S. Bureau of Labor Statistics, The Employment Situation — July 2026, USDL-26-1291, 7 August 2026 — bls.gov

• U.S. Department of Labor, Unemployment Insurance Weekly Claims, week ending 1 August 2026 — dol.gov

• Institute for Supply Management, July 2026 Manufacturing and Services PMI reports — ismworld.org

• The Conference Board, US Leading Economic Index and component descriptions — conference-board.org

• Federal Reserve Bank of St. Louis (FRED), Sahm Rule Recession Indicator (SAHMCURRENT), July 2026 — fred.stlouisfed.org

• Federal Reserve, FOMC statement of 29 July 2026 — federalreserve.gov

• AAII Investor Sentiment Survey, 6 August 2026 — aaii.com

• FGA method and working paper WP-2026-01 — DOI 10.5281/zenodo.20709770

In one line. For fifteen years, software was the ideal borrower. AI is turning the ideal borrower into the most exposed one. The BIS documents it; lenders haven't priced it. But the durable lesson here isn't about software — it's about the difference between an indicator that measures a level and one that measures a feedback loop, and why that distinction decides whether you get 2018 wrong.

The setup, in plain terms

Suppose you lend for a living. You want the perfect borrower: predictable revenue, high margins, no factories, customers who rarely leave.

For fifteen years that borrower existed, and it was called software. Subscriptions renew themselves. High gross margins. Almost no capital expenditure. A cash flow machine.

Private credit — funds lending directly to mid-sized companies, bypassing banks — concentrated there. And because most of that market is opaque, nobody could measure how much. Which is what makes the new BIS Bulletin No 128 (Avalos, Cornelli and Eren, 14 July 2026) worth reading: it found a window.

That window is the BDC — business development company. It's a US vehicle that, by statute, must report its loans one by one, quarterly, to the SEC. There are around 170 of them, holding some $550 billion in assets, originating roughly one fifth of all US direct lending. They are the visible slice of an otherwise unobservable market. Whatever is true of them is plausibly true of the rest.

What the BIS found

One. BDCs have lent about $115 billion to software firms — roughly a fifth of all their lending and over 80% of their technology portfolios. Around three quarters of that sits in productivity, application and workflow tools: precisely where AI substitutes most directly.

Two. Lenders are not charging for the risk. Spreads on software borrowers are indistinguishable from other sectors — and they have fallen, fastest on newly issued loans and on loans from non-traded BDCs, the least scrutinised segment. The BIS puts it plainly: lenders are being paid less to carry a risk that has grown.

Three. The exposure is doubly concentrated. 60% of software lending now goes to firms borrowing from seven or more BDCs at once — under 10% in 2015. Meanwhile the five largest BDCs hold about 37% of that credit, the ten largest more than half. A shock would surface across many balance sheets simultaneously, with losses landing on few.

And the honest counterpoint, which the BIS states first: less than 1% of these loans are past due, a smaller share than the rest of their portfolios. Loans are mostly senior and secured. Leverage is capped by statute. Nothing is broken.

Note the asymmetry, though. Public equity markets have been discounting AI disruption since December 2022 — software valuations went from a premium over the broader tech sector to a discount. Credit hasn't moved at all.

Why private credit is a harder object to observe

Before the methodological part, one structural point, because it explains why a bulletin like this exists at all.

A bank is observed continuously: a supervisor sees its book, capital ratios are public, stress tests are run on it. Private credit is not one thing but a family of vehicles under different legal regimes, most of them outside registration requirements, with thin disclosure and hard-to-trace links to banks and other intermediaries.

Three consequences follow, and they compound:

• Valuation is largely mark-to-model. These loans do not trade. Their reported value is the manager's own assessment — which is not fraud, but it does mean the number that reassures you was produced by the party it reassures.

• There is no system-wide view. No one can aggregate exposures across the sector the way a supervisor aggregates bank exposures. The BIS works around this by reasoning from BDCs outward, and says so explicitly.

• Funding structure varies enormously. Listed BDCs are equity-funded with no redemption right — genuinely stable. Non-traded vehicles offer periodic, capped redemptions, and caps are a promise about normal weather.

That is why the correct reading of Bulletin 128 is not "BDCs are dangerous". It is the opposite: BDCs are the better-lit room, and this is what the better-lit room looks like.

What the BIS was able to measure, in one table. Source: BIS Bulletin No 128 · FGA Research illustration.

The BIS measures each link separately. The question an early-warning indicator has to answer is whether they have started to push on one another. Source: BIS Bulletin No 128 · FGA Research illustration.

The part worth teaching: levels versus feedback

Here is where the episode becomes generally useful, well beyond private credit.

Most early-warning indicators measure a level. Credit spreads. Leverage ratios. Valuation multiples. Default rates. Each answers: how stretched is this one thing? And each shares the same failure mode — a single dimension can stretch a long way without a cycle turning, and it usually does.

The alternative is to measure feedback: not how stretched any one dimension is, but whether the dimensions have started to amplify each other. Whether a shock to corporate revenues tightens financial conditions, and that tightening damages the real economy, and that damage feeds back into corporate revenues. A vicious circle.

The distinction is not academic. It is the difference between two kinds of error:

• False alarms (type I). A level-based indicator fires whenever its dimension stretches. It calls crises that never come. Every level indicator has a long list.

• Missed turns (type II). An indicator too reluctant to fire arrives after the damage.

A feedback-based indicator makes a specific trade: it accepts being later on the level in exchange for being right about the regime. It refuses to fire on one stretched dimension. It waits for the loop.

Our own FGA Warning Signal is built on exactly that principle — it fires when three dimensions (macro, corporate/micro and financial markets) enter a vicious circle. And the record shows both sides of the trade:

• It hit −2.9 in 2008 and −2.8 in 2020, its historical lows. The loop closed.

• It did not fire in 1998 or in Q4 2018 — episodes where one dimension tightened sharply, markets fell hard, and the feedback never materialised.

  

  Why the Warning did not fire in 1998 or Q4 2018. Source: FGA Research · historical results, illustrative.

A pure credit indicator would have fired in Q4 2018. That is the whole argument, in one counterexample. The methodology is documented in the working paper — The Robustness of a Four-Layer Macro-Cycle Classification System (WP-2026-01, DOI 10.5281/zenodo.20709770) — with the full record: six of six episodes ahead of the break, exact binomial p ≈ 0.016.

Applying it here

So what does Bulletin 128 actually give us? Not a reading. An anatomy — the three links of a loop that has not yet started turning:

• Corporate/micro: AI compresses software revenues.

• Financial: that risk isn't in the price of credit, the loss buffer has thinned, and lenders are concentrated.

• Macro: if the loop closed, credit would withdraw from the US mid-market — not just from technology.

And then the first link would worsen again. That is a vicious circle.

It isn't turning today. The first link hasn't fired — real delinquency remains under 1% — and the other two aren't amplifying each other. Our cycle layer reads green and rising. The document changes what we watch. It changes no colour.

Four-layer reading as of 27 July 2026. The BIS report adds context to Layer 1; it does not move it. Source: FGA Research.

A second lesson: the metrics that reassure you are the ones looking backwards

The BIS is careful about something that deserves emphasis, because it generalises to almost any credit analysis.

Delinquency, impairment and write-down data are backward-looking by construction. A revenue shock takes time to become a missed payment. And there is an instrument — payment-in-kind, where the borrower settles interest by issuing more debt instead of paying cash — that postpones the moment strain becomes visible at all.

So a portfolio can be deteriorating and reporting well simultaneously. Today's calm numbers may simply be old numbers.

This is why the BIS chose BDCs. Not because BDCs are the problem — their statutory leverage caps and secured lending make them the better-protected part — but because they are the only part you can see. The analytical move is to reason from the observable to the unobservable, and state clearly that you are doing so. That is worth imitating.

What to watch — and why sequence beats snapshot

Three concrete markers, in order of importance:

1. Real delinquency, not accounting delinquency. Watch whether that "under 1%" starts to rise, and specifically whether it rises faster in software than in the rest of the portfolio. That would be the first link firing.

2. Whether spreads wake up. Watch for credit beginning to charge software borrowers more than everyone else. Today it doesn't discriminate. The day it does, the market will have priced the risk — late, but priced.

