The Convergence Point

The March Consumer Price Index, released April 10 by the Bureau of Labor Statistics, showed all-items inflation at 3.3% year-over-year — the highest reading since May 2024. On a monthly basis, the all-items index rose 0.9% in March alone, following 0.3% in February. The energy index surged 10.9% in the month, the largest monthly increase since September 2005. Gasoline rose 21.2% in March — the largest single-month increase for that series since its measurement began in 1967.

Core inflation — all items less food and energy — came in at 2.6% year-over-year, up from February, with shelter rising 0.3% in the month and airline fares up 2.7%.

These are not ambiguous numbers. They are the Iran war energy shock showing up in the consumer price data with a one-to-two month lag, compounded by tariff-driven price pressure in goods categories. Austan Goolsbee, president of the Federal Reserve Bank of Chicago, described the inflation outlook as "going from orange to red lately," said inflation is "stalled out at kind of 3%," and characterized the Iran conflict as a "stagflationary shock." He had previously been among the FOMC members most open to rate cuts in 2026. He now says rate cuts are off the table unless there is "proof that we're back on an inflation headed to 2%," and has indicated that absent that proof, rate reductions could be pushed to 2027 at the earliest.

Two assumptions, two doors

The financial models underwriting the AI infrastructure buildout made two structural assumptions about the macro environment.

The first was cheap energy. Goldman Sachs projected hyperscaler AI capex at over $500 billion in 2026. Oracle committed to $50 billion in data center construction, issued $43 billion in new debt, and restructured its workforce to free the cash flow to service it. Microsoft, Meta, Amazon, and Google made comparable commitments. Every model that justified these commitments assumed electricity costs within a foreseeable range, stable natural gas pricing, and supply chains uninterrupted by commodity shocks.

The $126 Barrel Problem documented the first door closing: the Iran war sent oil above $110, disrupted the Strait of Hormuz, removed 44% of Gulf oil exports in March, and subjected the AI buildout's energy assumptions to stress they were not designed to absorb. The Ceasefire Asterisk documented that the door had not fully reopened: spot prices eased to $92 with the ceasefire, but structural energy market scarring — damaged infrastructure, elevated insurance premiums, disrupted helium supply, South Korea's LNG security — persisted on a timeline measured in months to years.

The second assumption was cheap money. Every capex commitment made at the current scale was underwritten at a cost of capital calibrated to an environment in which the Federal Reserve was expected to cut rates in 2026 — reducing the financing cost on new debt issuance, supporting equity valuations that make large equity-financed buildouts feasible, and keeping the hurdle rate for long-duration infrastructure investment manageable.

March CPI at 3.3% closes the second door.

Why the Fed is boxed

The Federal Reserve's dual mandate — price stability and maximum employment — is being pulled in opposite directions by the current data, and the March CPI has resolved the ambiguity in the wrong direction for the rate-cut thesis.

The labor market is fragile. Goldman Sachs documented net negative AI-driven payroll growth of 16,000 per month over the past year. The Atlanta Fed's survey of corporate executives found larger companies planning AI-driven workforce reductions not yet visible in aggregate employment data. The Goldman scarring literature projects decade-long wage depression for tech-displaced workers. The labor market argument for rate cuts is real — a softening jobs market is exactly what the Fed is supposed to address with accommodation.

But the Fed cannot cut into 3.3% inflation. The FOMC's credibility depends on its 2% target being treated as a target rather than a suggestion. The energy shock that drove March's 0.9% monthly increase is partially transitory — if the ceasefire holds, gasoline prices will retrace. But core inflation at 2.6% and climbing reflects tariff-driven price pressure in goods categories that is not transitory. The Section 122 surcharge (15% global tariff, expiring July 24), the Section 232 restructuring (full customs value assessment on metals, effective April 6), and the pharmaceutical tariff (100% on patented imports, arriving mid-year) are supply-side cost increases that do not resolve with a ceasefire. They are policy instruments that raise consumer prices structurally until the policy changes.

Goolsbee used the phrase "stagflationary shock" deliberately. Stagflation — rising inflation combined with stagnant or contracting growth — is the Fed's worst scenario because the two halves of the dual mandate require opposite policy responses. Cutting rates to support employment worsens inflation. Holding or raising rates to control inflation tightens financial conditions in a fragile labor market. The Fed holds, because it cannot move in either direction without making the other problem worse.

What holding means for the buildout

The capex commitments are made. Oracle has the debt on its balance sheet. The construction contracts are signed. The Stargate facilities are under development. These programs do not stop because the Fed holds rates.

What changes is the cost of the next tranche.

The AI buildout is not a single investment decision made in 2026. It is a multi-year program that requires ongoing capital markets access: new debt issuances to fund expansion, equity valuations that support stock-based compensation and acquisitions, and credit conditions that allow the second and third tier of the buildout — colocation operators, enterprise data center companies, the firms supplying the infrastructure rather than owning it — to finance their share.

Every month the Fed holds rather than cuts, the cost of that ongoing capital access is higher than the models assumed. Not dramatically higher — the federal funds rate at 3.64% is not a crisis-level rate. But higher than the 2025 baseline, higher than what was underwritten in the capex projections, and now higher for longer than the rate-cut timeline the models assumed when the commitments were made.

The second-order effect is on valuations. AI infrastructure companies — hyperscalers, semiconductor manufacturers, data center REITs — are priced on discounted future cash flows. Higher discount rates compress valuations. Compressed valuations reduce the equity market's capacity to support the next tranche of investment without dilution. This is not a collapse scenario; the companies involved have the balance sheets to absorb it. It is a sustained headwind that was not in the original model.

The convergence

The $126 Barrel Problem identified the energy shock. The Ceasefire Asterisk established that the shock was not fully reversed. The South Korea piece documented the manufacturing tier exposure that US-centric coverage missed. The Bond Market's Forced Reckoning showed the tariff inflation signal becoming legible once the war-recession hedge unwound.

March CPI at 3.3% is where those threads converge. The energy shock generated the gasoline price surge that drove the monthly number. The tariff stack generated the goods price pressure driving core above 2.6%. The labor market fragility created by AI-driven displacement generates the growth risk that prevents the Fed from simply hiking through the inflation. The result is a central bank that cannot move, a cost of capital that cannot fall, and a multi-trillion-dollar buildout that assumed both.

The convergence point is not a crisis. The AI buildout will continue. The capex commitments are too large, the strategic logic too compelling, and the competitive pressure among hyperscalers too intense for 3.3% CPI to stop it.

What it means is that the financial architecture of the buildout — the debt structure, the equity valuations, the cost-of-capital assumptions embedded in every model that justified $500 billion in annual AI infrastructure spending — was built for a world that no longer exists. Cheap energy is gone. Cheap money is off the table, Goolsbee says, until inflation proves it is "headed to 2%." That proof is not in the March data.

The models need updating. The infrastructure is being built regardless.