The OECD stops believing in soft landings

The forecast didn't move. The world did.

Four OECD reports, from December 2025 to September 2026. Headline GDP projections: barely touched. The narrative underneath: completely rewritten. Nine months ago the organization expected inflation to converge and AI to carry growth. Now it sees sticky prices, an AI capital-spending bubble, and trade policy as pure unpredictability.

The OECD's Economic Outlook is the closest thing the rich world has to an official mood ring — published twice a year with interim updates, read by every finance ministry and central bank on earth, written in the cautious prose of an institution that cannot afford to panic and cannot afford to lie. Which is what makes the four-report sequence so telling: the numbers barely moved, but the meaning of the numbers did. An institution that revises its story while holding its forecasts is telling you the old story stopped working before it was willing to admit it.

The numbers didn't change. The meaning of the numbers did. That is what a regime change looks like in institutional prose.

The inflation story is the pivot. "Expected convergence" became a persistent, sticky reality — the kind that locks in higher-for-longer rates. Kashkari warned that price pressure had spread from energy into services — "all aspects" of the economy. When the central bankers stop saying "transitory" and start saying "all aspects," the meeting has a different mood.

"Sticky" is the word that replaced "transitory," and the replacement took five years. In 2021, the central banks called the post-Covid price surge transitory and were wrong; in 2026, they call the persistence sticky and nobody argues. The mechanism is the one Kashkari named: inflation that starts in energy migrates into services — rents, wages, insurance, the haircut and the hospital bill — where it settles in and stops responding to interest rates. Energy prices can fall; a wage-price rhythm, once established, has to be broken. "All aspects" is central-banker for "it is in the walls now."

The Fed hiked on September 16 — unanimously, to 3.75–4.00% — with 16 of 18 officials seeing another hike this year. The bond market answered with the 10-year Treasury briefly exceeding 5% for the first time since 2007, and the 30-year hitting its highest since 2004. Markets are pricing persistent inflation the way institutions now describe it.

Unanimity is the tell. A divided Fed hikes with dissents and hedged language; a unanimous Fed, with 16 of 18 officials penciling in another hike, is an institution that has stopped debating whether inflation is beaten and started debating how much pain the cure requires. The 3.75–4.00% range is not the story — the trajectory is. Higher for longer is no longer a warning. It is the policy.

The 5% on the 10-year is the number that reprices everything. The last time the benchmark Treasury paid 5%, the iPhone was new and subprime was a rumor. At 5%, the risk-free rate becomes the gravity of the entire financial system: mortgages reset against it, corporate borrowing costs climb with it, and every risky asset on earth has to justify itself against a government bond paying five percent for doing nothing. The 30-year at its highest since 2004 tells the same story in a longer maturity: the market has stopped believing that rates come back down. Duration is no longer a bet on normalization. It is a bet on persistence.

Then there is the AI re-rating, and it is delicious. The OECD reclassified artificial intelligence from a pure growth driver to a multi-faceted risk: massive capital spending plus power-grid bottlenecks could make AI a source of financial-market stress. The secular savior is now a capex sinkhole. The data centers need power that Europe — see today's other story — does not have.

The re-rating is the OECD admitting, in institutional language, that the AI boom has a physical body. Data centers are not software — they are concrete, copper, and gigawatts, and the gigawatts have to come from grids that were not built for them. The capital spending is the financial side of the same constraint: hundreds of billions committed to build the infrastructure of the AI economy, financed — in many cases — with debt that assumed the old rate regime. An AI capex boom built on cheap-money assumptions, meeting a power grid built for a smaller economy, inside an inflation regime that keeps money expensive: that is the multi-faceted risk, and every facet is a bill coming due.

There is a circularity the OECD does not quite name. AI needs power; power needs investment; investment needs cheap money; cheap money needs low inflation; low inflation needs the economy to cool; a cool economy cannot fund the AI buildout. The snake is not only eating the trade-policy tail. It is eating its own power cord.

The economy's current composition is what one analyst called a "pincer": prosperity and inflation advancing simultaneously. The Dallas Fed's Weekly Economic Index hit 3.07% for the week ending September 12 — a third straight week above 3%, the first such run since 2022. Growth above trend, prices refusing to fall. The economy is strong enough to keep inflation alive and fragile enough to break under the cure.