3. Non-traded funds. About a quarter of this credit sits in semi-liquid vehicles with quarterly redemption caps around 5% of assets. In early 2026, a large one breached its cap for the first time. A repeat would be funding fragility surfacing in fair weather.

And this is the methodological point to end on. A vicious circle is not a state you can read from a single document — it is a process, and processes are only visible in sequence. One report gives you the photograph. Tracking the layers week after week gives you the film. For a feedback-based indicator, the film is the only thing that matters: the question is never "how bad is this number today" but "have these three started pushing on each other yet".

That is why the layer tracking is where this piece actually gets resolved — not here, but over the coming weeks.

A practical checklist: how to read an institutional warning

If there is one transferable thing in this piece, it is this. Reports from the BIS, the IMF, the ECB or the Fed are written to be careful, which means the strength of the claim is often buried in qualifiers. Four questions extract it quickly.

1. Is this a measurement or a projection? Bulletin 128 is overwhelmingly measurement — loan-level data through Q4 2025, 18,097 borrowers, a panel of 152,116 spread observations. That deserves far more weight than a forecast. The distinction matters because the two have very different track records: we looked at exactly that in our study of how often the official reports get it right.

2. What is the institution's loss function? The BIS is a financial-stability body: it is structurally more willing to warn early and be wrong than to stay silent and be late. The IMF, forecasting growth for member states, leans the other way. Neither is a flaw — but you should not read a BIS warning with the same prior as an IMF projection.

3. Does the document claim a mechanism, or only a correlation? This one claims a mechanism, and names each link. That is what makes it usable: a mechanism tells you what to watch next. A correlation only tells you what already happened.

4. What would have to be true for this to matter? The single most useful question, and the one most reports leave to the reader. Here the answer is concrete: real delinquency rising faster in software than elsewhere, spreads starting to discriminate, redemption caps under pressure. Write it down before the next headline arrives.

Notice that none of the four asks whether the warning is "right". That is the wrong question at this stage. The right question is what state of the world the warning is describing, and how you would know if it changed.

In this series

This piece is the third chapter of a thread we have been following since June:

• AI bubble? What the BIS warns — and how to read it by layers (30 June) — the BIS Annual Report and Bulletin 120: how the build-out of AI is financed, from cash flows to debt. Today's piece is the mirror image: what happens to credit lent to the firms AI may displace.

• War or technology: the summer's BIS–IMF diptych (14 July, in Spanish) — the same engine the BIS treats as a risk is what sustains the cycle for the IMF. Both readings coexist; the framework orders them.

• How accurate are the official reports? (16 July) — the measured track record of the WEO and the BIS as instruments: what you can ask of each and what you cannot. Useful reading right before a warning like this one.

Get the weekly cycle reading — free

Why we read this particular story the way we do

A closing note on standing, because this piece sits at the intersection of two things we have spent a long time on.

On AI, we are not commentators — we are practitioners, and early ones. Francisco Salvador founded one of the first research firms to apply artificial intelligence to market sentiment analysis in 2001: two decades before the current cycle, when doing this meant building the tooling yourself. That history is why we read the AI-disruption claim in Bulletin 128 as an operating question — which software categories actually get substituted, and how fast — rather than as a headline. The BIS points at productivity, application and workflow tools. Anyone who has built in that space knows why those go first.

On the cycle, we publish the record. The four-layer framework is documented in a working paper with a DOI — The Robustness of a Four-Layer Macro-Cycle Classification System (WP-2026-01) — and the results are stated in full: six of six episodes ahead of the break (exact binomial p ≈ 0.016), defensive before 2000, 2008, 2020 and 2022, with no false exit in 1998 or Q4 2018. Applied in production since 2007–2008, without re-optimising on the episodes being judged. Drawdown −29% against the market's −52%, 7.5× against 5.5× (2008–2026, illustrative). And a pre-registered public forward test on OSF: every regime change is timestamped live, before the outcome is known.

We say this for one reason only: a claim about feedback loops is worth exactly as much as the record of the person making it. Ours is open, dated and checkable — including the misses.

And you — when you read an institutional warning like this one, what tells you it's a change of regime rather than a strong headline? I read every comment.

Editorial and educational content, not investment advice. No buy or sell signals, no price targets. Past results — real or simulated — do not guarantee future results; investing involves risk, including total loss of capital.

Information, not advice. Independence. Capital. Conviction.

Independence. Capital. Conviction. · FGA Research & Advisory · Est. 2006 · 33 years of study

Sources

• BIS Bulletin No 128, "AI disruption in private credit: exposure to software firms in BDCs", F. Avalos, G. Cornelli and E. Eren, 14 July 2026 — bis.org/publ/bisbull128.pdf

• BIS Bulletin No 120, "Financing the AI boom: from cash flows to debt", January 2026 — bis.org/publ/bisbull120.pdf

• BIS Quarterly Review, March 2026, "Private credit's software lending meets AI disruption"

• F. Salvador, "The Robustness of a Four-Layer Macro-Cycle Classification System", WP-2026-01 — DOI 10.5281/zenodo.20709770

The verdict. The IMF's WEO forecasts levels, and its record at turning points has been measured: of 153 recessions between 1992 and 2014, the consensus anticipated only 5. The BIS report monitors vulnerabilities, and its record too: it warned of 2008 from 2003 — and called for rate hikes in 2011, three years too early. Neither is an oracle. We use them for what they are: two first-class inputs, with known strengths and weaknesses — read through a methodology of our own.

First, in plain words

If this is not your trade, one image is enough.

The IMF produces the weather forecast of the world economy. Twice a year — plus two updates — it publishes how much each country will grow, the way the weatherman says whether it rains tomorrow. That forecast is called the World Economic Outlook, the WEO.

The BIS is something else: it is the building inspector. It does not tell you whether it rains tomorrow. It checks the beams: how much debt has accumulated, who holds it, and where the crack is showing. Its annual diagnosis has come out every June for nearly a century.

Why should any of this matter to a normal person? Because these two documents move decisions that end up in your life: the interest rates that price your mortgage, government budgets, the portfolios where your pension sits. When a headline says "the IMF projects" or "the BIS warns", it is quoting one of these two reports — almost always without telling you the essential part: how often they actually get it right. And misreading them is not free: it is paid in decisions.

Keep one number in mind as you read: of 153 recessions, the experts anticipated 5. By the end of this piece you will understand why that figure does not make the experts useless — it makes it useless to ask them for what they cannot give.

The piece starts gentle and hardens as it goes: by the end it is a report. Stay as long as you like — the 30-second summary waits below.

Of 153 recessions across 63 countries (1992-2014), the consensus anticipated only 5 a year ahead. FGA illustration after An, Jalles and Loungani (2018).

What they are, and where they come from

The World Economic Outlook was born as an internal staff paper at the IMF in June 1969, for an informal Board discussion. For eleven years it circulated only internally. In May 1980 it was published for the first time; in 1984 public interest made it semiannual (April and October), and crises added special editions — December 1997 (Asia), December 2001 (after 9/11) — until the January and July updates became standard. Today its database covers 196 economies, projects five years out, and is public and free: probably the largest statistical public good in the economic world.¹

The BIS Annual Report is older than the IMF itself. The Bank for International Settlements was founded on 20 January 1930 under the Young Plan to administer Germany's Versailles reparations — hence "Settlements". It opened in Basel on 17 May 1930 and is the world's oldest international financial institution; it has published its annual report since 1931. Its biography has a dark chapter — during World War II it accepted from the Reichsbank 3.7 tonnes of gold that turned out to be looted from Belgium and the Netherlands, returned in 1948 — and a near-death: the 1944 Bretton Woods conference adopted a resolution calling for its liquidation "at the earliest possible moment". It survived. Today it groups 63 central banks representing about 95% of world GDP, and has hosted the Basel Committee since 1974 — the source of Basel I, II and III.²

Two genealogies that explain two temperaments. The IMF was born to coordinate policy among governments: its numbers are read in 190 capitals and negotiated with 190 sensitivities. The BIS answers only to central bankers: it can afford to be the uncomfortable voice.

A century of two reports: the BIS (above) and the IMF (below). FGA illustration.

How an official forecast is made

The WEO does not come out of one big model. It is built bottom-up: country-desk economists produce each national forecast; a central memo fixes global assumptions (oil, interest rates); successive aggregation rounds check consistency — for instance, that the world's current accounts sum to roughly zero. When the process converges, it is published.

Are the numbers politically massaged? The IMF's Independent Evaluation Office (IEO) investigated in 2014, and its answer has two parts. It found no evidence of political manipulation of WEO forecasts. But it did document the bias where it hurts: in countries under exceptional-access rescue programs — which concentrated more than 80% of the dollars disbursed — forecasts were systematically optimistic. When the lender and the forecaster are the same house, the number tends to help the program.³

The WEO's record, measured

At the turning points, the result is emphatic. An IMF working paper — An, Jalles and Loungani (2018) — examined 153 recessions across 63 countries between 1992 and 2014 (1,306 country-year observations; recession occurs in 12% of them). In April of the preceding year, the consensus of forecasters expected output to fall in only 5 of the 153 cases. By October of the recession year itself — with the year nearly over — 35 were still unforeseen. And the magnitude is missed "by a wide margin until the year is almost over": on average, forecasters started from ~3% growth in advanced economies (~4.5% in emerging ones) for years that ended at −2% and −3%. Do official forecasters do worse than private ones? No: their errors are "virtually identical", with a correlation above 0.9. Prakash Loungani had already written it in 2001, examining the 1990s: "the record of failure to predict recessions is virtually unblemished."⁴

One nuance almost nobody quotes: booms are missed too. In December of a boom year, forecasts still sit ~1.5 points below actual growth. If it were only about incentives (nobody wants to announce recessions), booms would be overestimated. Missing both tails points to something deeper: limited information, and models that converge to the mean.

Two episodes tell it better than any statistic.

2008-09, the error in slow motion. In April 2008 — with Bear Stearns already rescued — the WEO projected 3.8% global growth for 2009. Then came the staircase down: 3.9% in July, 3.0% in October, 2.2% in November, 0.5% in January 2009, −1.3% in April. The final figure was −0.6% (and −2.0% at market exchange rates). Four and a half points of error at an eighteen-month horizon, corrected at roughly a point per quarter. It is not that nobody saw the recession: it was announced in instalments.⁵

The WEO's revision path for 2009: four and a half points of error, corrected in instalments. FGA illustration on IMF data.

2020, the fastest reversal in WEO history. In January 2020 it projected +3.3% for the year; in April, −3.0%. Six and a half points of revision in less than three months. The final figure: −3.1%. Here the forecast corrected fast — because the shock was visible to the naked eye.⁶

The bias has a known shape, and it is quantified. Timmermann (2007), in the external evaluation the IMF itself commissioned, found systematic overprediction of next-year growth (0.36-0.55 points in advanced economies) and underprediction of inflation. The 2004-2017 evaluation (Celasun et al., 2021) pinned it down by horizon and by group: the typical error (RMSE) in advanced economies runs from ~0.7 points (autumn forecast for the current year) to ~1.8 points (year ahead), and is larger in emerging and commodity economies — up to ~3.7 points at 3-5 years for fuel exporters. At short horizons the bias is small and accuracy comparable to Consensus Economics; but at 3-5 years the median error is −0.5 points, a quarter of countries accumulate optimism above one full point, and in up to half of countries the WEO loses to a naïve forecast (the historical average). And one source is identified (Ismail et al., 2020): planned fiscal adjustments. Each additional year of horizon adds ~0.2 points of optimism; over five years, a full point. The plan needs growth to cooperate, and the forecast grants it.⁷

The WEO's record according to academic and IMF self-evaluations. FGA elaboration.

Forecasters also ration their bad news. That is Nordhaus's (1987) finding, confirmed again and again in these papers: today's revisions predict tomorrow's — if a forecaster cuts, the odds are it will cut again. It is the efficiency test that the WEO and the private consensus fail alike (the stickiness only disappears inside the recession, when it is already late). In Nordhaus's words: "we break the good or bad news to ourselves slowly." The 2008-09 staircase was not bad luck: it is the pattern.⁸

And the most expensive error is confessed in writing. In October 2012 the IMF published, in its own flagship, the box "Are We Underestimating Short-Term Fiscal Multipliers?", which Blanchard and Leigh turned into a paper in 2013. The finding: euro-crisis forecasts assumed a fiscal multiplier of ~0.5 — each euro of cuts would cost 50 cents of GDP. Actual multipliers were "substantially above 1", possibly around 1.7. Across 26 European economies, each point of planned consolidation was associated with a full point of GDP lost relative to forecast (coefficient −1.1; R² ≈ 0.5). The consequence has a name: Greece. The 2010 program projected a GDP decline of 5.5% between 2009 and 2012; the actual fall was 17%, with unemployment at 25% where 15% had been projected. The IMF's own post-mortem (2013) calls them "notable failures". The second half must be said too: it was the IMF itself that measured, published and corrected its error. Few private forecasters do that.⁹

The confessed error: assumed versus actual multiplier, and the Greek program. FGA illustration after Blanchard-Leigh (2013) and IMF (2013).

The BIS: the one that warned — and the one that always warns

The 2003 warning is documented, not legend. At the August 2003 Jackson Hole symposium, Claudio Borio and William White — the BIS's Economic Adviser from 1995 to 2008 — presented the thesis that a liberalised financial system, combined with monetary policy that only watches near-term inflation, lets financial imbalances build up unchecked. They proposed what was then unthinkable: raising rates when credit runs hot, even with inflation quiet. Greenspan, present, defended the opposite doctrine at that same symposium — let it run and clean up afterwards.¹⁰

The BIS Annual Report of June 2007 — weeks before August blew up — raises the warning to official text. Three literal quotes: "virtually no one foresaw the Great Depression of the 1930s, or the crises which affected Japan and Southeast Asia… each downturn was preceded by a period of non-inflationary growth exuberant enough to lead many commentators to suggest that a 'new era' had arrived." On securitisations: "who now holds these risks, and can they manage them adequately? The honest answer is that we do not know." And the warning that tail events "might at some point have much higher costs than is commonly supposed." When the crisis hit, Der Spiegel titled its profile of White "The Man Nobody Wanted to Hear".¹¹

The framework behind it has numbers. Borio (2012) measured the financial cycle — the joint swing of credit and property prices — in industrialised economies: it lasts 16 years on average (nearly 20 in cycles peaking after 1998), against at most 8 for the classic business cycle. Its peaks coincide with banking crises — only three exceptions since 1985. And recessions that fall in its contraction phase are ~50% deeper.¹²

The financial cycle (16-20 years) versus the business cycle (up to 8). FGA schematic after Borio (2012).

From that literature came an operational indicator, the credit-to-GDP gap, whose early-warning power is measured: an area under the ROC curve of ~0.8 up to five years before a crisis, and the debt service ratio is even sharper near the outbreak — AUC of ~0.94 in the final quarters (Drehmann and Juselius, 2014; for calibration: a medical test like the PSA scores around 0.8). Basel III adopted it officially as the guide for the countercyclical capital buffer: gap below 2 points, zero buffer; above 10, buffer at maximum. According to the Basel Committee, that rule would have required building up capital 2-3 years in advance of every major banking crisis in its sample. The world's regulator turned the BIS thesis into law.¹³

But the flip side is documented too. In June 2011, with the recovery still fragile, the BIS Annual Report literally called for "higher policy rates" and stated that "tighter global monetary policy is needed". Advanced-economy inflation then spent years below target. Krugman demolished it in 2014 in a post titled "Stability or Sadomonetarism?": "the BIS basically just wants to raise rates, and is always looking for a reason." Martin Wolf, the same week: "bad advice from Basel's Jeremiah". The BIS defence also has literature: Schularick and Taylor (2012) showed with 140 years of data across 14 countries that "credit growth is a powerful predictor of financial crises" — ignore it at your peril. But the criticism stands: the BIS gets the crises right, not the calendars. It always warns; sometimes the world takes a decade to prove it right, and sometimes it never does.¹⁴

The statistical lesson: two loss functions

None of the above is a scandal. It is elementary statistics, and it is the part of this piece you can reuse for life. Bring back the two figures from the beginning: the weatherman and the building inspector.