The Dallas Fed's index is worth understanding, because it is the pincer measured weekly. The Weekly Economic Index distills the high-frequency data — labor, production, spending — into a single growth reading, and 3.07% means the economy is running meaningfully above its long-run trend. Three straight weeks above 3%, the first such run since 2022, is not a blip; it is a pace. And the pace is the problem: an economy growing above trend is an economy generating the demand that keeps prices sticky, which keeps rates high, which is supposed to slow the economy — except it hasn't. The pincer is the transmission mechanism failing in real time. The medicine is being administered. The patient is still running.

Trade policy is the third fracture. The US average effective tariff rate swung from 14.0% (December) to 9.9% (March) to 9.6% (June) to 10.9% (September) — a number that, the OECD says, no longer matters. The damage is not the rate. The damage is the unpredictability and the supply-chain restructuring it forces.

Read that carefully: the tariff number is "repeatedly rising and falling" and tells you nothing. Companies restructuring supply chains creates structural cost increases, which create more inflation. The snake eats its tail, and the tail is labeled "trade policy."

The OECD's dismissal of the number is the real insight. A tariff rate that swings from 14% to 9.6% and back to 10.9% in nine months is not a policy — it is weather. And companies do not optimize for weather; they insure against it, by duplicating supply chains, by nearshoring, by holding inventory, by paying the premium for optionality. Every one of those insurance payments is a structural cost increase that does not reverse when the tariff number falls again. The restructuring is the inflation. The rate is just the excuse the quarter's headlines use.

France is the European fracture line. The OAT-Bund spread hit 110 basis points — the worst since the 2012 debt crisis. Fiscal anxiety is back in Europe, wearing French clothes this time.

The spread is the market's verdict on a sovereign, expressed in the only language markets speak. French government bonds (OATs) now pay 1.1 percentage points more than German Bunds — the premium investors demand to hold France's debt instead of Germany's. The last time the gap was this wide, the euro itself was in question. France is not Greece in 2012; but the spread says the market has started pricing French fiscal politics as a risk, not a given. Fiscal anxiety, the OECD's polite phrase, is back in Europe — and it is wearing the clothes of the eurozone's second economy.

Japan is the monetary mystery. The Bank of Japan hiked to 1.25% and the yen still fell 2% into the 157s. Rate hikes aren't working the way the textbooks say. When tightening weakens your currency, the textbook is on fire.

The yen's defiance has a mechanics. Rate differentials drive currencies — except when they don't, because the market is positioned the other way, or because the hike was already priced, or because Japan's structural outflows (its insurers, its pension funds, its carry traders) overwhelm the policy signal. A central bank that hikes into a falling currency is a central bank discovering the limits of its own instrument: the textbook says tightening strengthens the currency; the market says the textbook was written for a smaller, simpler world. The fire is not Japan's alone. Every central bank watching is taking notes.

The market week told the story in miniature: Nasdaq record on Monday on Meta's AI agent announcement, bond rout by Thursday, VIX spiking 6.8% on Wednesday. Only the Magnificent Seven held — up 5% on the week while everything else wobbled. Concentration risk wearing a party hat.

The week is the regime in five trading days. Monday's Nasdaq record, on an AI product announcement, is the growth narrative — the secular savior, still selling the dream. Thursday's bond rout is the rate narrative — the sticky reality, repricing everything the dream is built on. Wednesday's VIX spike is the market noticing it believes both at once. And the Magnificent Seven, up 5% while everything else wobbled, are the concentration at the heart of it: seven stocks carrying the index, the AI capex, and the growth story, all at once. A party hat on concentration risk is still concentration risk. The week ended the way the OECD's four reports did: the numbers holding, the meaning changed.

The dark read is the honest one: the six variables the OECD says to watch are long-term yields, energy reserves, AI capex sustainability, and food prices. Four of those are outside any central bank's control. The adults in the room just lowered their voice because the room got louder.