The base rate rules. In advanced economies, recessions occupy roughly 10% of quarters (122 recessions in 21 countries between 1960 and 2007; 12% of country-years in the An-Jalles-Loungani sample). A weatherman who never announces a storm is right 90% of the days. That is the trap against which everything else is measured.¹⁵

The level forecaster (WEO) minimises average error. With a 10% base rate, that pushes it mathematically toward the centre of the distribution: projecting the normal year is almost always the lowest-error bet. Its characteristic failure is type II — missing the rare event. Add the institutional incentive (the forecaster is also lender and adviser) and Nordhaus's rationing, and the 2008 staircase stops being a mystery.

The early-warning system (BIS) optimises the opposite: never missing the crisis, accepting false alarms. A serious inspector would rather close a sound building than leave a cracked one open. The classic calibration of the credit gap requires capturing at least two thirds of crises, then minimising the noise. Its characteristic failure is type I — the chronic "too early". Not a BIS defect: its mandate.

For the technical reader — two notes.Bayes and the base rate. With crises in 10% of years, an excellent indicator — one that catches 80% of crises with only 20% false alarms — is right ~3 times out of 10 when it sounds: 0.8×0.1 / (0.8×0.1 + 0.2×0.9) ≈ 0.31. Cutting false alarms to 10% lifts that to ~1 in 2. That is why no serious early-warning system promises certainty: it promises odds better than the base rate. The loss function. The literature (Alessi and Detken, 2009) formalises the regulator's choice as L = θ·(missed crises) + (1−θ)·(false alarms), where θ measures how much each type of error hurts. The WEO operates, de facto, with a low θ; the BIS, with a high one. Everything else — the optimism of one, the alarmism of the other — follows from there.¹⁶

Comparing their raw "accuracy" is therefore a category error. The right question: the WEO answers "where will the economy be if nothing unusual happens?"; the BIS answers "where would it hurt if something does?"

Two trades, two failure modes: the forecaster and the watchman. FGA elaboration.

How we put them to work — with a methodology of our own

Two things must be kept apart here, because they are often confused. Our methodology does not come from these reports. The four layers through which we read the cycle — cycle and credit, crowd, liquidity, real economy — are built on FGA's proprietary series, developed and tested over years and documented in a public working paper. The BIS and the WEO are not our method: they are two context inputs that are valuable precisely because their record — strengths and weaknesses — has been measured, as we have just seen. Once the weakness is known, you sidestep it; once the strength is known, you exploit it.

• From the WEO we exploit its strength: the best map of the central scenario at short horizons and an unbeatable database. We use it as context for the real-economy layer. And we sidestep its measured weakness: we never ask it to warn us of a turn — it is proven that it will not do so in time. Watching the turn is the job of our own leading indicators, built for exactly that.

• From the BIS we exploit its strength: the most serious inventory in existence of credit and liquidity fragilities. We use it as a cross-check on our cycle and liquidity layers. And we sidestep its measured weakness: it is not a calendar. A BIS warning does not by itself change the colour of any layer; our series change it, when the data confirms.

• The regime — the name we give to the moment of the cycle — comes from the cross-reading of the four layers, not from any external document. That is why our verdict can agree with the BIS one month and disagree the next: the methodology is ours; they are two first-class cross-checks.

In one sentence: we don't predict, we classify — and we use the world's best forecaster and the world's best inspector knowing exactly how each one fails.

Summer 2026 is the live example. July's WEO describes a cycle that holds (3.0% global) standing on a single technological leg; June's BIS report describes exactly that leg as the vulnerability — AI capex financed with debt and circular financing (we analysed it in our reading of the BIS report). They do not contradict each other: each is doing its job. Our framework read them together in the BIS–IMF diptych — and no layer changed colour because of it.

In this series: our reading of the BIS report (30 June) · the BIS–IMF diptych (14 July, in Spanish) · this study.

In 30 seconds

• The WEO (IMF, public since 1980) is the weather forecast: it projects levels for 196 economies. At short horizons, as good as the best private consensus. At the turns, it is not: 5 of 153 recessions anticipated.

• The BIS report (since 1931) is the building inspector: it does not forecast, it inventories fragilities. It warned of 2008 from 2003 — and called for hikes in 2011 with inflation below target. Alarmist by design.

• Opposite errors, by construction: the WEO fails type II (missing the turn); the BIS type I (over-warning). With recession only 10% of the time, both arithmetics are inevitable.

• Our rule: a proprietary four-layer methodology on our own series; the WEO as central-scenario context, the BIS as a fragility cross-check — and the turn, watched with indicators built to see it. Neither one is asked for dates.

Quick user's manual

• WEO: April and October (full editions), January and July (updates). Free at imf.org/weo. Reading order: chapter 1, the projections table, and the analytical chapters — where the IMF actually thinks.

• BIS: Annual Report in late June; Quarterly Review in March, June, September and December. Free at bis.org. Reading order: the concluding chapter first — the house always writes it as a diagnosis.

Notes and sources

1. Boughton, J., "Keeping Score: The World Economic Outlook", official IMF history, ch. 5 · all WEO editions since 1980.

2. BIS, official history: foundation · World War II and Bretton Woods · member central banks · Basel Committee history · first Annual Report (1931).

3. IMF IEO (2014), "IMF Forecasts: Process, Quality, and Country Perspectives".

4. An, Z., J. Jalles and P. Loungani (2018), "How Well Do Economists Forecast Recessions?", IMF WP/18/39 · Loungani (2001), "How accurate are private sector forecasts?", International Journal of Forecasting 17(3).

5. IMF, WEO April 2008, ch. 1 · January 2009 update · final figure: WEO October 2010, tables.

6. IMF, WEO Update January 2020 · WEO April 2020 · final figure: WEO October 2021, ch. 1.

7. Timmermann, A. (2007), "An Evaluation of the World Economic Outlook Forecasts", IMF Staff Papers 54(1) · Celasun, Lee, Mrkaic and Timmermann (2021), IMF WP/21/216 · Ismail, Perrelli and Yang (2020), IMF WP/20/229.

8. Nordhaus, W. (1987), "Forecasting Efficiency: Concepts and Applications", Review of Economics and Statistics 69(4).

9. Blanchard, O. and D. Leigh (2013), "Growth Forecast Errors and Fiscal Multipliers", IMF WP/13/1 (also in American Economic Review 103(3)) · the original box: WEO October 2012, Box 1.1 · Greece: IMF Country Report 13/156 (2013).

10. Borio, C. and W. White (2004), "Whither Monetary and Financial Stability?", BIS Working Paper 147 — the paper presented at Jackson Hole, August 2003.

11. BIS, 77th Annual Report (June 2007), ch. VIII · White (2006), "Is price stability enough?", BIS Working Paper 205 · Der Spiegel (2009), "The Man Nobody Wanted to Hear".

12. Borio, C. (2012), "The financial cycle and macroeconomics: what have we learnt?", BIS Working Paper 395.

13. Drehmann, M. and M. Juselius (2014), "Evaluating early warning indicators of banking crises", BIS Working Paper 421 · Borio and Drehmann (2009), BIS Quarterly Review, March · Basel Committee (2010), countercyclical capital buffer guidance.

14. BIS, 81st Annual Report (June 2011), ch. IV · Krugman, P. (2014), "Stability or Sadomonetarism?", NYT · Wolf, M. (2014), "Bad advice from Basel's Jeremiah", Financial Times · Schularick, M. and A. Taylor (2012), "Credit Booms Gone Bust", American Economic Review 102(2).

15. Claessens, S. and M.A. Kose, "Recession: When Bad Times Prevail", IMF Finance & Development.

16. Alessi, L. and C. Detken (2009), "'Real time' early warning indicators for costly asset price boom/bust cycles", ECB Working Paper 1039.

Our own methodology: FGA Research, four-layer framework working paper, DOI 10.5281/zenodo.20709770.

Editorial and educational content, not investment advice. No buy/sell signals, no price targets.

Independence. Capital. Conviction. · FGA Research & Advisory · Est. 2006 · 33 años de estudio

The question almost every business owner asks today isn’t “which AI do I buy?” It’s “where do I start, and whom do I trust so I don’t waste a fortune on hype?” The answer starts with the data, not the technology.

Artificial intelligence is the next industrial revolution. And like every industrial revolution, it rewards not who spends most on machines, but who decides well what to automate, and in what order. At FGA we have worked with AI since 2001 —we founded one of the first Spanish firms doing AI-based analysis— and one thing hasn’t changed in 25 years: the technology is the easy part. The decision is the hard part.

The problem isn’t AI. It’s the command room.

Almost every SME we work with arrives with the same sentence: “we have all the information; what we don’t have is a way to decide with it.” Thirty years of sales records in folders, the business intelligence living in the director’s head, and every opportunity depending on someone remembering in time. It’s not a software problem. It’s a method problem.

A real case, anonymised

A Spanish B2B supply company: three decades in the market, tens of thousands of SKUs, a book built on craft. Healthy, profitable —and with an invisible ceiling. Before proposing anything, we did what an auditor does: digitise the full sales archive and verify it document by document. What surfaced wasn’t visible from the inside:

• Nine out of ten inbound contacts had no scheduled next step.

• Six out of ten lost customers never received a proposal after their last purchase. They didn’t leave: they stopped being called.

• A contract renewed for seven years with “the same” had lost almost half its price.

None of these findings was an AI problem. They were focus and process problems — the ones AI amplifies if you don’t fix them first.

What changed (and what didn’t)

The first phase wasn’t technology: it was repositioning the business. A clear value proposition; digital presence that shows up when the client searches —also when they ask an AI—; a sales process with cadence where every contact has an owner, a date and a next step; and the archive digitised and verified as raw material. Only then did we switch on the AI (our Cortex engine) over its own archive.

Same team, same product, same market. In one quarter: 2.4× the deals signed, 3.3× the value signed, +23% average ticket with no discounts, 38% conversion —a series record. (Illustrative and anonymised case; results not extrapolable nor guaranteed.)

Order matters: data first, AI after

The lesson for a board is uncomfortable and liberating at once: AI doesn’t fix a business; it multiplies the one you already have. If the process is broken, AI multiplies the mess. If the archive isn’t verified, AI learns from dirty data. That’s why we audit before we automate — the same reasonableness principle we apply to macro, applied to your company: we don’t believe the story, we verify the number.

In 30 seconds

🔴 AI is a board-level decision, not a technology purchase.
🟡 Data first (audited), then process, and only then AI.
🟢 Well ordered, the effect multiplies; badly ordered, it amplifies the problem.

FGA Advisory guides boards and management teams through that decision: an AI implementation strategy with method and evidence, from the “AI in the Boardroom” white paper to the real case. Demonstrated vision since 2001, not hype opportunism.

Talk to FGA Advisory

Informational and commercial content. Illustrative and anonymised case; results not extrapolable nor guaranteed. FGA Research & Advisory · Est. 2006.

Independence. Capital. Conviction. · FGA Research & Advisory · Est. 2006

The reasonableness principle applied to the economy is FGA’s own procedure: treat every macro data point the way an auditor treats an account. Don’t believe the story, verify the number, and cross-check sources until the whole is reasonable. Francisco Salvador has called it that for years — and it is what separates auditing the economy from opining on it.

Most macro analysis in circulation is well-written opinion: someone reads a headline, frames it in a narrative and publishes. It sounds convincing. But a narrative is not a verified fact, and in markets that difference costs money.

It wasn’t born in statistics. It was born in auditing.

FGA’s process didn’t start in an econometric model. It started in auditing — Arthur Andersen —, where the first lesson is uncomfortable: the auditor doesn’t believe what he’s told; he verifies the number. No matter how coherent management’s explanation sounds: if the figure doesn’t reconcile with its source and its history, it doesn’t get in.

Carrying that discipline into macroeconomics is the reasonableness principle. We don’t ask "which narrative fits today?"; we ask "is this data reasonable in light of everything else?". It’s a boring, slow, distrustful way of looking — and that’s why it works.

Auditing macro demands what almost nobody has

The reasonableness principle is not a slogan: it sets a very high bar. To truly verify and cross-check you need four things that can’t be improvised:

Exceptional command of series and data. Auditing means comparing every series against its source and its history. That ability to handle large volumes of economic series placed Francisco, according to Bloomberg, at the level of the major international consultancies in Spain. It’s not showing off data: it’s the precondition for being able to verify it.

Deep knowledge of markets and their institutions. A macro figure can’t be understood without knowing how a central bank, a trading desk or a credit cycle work from the inside. Two and a half decades in the sector — banking, asset management, brokerage — give the context a model can’t see.

Proven multisector experience. Decades analysing a huge variety of sectors and companies teach you to separate the structural problem from cyclical noise. The same eye that audits a company audits an economy.

And from all of it comes the only thing that matters: the capacity to anticipate. When you verify the leading indicators instead of reacting to the coincident ones, the turn is visible before it becomes a headline. It’s not predicting the future — it’s classifying the present with enough rigor that the future holds fewer surprises.

From reasonableness to the four layers

FGA’s four-layer framework —the Warning, the Crowd, Money/the Fed and FGA Macro— is simply the reasonableness principle turned into a process: four independent readings crossed until the whole is coherent. None decides alone. The regime only changes when the weight of evidence makes it reasonable.

The best part: this now reaches the retail investor

For years, this rigor —auditing macro, mastering the series, anticipating the turn— was a service for institutional investors. What changes with FGA Macro is access: the same process, the same discipline, published openly every week so the individual investor can also read the cycle with institutional-grade judgment. Free, in Spanish and English.

In 30 seconds

🟢 Reasonableness principle = audit the economy, don’t opine on it.
🟡 It demands rare command of series, inside knowledge of markets, and decades across sectors.
🔴 Its fruit is anticipation — and now, via Substack, it’s within the retail investor’s reach.

Every week we classify the market regime with this rigor and publish it —dated, before the outcome is known— in FGA Macro.

Read the cycle with institutional-grade judgment — free

The full framework is in The FGA Method; and the firm, with its two tracks —Research and AI & Advisory—, at fgaresearch.com.

Informational and educational content. Not financial advice, not a buy or sell recommendation, and containing no signals or price targets. FGA Research & Advisory. Francisco Salvador is an agent of Miraltabank, an entity regulated by the Bank of Spain.

Independence. Capital. Conviction. · FGA Research & Advisory · Est. 2006 · 33 years of study

"Will the market crash?" has no useful answer, because it is a prediction. The question you can actually answer is different: is the cycle damaged inside? That one separates a drop that reverses from one that starts something — and, for protecting capital, it is the only one that matters.

It is the most repeated search every time the market shakes: "will the market crash?". And the honest answer from anyone —bank, asset manager or analyst— is always the same: nobody knows. Tomorrow’s prices aren’t written. Asking it is asking someone to guess.

But there is a second question, far less glamorous, that does have an answer. And it changes everything.

Predicting is guessing. Classifying is diagnosing.

Predicting means betting on a specific future —"it will fall in September"— and waiting to see if you’re right. You depend on luck. Classifying is different: diagnosing the state of the present with fixed rules, like a doctor who doesn’t guess when you’ll fall ill but can tell whether you’re healthy today. You depend on discipline, not fortune.

A market regime is that diagnosis: the state of the market, not its next move. And unlike a prediction, it can be verified without waiting for the future.

Why the same drop means opposite things

A 5% drop with a healthy cycle is an opportunity: the market breathes and moves on. The same 5% drop when the cycle is damaged inside can be the first step of a structural bear market. The number is identical. What changes is the terrain.

That’s why "will it crash?" is the wrong question: even if you got the "yes" right, you wouldn’t know if it’s noise or a beginning. The useful question is not whether it falls, but what state the market is in when it does.

How it’s answered: four layers

At FGA we don’t guess: we classify the regime by crossing four independent readings. None decides alone; the regime only changes when two or more turn at once.

The Warning checks whether the cycle is healthy or damaged inside (macro, corporate, markets). The Crowd reads sentiment contrarian: bottoms are made on fear. Money · the Fed measures whether the price of money pushes or brakes. FGA Macro confirms where in the cycle we are. Together they answer the right question.

In 30 seconds

🔴 "Will the market crash?" = prediction. Nobody knows.
🟡 "Is the cycle damaged inside?" = diagnosis. This one has an answer.
🟢 The same drop is noise or a chance to break depending on the regime. That’s why the diagnosis matters more than the prediction.

Every week we classify the market regime with the four layers and publish it —dated, before the outcome is known— in FGA Macro. Free, in Spanish and English.

Subscribe to FGA Macro

The full framework —the four layers, the Warning and the traffic light— is in The FGA Method. And what a regime is exactly, in this piece.