Count what the central banks cannot touch. Long-term yields are set by the bond market's verdict, not the policy rate. Energy reserves are geology and geopolitics. AI capex sustainability is a corporate boardroom decision multiplied by thousands. Food prices are weather, war, and harvests. Four of the six variables that will decide whether the soft landing holds are outside any central bank's control — which means the institutions with the most responsibility for the outcome control the least of its determinants. The lowered voice is not calm. It is the sound of adults doing the math.

What nobody will say out loud: inflation converging was a forecast; inflation persisting is a fact. The OECD's four reports trace the exact moment an institution stopped believing its own baseline — and published the disbelief anyway.

Watch December. If the fifth report moves the GDP numbers down while the narrative stays dark, the soft landing is officially dead. If the numbers still don't move, the institution is managing expectations, not forecasting them.

December is the test the article sets for itself. A fifth report that finally cuts the GDP numbers, with the narrative still dark, is the institution admitting the regime change in the only language it trusts — the forecast. A fifth report that holds the numbers again, with the story darker still, is the institution doing something else: managing what the world expects, rather than predicting what the world does. Either way, the four reports from December to September will read, in hindsight, as the moment the prose changed before the numbers dared to. The forecast didn't move. The world did. December decides whether the forecast catches up.

Western coverage — BWFA and US wealth managers — reads the outlook as a trading regime: rates versus AI, winner undecided.

The frame: the Fed hike is necessary medicine for persistent inflation, and the market's job is to price the dosage. European press centers the French spread — the return of sovereign-debt fear as the real story behind the OECD's diplomatic language.

The Western bet is that credibility compounds: hike now, anchor expectations, and the shock stays contained.

The Western lens also prices the December test as a trade. If the fifth report cuts GDP with the narrative still dark, the soft-landing thesis reprices violently — duration rallies, credit widens, the Mag Seven premium compresses. If the numbers hold again, the market reads it as the institution choosing expectation management over forecasting, and prices the credibility discount. Either way, the West's answer to the open question — does another hike cure inflation or break something first — is that credibility is the only instrument that compounds. Hike now, anchor expectations, contain the shock. The medicine is the message.

Eastern coverage — Xinhua, Russian outlets — frames Western rate turmoil as self-inflicted damage from sanctions and energy policy. The supply shock did not fall from the sky; it is the bill for years of Western choices, now being billed to every mortgage holder on earth.

Russian commentary notes the irony with particular relish: the West's inflation fight constrains its own Ukraine spending. Tight money in Washington is loose strategy in Kyiv.

The Eastern lens also reads the AI re-rating as a Western contradiction made visible. The same economies that sanctioned their way into an energy shock are now financing a data-center buildout their grids cannot power, with debt priced for a rate regime that no longer exists. In this telling, the multi-faceted risk is not a market accident — it is the compound interest on a decade of Western choices. The bill, as ever, is presented to every mortgage holder on earth.

The South — emerging-market press, Indian and Brazilian outlets — reads higher-for-longer as capital-flow pain: expensive dollars, squeezed budgets, borrowing costs that rise because Washington sneezes.

When the 10-year Treasury pays 5% risk-free, foreign money follows the rate differential home. Central banks from São Paulo to Jakarta are forced to hike against their own weakening demand — importing the West's medicine for the West's disease. India's analysts put it plainly: every hawkish speech from Washington does the opposite of what their import bills need.

The South's lens also counts the food-price variable the hardest. Of the OECD's six variables to watch, food prices are the one that hits the South first and worst — weather, war, and harvests translated directly into the price of bread, in economies where food is half the household budget. Long-term yields are a Western portfolio problem; AI capex is a Western boardroom problem. Food is everybody's problem, and the South pays it first. Higher-for-longer, in this reading, is not a trading regime. It is imported punishment with a grocery bill attached.

The stakes are December. Four reports traced the arc from "expected convergence" to sticky reality, from AI as savior to AI as multi-faceted risk, from tariff numbers that mattered to unpredictability that does. The GDP forecasts barely moved through all of it — the institution holding its numbers while its prose changed underneath. The fifth report is the moment of truth: cut the numbers with the narrative still dark, and the soft landing is officially dead; hold the numbers again, and the OECD is managing expectations, not forecasting them. The forecast didn't move. The world did. What remains is to see whether the forecasters admit it.