Informational and educational content: dissemination of an analytical framework. Not financial advice, not a buy or sell recommendation, and containing no signals or price targets. FGA Research & Advisory. Francisco Salvador is an agent of Miraltabank, an entity regulated by the Bank of Spain.

Independence. Capital. Conviction. · FGA Research & Advisory · Est. 2006 · 33 years of study

A market regime is the name of the cycle phase: it defines how much risk the market tolerates — not what to buy. Knowing which regime you are in matters more than calling the next move, because the regime decides whether that move is noise or the start of something.

Almost everyone invests by asking the wrong question. "What do I buy?" "Will Nvidia go up?" "Is now a good time to get in?" These are questions about the next step. But the next step depends on the terrain — and the terrain has a name: the regime.

The word that changes how you see the market

Think of the sea. A good swimmer doesn't ask "should I go in or not?". They ask "how is the sea today?". In calm water a bad move is forgiven. In a rip current the same move drowns you. The swimmer doesn't guess the next wave: they classify the state of the sea and adjust their behaviour.

A market regime is exactly that: the state of the sea. It is not a forecast ("it will rise tomorrow"); it is a diagnosis of the present ("the market is in a state that tolerates little risk"). And unlike a forecast, a diagnosis can be verified without waiting for the future.

Why it matters more than calling the next move

Because the same event means opposite things depending on the regime. A 5% drop in a healthy regime is an opportunity; the same 5% drop when the cycle is damaged inside can be the first step of a structural bear market. The number is identical. What changes is the terrain.

The investor who chases moves lives reacting to every headline. The one who reads the regime knows, before the headline, how much it weighs. That is the whole difference between panicking and staying calm with a reason.

How a regime is classified: four layers

At FGA we don't guess the regime: we classify it, crossing four independent readings. None decides alone; the regime only changes when two or more turn at once.

The Warning — protects. It separates noise (corrections that reverse) from the start of a structural bear market, watching three dimensions: macro, corporate and financial markets. Answers: is the cycle healthy or damaged inside?

The Crowd — times. Sentiment read contrarian: bottoms are made on fear, tops on euphoria. Answers: euphoria or fear?

Money · the Fed — triggers. The price of money is the lid that pushes or brakes; its turn has marked the great bottoms. Answers: tailwind or headwind?

FGA Macro — confirms. Leading versus coincident indicators; the turn forms before it is visible. Answers: where in the cycle are we?

Predict vs. classify

Here is the sentence that sums up how we work: "We don't predict. We classify." — markets are not to be guessed; they are to be classified.

Predicting means betting on a specific future and waiting to see if you're right. Classifying means diagnosing the present with fixed rules and adjusting behaviour to that diagnosis. The first makes you dependent on luck; the second, on discipline. A regime doesn't tell you what happens tomorrow. It tells you how much it will hurt if you're wrong today — and, for protecting capital, that is worth more.

In 30 seconds

🟢 Regime = the state of the sea, not the next wave.
🟡 It is classified with rules (four layers), not guessed.
🔴 The same event means opposite things depending on the regime: that's why it matters more than the move.

This is a term from the FGA lexicon. For the full framework — the four layers, the Warning, the traffic light — with the canonical definitions, see our method page.

Every week we classify the market regime with the four layers and publish it —dated, before the outcome is known— in FGA Macro. Free, in Spanish and English.

Subscribe to FGA Macro

Want the full framework? The four layers, the Warning and the traffic light, with the canonical definitions, are in The FGA Method.

Informational and educational content: dissemination of an analytical framework. Not financial advice, not a buy or sell recommendation, and containing no signals or price targets. FGA Research & Advisory. Francisco Salvador is an agent of Miraltabank, an entity regulated by the Bank of Spain.

Independence. Capital. Conviction. · FGA Research & Advisory · Est. 2006 · 33 years of study

This week's verdict. The regime is still Compressed Spring — unchanged. The news is inside the layers: the crowd's fear, which had been loading the spring, is starting to ease. No euphoria (which would mark a top), no macro damage (which would mark a turn).

The S&P 500 closed the week near highs again, brushing 7,480 (about +19% over twelve months). Once again, price isn't the story. The story is in Layer 2: the fear we've been flagging for weeks has begun to ease. Our sentiment gauge is coming off its cycle lows. Careful, though — this isn't optimism arriving; caution still rules — but the spring that had been pinned to the floor for weeks is starting to release.

And last week's other spring still holds: in Layer 4, the leading indicator stays above its neutral level. It's no longer turning; it's expanding.

Regime: Compressed Spring. The cycle holds, the leading indicator confirms the expansion; fear starts to ease and monetary policy is still the lid.

In 30 seconds

Regime: Compressed Spring — unchanged (it only changes if two or more layers turn).

• Layer 1 · Warning — 🟢 out of the alert zone; no structural damage.

• Layer 2 · Crowd — 🟡 this week's change: fear starts to ease, still short of optimism.

• Layer 3 · Money (the Fed) — 🟡 monetary policy is still the lid; no Fed meeting this week.

• Layer 4 · FGA Macro — 🟢 the leading indicator holds in expansion, above neutral.

  The so-what: only one layer moves — fear eases — so the regime doesn't change. The spring stays loaded, but with one contrarian support less. No euphoria (which would mark a top), no macro damage (which would mark a turn).

  This is how we read the cycle every week: four layers, made clear. Get it free in your inbox.

  Subscribe now

  ---

  This week's composite

  (The four-layer composite and the scenario map with probabilities are in the report gallery, at the foot of this edition.)

  Regime in force: Compressed Spring. The regime only changes when two or more layers turn, and this week only one moves — sentiment. So the name doesn't change. What changes is the internal tension: one of the springs loading the move higher — the crowd's fear — is starting to ease.

  ---

  The four layers, in detail

  Layer 1 · Warning — 🟢 Green. The Warning Signal remains clearly out of the alert zone. Worth remembering where the real danger sits: the crisis lows were −2.9 (2008) and −2.8 (2020), and we're nowhere near them. It's easing slowly from the January peak, yes, but that's losing a little momentum, not breaking. As long as this layer stays green, this is a reset with the spring loaded, not a turn lower.

  Layer 2 · Crowd — 🟡 Amber, and here's the news. For two editions we've told the "wall of worry": price at highs, a frightened crowd. Last week the first crack appeared — the professionals (very bullish advisors, active managers almost fully invested) had already gone long while retail stayed out. This week that crack widens and sentiment leaves its cycle lows. Now, careful with the word. Fear eases; it hasn't given way to optimism. We're still in caution, far from the complacency that builds tops. And the honest read is a touch uncomfortable: one of the contrarian supports loading the spring has been trimmed. Less fear, less cushion. It's not a top signal — it's the signal to start watching the other extreme.

  Layer 3 · Money (the Fed) — 🟡 Amber, drifting lower. Here Layer 3 measures one specific thing: the monetary policy of central banks, above all the Fed. Not market conditions, but whether money blows with you or against you. And for now it blows against, gently. Warsh's Fed keeps the June bias, and there was no meeting this week, so no news. It's not restrictive yet; it's less accommodative. It's the lid that keeps the spring compressed. What we watch isn't what the Fed has already done, but what the curve starts to price: that forward-looking clock moves first.

  Layer 4 · FGA Macro — 🟢 Green, rising. No surprises, and that's good. The leading gauge bottomed in mid-2025 and crossed its neutral level this spring; it's still above, in expansion. The coincident still lags, a little soft. Leading ahead and coincident lagging is the signature of an early expansion, not of a cycle running out. Recessions are announced the other way round: when the leading gauge collapses with the lagging stretched. That's absent today.

  ---

  The verdict

  Regime: Compressed Spring — with the fear spring easing. The pieces still fit. The cycle isn't broken, the leading indicator is expanding, monetary policy presses from above. The only thing moving this week is the crowd. Nothing more — but not nothing. That fear, which for weeks worked in our favour in contrarian terms, is starting to give — without yet making the jump to open optimism. The spring keeps pushing and has rope left: no euphoria (which would mark a top), no macro damage (which would mark a turn). As long as Layer 1 stays green, this is still a loaded reset.

  Scenario map (model probabilities)

  (The four-layer composite and the scenario map with probabilities are in the report gallery, at the foot of this edition.)

  This is a regime read and its probabilities, not a recommendation or a price target.

  ---

  Analogues & Bubble Watch

  The pattern "intact cycle + tightening central bank + leading gauge in expansion + fear starting to ease" has useful precedents in 1994–95 (the Fed hikes, the market digests it without a recession) and 1998 (a sharp scare inside a live cycle, distinct from 1999, already late). Illustrative analogues; every cycle is different.

  In the Bubble Watch, valuation still ticks ✓ —the CAPE at a historic extreme— and index concentration; they do not tick cycle, credit or the curve. Fuel yes, spark no. What's changed versus a month ago? With fear easing, that contrarian support weighs less. And that forces us to look at the other end of the board —complacency—, which until now we weren't even touching.

  ---

  What we're watching this week

  1. Warning Signal (Layer 1): if it speeds its fall toward zero, that would be the first real warning of a phase change (→ First Leg Down).

  2. Sentiment (Layer 2): now that fear is starting to ease, the question switches sides. Does it settle into caution, or accelerate toward the optimism and euphoria that build tops? That's what's new.

  3. Coincident (Layer 4): if it starts to rise, confirming the leading gauge, it validates the early expansion — the piece that matters most to us.

  4. The yield curve (Layer 3, forward-looking): the first hint that the Fed will loosen the lid shows up in the curve before it shows up in rates.

  We flag the day the regime changes — that's the moment of maximum value.

  ---

  The proof, in the open

  The methodology and the 33-year backtest are published and verifiable: Evidence — the proof is open →

  How is this read built, layer by layer? It's in the Edition Zero →; and last week, in The leading indicator confirms the turn →.

  And this week, our read of the BIS report on extreme valuation: AI bubble? What the BIS warns →.

  ---

  You don't need to predict the top. You need to know which phase of the cycle you're in — and to be flagged the day it changes. It works for a portfolio and it works for a business: the cycle regime is the same backdrop when you decide when to invest, expand, hire, take on inventory or finance.

  Subscribe free and get the cycle read every Sunday, before the open. Early subscribers get founder terms.

  Subscribe now

  ---

  Reading this from an investment committee?

  FGA Research distributes, on demand and subject to prior qualification, its Institutional Macro & Markets Report: the same four-layer process, with the full proprietary series and watchlist. If you run capital, manage risk or build strategy for a committee, write to us stating your role and mandate. We confirm eligibility and send the full report within 24h.

  Request the institutional report →

  ---

  Information, not advice. Independence. Capital. Conviction.

  ---

  About the author. Francisco Salvador, founder of FGA Research. The four-layer process has served the institutional sector for years; here we open it to the non-institutional reader —investors and business owners who decide with the cycle in front of them. I comment on the cycle every week on LinkedIn → Francisco Salvador on LinkedIn

  ---

  Notice. Informational and educational content: dissemination of proprietary macro indicators and cycle analysis, not a regulated investment service. It is not advice or a recommendation —general or personalised— nor buy/sell signals. Market data from public sources; proprietary indicators of FGA Research; past results —real or simulated— do not guarantee future results; investing carries risk, including total loss of capital.

  Independence. Capital. Conviction. · FGA Research & Advisory · Est. 2006 · 33 years of study

📄 The full weekly report — page by page

• It is not "revolution or bubble": it is both. Like the railway or the internet, AI is a real transformation and, at once, an episode with a bubble profile.

• What decides the outcome is productivity, not the technology. Task-level gains are real (20–50%); the aggregate kind —the one that pays for the capex— is still below 1%.

• The main risk is not valuations: it is the financing —with debt, opaque and circular.

• The BIS warns, it does not predict. And our 4-layer model places the regime today in a compressed spring: stored tension under a lid, no damage yet.

This is how we read the cycle every week: four layers, made clear. Get it free in your inbox.

The BIS report, in brief

On 28 June the BIS —the central bank of central banks— published its Annual Economic Report; the chapter is titled "Progress and peril". In one line: AI has sustained growth, but its financing is creating risks. What the BIS states, plainly:

• The five largest hyperscalers will spend over a trillion dollars on AI capex in 2025–26, beyond their earnings and free cash flow; some are already issuing debt.

• AI-specific investment is around 1% of GDP; broad IT investment, about 5% —above the dot-com peak of 2000.

• Task-level efficiency rises 20–50%, but aggregate productivity is estimated below 1% over the long run.

• The BIS recalls the precedents —canal (1830s), railway (1840s), electrification (1920s), dot-com (90s)— and places the current multiple trajectory as the steepest of all.

The BIS’s own conclusion is nuanced: it does not declare a bubble; it warns of a risk conditional on very high expectations being met. So much for the X-ray. From here, our reading.

Our reading: why revolution and bubble coexist

The framing error is to pose it as a dilemma: either a revolution or a bubble. Economic history says it is almost always both, in this order.

Kindleberger and Minsky described the anatomy of every bubble decades ago: a displacement (a real novelty) attracts capital; capital leverages up; price decouples from fundamentals; tension builds; and, finally, the reversal. Minsky’s mechanics are exactly what the BIS describes today: agents move from funding with their own cash to funding with debt.

Carlota Pérez put it clearly: every technological revolution has an installation phase —when financial capital builds the infrastructure, with frenzy and bubbles— and a deployment phase, when the technology permeates the real economy. Between them there is usually a collapse. The bubble does not contradict the revolution: it is how capital finances, in excess, its own infrastructure.

The railway is the best mirror. The network was built and transformed the world for decades. And even so, most first-wave investors lost money, because capital was destroyed before returns matured. Being right about the technology and losing money on this vintage of capital are two different things.

The four layers: from an unanswerable question to four with answers

Here is our value. The BIS gives the global X-ray once a year. We read the cycle layer by layer, every week. And the BIS report, without meaning to, feeds our four layers.

Layer 1 · Warning. Is the system in the alert zone or with room? The BIS describes vulnerabilities present —circular financing, off-balance-sheet risk— but not triggered. There is latent tension, not damage.

Layer 2 · Crowd. Euphoria or fear? Implied valuations are well above their historical benchmarks and risk premia are compressed. But euphoria is not uniform: you can buy the index and fear it at once. That divergence is information, not noise.

Layer 3 · Money (the Fed). Is money easing or tightening? It is the decisive layer. Bubbles burst when credit gets dearer, not when the technology disappoints. The BIS puts it another way: there are two independent triggers —a rate hike to contain inflation, or a disappointment in AI earnings— and both pass through the same financial channel.

Layer 4 · Macro. Is the underlying economy accelerating or softening? Here it connects with productivity: while the aggregate stays below 1%, capex rests on a promise, not a result.

The four layers turn an unanswerable question —"is it a bubble?"— into four questions with an observable answer. They do not predict the date. They locate the regime.

And it bears saying without false modesty: our model is not theory. Over the 33-year study, the four big structural declines —2000, 2008, 2020 and 2022— were characterised ahead of time, and the two corrections that led nowhere —1998 and Q4 2018— were filtered out without taking anyone out of the market falsely.

The real focus of risk: the financing

If you remember only one idea from the report, let it be this: the problem is not valuations, it is the financial plumbing.

The BIS documents it. Direct lending funds have quadrupled their lending to AI and IT in five years, to about 15% of their portfolios; the outstanding private credit to AI firms already exceeds 200 billion dollars. Loans to AI are larger than average, yet with collateral, tenor and spreads similar to the rest: either lenders underprice the risk, or the equity market overvalues the cash flows.

To this add three traits that amplify any shock. Circular financing: chip makers take equity stakes in labs that, in turn, commit to buying their compute. Off-balance-sheet risk: long-dated, poorly disclosed contracts where the same asset can be pledged several times. And the wealth effect: with households more exposed to equities than ever and the US at about 64% of the global index, a US correction does not stay at home.

The BIS sums it up in a line worth keeping: "leverage does not disappear by being out of sight."

The decisive variable and the shape of the risk

It all reduces to one measurable variable: whether, in the next 12 to 24 months, aggregate productivity and AI earnings converge toward what the market already discounts, or disappoint.

What matters is the shape of the risk. If AI delivers, the cycle continues in an orderly way. If it disappoints, the adjustment is non-linear: deleveraging feeds on itself. That is why this is a tail risk, not the central case. The upside is gradual; the downside, abrupt. That asymmetry is the message.

The bull case (because it must be made too)

It would be dishonest to close only on the dark side. Unlike the dot-com, here there are real and growing revenues; the big tech firms are profitable and still fund most of it with cash. The cost of compute falls fast, which can validate the capex with less revenue than expected. Lagging productivity is not the same as absent: general-purpose technologies show a delay —the J-curve— before the aggregate jump. And the epicentre of fragility is confined to the periphery, not the profitable core.

The revolution is real. The question is not whether AI changes the world. It is whether this particular vintage of capital recovers what was invested before credit tightens.

The panel today

Translated into our framework, and publishing only the state, not the series: the cycle, out of the alert zone; the crowd, fearful despite indices at highs; liquidity, no longer easing; and our own macro, turning up. The regime that combination describes is what we call a compressed spring: energy stored under a lid. It is not the start of a crisis. It is a system under tension, held up by liquidity.

The BIS report does not change the regime. It contextualises it: it tells us where the tension to watch sits.

Signals to watch

• Credit spreads of AI firms widening while stocks rise —the equity-debt divergence. (Layers 1 and 3.)

• Redemptions in retail private-credit funds —early signs already. (Layer 1.)

• A hyperscaler cutting its capex guidance. (Layer 4.)

• Supply bottlenecks —power, chips, grid— as an amplifier of overinvestment. (Layers 3 and 4.)

None predicts the date. All move the regime. We track them every week.

Sources

• BIS, Annual Economic Report 2026, Ch. I "Progress and peril" (28 Jun 2026): capex >1tn; task-level productivity 20–50% and aggregate <1% (Graph 10.A); historical parallels (Graph 11.C); circular financing and same asset pledged several times (Graph 13.C); ~15% of direct-lending portfolios (note 2); US ≈64% of the MSCI (Graph 12).

• BIS, Bulletin 120 "Financing the AI boom" (7 Jan 2026): AI investment ≈1% of GDP and broad IT ≈5% (Graph 4.A); private credit >200bn and ≈8% (Graph 3.A); the leverage quote.

• FGA Research, working paper (DOI 10.5281/zenodo.20709770): 33-year backtest. Results on historical data: they validate the process, not a recommendation.

Editorial content by FGA Research. It is not investment advice or a personal recommendation. It is not a recommendation on any financial instrument. References to BIS publications are provided for educational purposes; figures come from the cited sources. Past results —real or simulated— do not guarantee future results; investing carries risk, including total loss of capital. Independence. Capital. Conviction.

Reading this from an investment committee?

FGA Research distributes its institutional Macro & Markets report on demand and subject to prior qualification: the same four-layer process, with the full proprietary series and watchlist. If you allocate capital, manage risk or set strategy for a committee, write to us stating your role and mandate. We confirm eligibility and send the full report within 24h.

Request the institutional report →

Read the full BIS report →

Subscribe free · no card

This is how we read the cycle every week: four layers, made clear. Get it free in your inbox.

First, in the proprietary macro: our leading indicator crosses above its neutral level and confirms the early turn that last week was only hinted at.

Second, in sentiment: our composite still reads fear, but advisors and professional managers are turning long. The "everyone is afraid" that held the spring is starting to crack on the professional side.

The four layers: the cycle holds (green); sentiment in fear, with that first crack; liquidity — the Fed, with no meeting this week, keeps its tighter bias; and the proprietary macro, with the leading indicator now above neutral.

The model keeps the regime at compressed spring, unchanged. The cycle isn't broken, the leading indicator confirms, the base still fears, and the Fed presses from above.

This describes a state of the cycle; it is not a recommendation. It helps locate the moment for decisions of portfolio and of business: when to invest, hire or finance.

Three paths, as model probabilities, not a forecast. Base: still compressed. A loosening of the regime — a conditional scenario — if fear leaves its lows behind. Or a first leg down, if the cycle turns.

Each Sunday we read which phase of the cycle the market is in — and flag the day it changes. Editorial content, not investment advice.

The proof is open: Evidence — the proof is open

Follow Francisco on LinkedIn: linkedin.com/in/francisco-salvador

📄 The full weekly report — page by page

Reading this from an investment committee?

FGA Research distributes its institutional Macro & Markets report on demand and subject to prior qualification: the same four-layer process, with the full proprietary series and watchlist. If you allocate capital, manage risk or set strategy for a committee, write to us stating your role and mandate. We confirm eligibility and send the full report within 24h.

Request the institutional report →

Information, not advice. Independence. Capital. Conviction.

Not subscribed to FGA Macro yet? The weekly cycle read, in your inbox. Free.

This is how we read the cycle every week: four layers, made clear. Get it free in your inbox.

Only macro: no tips, no price targets. A reading of the cycle to decide better — for your portfolio or your business.

The four layers: — Cycle: out of alert. — Sentiment: fear. — Liquidity: the Fed stays tighter. — Proprietary macro: the leading indicator turns up.

We don't predict — we classify the regime, and flag the day it changes. Comprehension of the regime, before the consensus.

Editorial content, not investment advice. No targets, no buy/sell calls.

Reading this from an investment committee?

FGA Research distributes its institutional Macro & Markets report on demand and subject to prior qualification: the same four-layer process, with the full proprietary series and watchlist. If you allocate capital, manage risk or set strategy for a committee, write to us stating your role and mandate. We confirm eligibility and send the full report within 24h.

Request the institutional report →

Information, not advice. Independence. Capital. Conviction.

Not subscribed to FGA Macro yet? The weekly cycle read, in your inbox. Free.

The proof is open (working paper · DOI · forward test): Evidence — the proof is open

Follow Francisco on LinkedIn: linkedin.com/in/francisco-salvador

📄 The full weekly report — page by page

This is how we read the cycle every week: four layers, made clear. Get it free in your inbox.

That is why we publish FGA Macro. Not one more opinion: the proprietary series and audit methodology of FGA — the same process used to locate where we are in the cycle — explained clearly.

We cross four layers, and from their combination a single regime emerges:

— Warning Signal: is there structural damage? — Sentiment: how many are already in? — Monetary policy: is money blowing in your favour? — Proprietary macro: which phase of the cycle are we in?

This week's reading: a compressed spring. The cycle holds and the macro sees no recession; sentiment sits at an extreme. The market can keep rising, but with less margin.

In the 33-year study, the four big contractions were characterised in advance by the deterioration of these signals. Illustrative historical results.

We don't say what to buy or sell. We give the map: which phase of the cycle the market is in, before the consensus.

Editorial content, not investment advice.

Reading this from an investment committee?

FGA Research distributes its institutional Macro & Markets report on demand and subject to prior qualification: the same four-layer process, with the full proprietary series and watchlist. If you allocate capital, manage risk or set strategy for a committee, write to us stating your role and mandate. We confirm eligibility and send the full report within 24h.

Request the institutional report →

Information, not advice. Independence. Capital. Conviction.

Not subscribed to FGA Macro yet? The weekly cycle read, in your inbox. Free.

The proof is open — working paper (DOI), independent indexing and a live forward test: Evidence — the proof is open

Follow Francisco on LinkedIn: linkedin.com/in/francisco-salvador

This is how we read the cycle every week: four layers, made clear. Get it free in your inbox.

Read the working paper (DOI)

See the live forward test (OSF)

Download the dossier (PDF · ES/EN)

Across historical data, the system classified the six major episodes of the last 25 years correctly: it anticipated the four bear markets (2000, 2008, 2020, 2022), gave no false exits on the two big corrections that weren't (1998 and Q4-2018), and flagged the turns back up. Hypothetical results on partly reconstructed data; they do not guarantee future results.

The record supports robustness; the forward test puts it to the test on the only ground that allows no hindsight: the future. The model's construction is proprietary and not disclosed. A descriptive reading of the cycle, for informational and educational purposes: not investment advice or a recommendation (MiFID II/MAR), not personalized, and not about specific instruments. Past results, real or simulated, do not guarantee future results.

Read the weekly cycle in English — free.

Follow Francisco on LinkedIn: linkedin.com/in/francisco-salvador

Reading this from an investment committee?

FGA Research distributes its institutional Macro & Markets report on demand and subject to prior qualification: the same four-layer process, with the full proprietary series and watchlist. If you allocate capital, manage risk or set strategy for a committee, write to us stating your role and mandate. We confirm eligibility and send the full report within 24h.

Request the institutional report →

Information, not advice. Independence. Capital. Conviction.

Not subscribed to FGA Macro yet? The weekly cycle read, in your inbox. Free.