The Great Bifurcation

“The AI race has crossed another boundary: from technology into monetary policy.”

The Great Bifurcation is becoming visible. AI is producing two different macroeconomic regimes at two different speeds — and the critical issue is timing.

Phase I — AI Infrastructure Inflation: accelerated computing → data centers → electricity → grids → construction → financing → labor → higher aggregate demand and capital costs. Phase II — Industrial-AI Deflation: AI diffusion → automation → productivity → lower unit costs → expanded supply → disinflation.

Central banks are confronting Phase I before economies have captured the full productivity dividend of Phase II. And yesterday brought an important confirmation.

1. 01 The Fed Is Starting to See AI as a Demand Shock

   Chicago Fed President Austan Goolsbee explicitly connected booming AI investment with unusually strong U.S. demand and persistent inflationary pressure, arguing that if inflation is increasingly demand-driven rather than the residue of supply shocks, monetary policy may need to respond more forcefully.

   Until recently, AI was predominantly discussed by central banks as a future productivity story. It is increasingly becoming a current macro-demand story. The BIS reached a similar conclusion in July: the AI boom is generating a large, increasingly debt-financed investment surge that affects both aggregate demand and potential supply. signal: AI CapEx ↑ → Aggregate Demand ↑ → Inflation Persistence ↑ → Policy Constraint ↑.

2. 02 The Fed's Numbers Confirm the Constraint

   The September FOMC projections underline the problem: 2026 GDP growth of 2.3%, PCE inflation of 3.7%, Core PCE of 3.4%, and an end-2026 federal funds rate of 4.1% — moved sharply higher from the 3.8% median projected in June.

   The stronger the investment boom becomes before productivity arrives, the greater the possibility that monetary policy must restrain the rest of the economy to accommodate it. signal: The projected policy rate is rising precisely because AI demand is arriving before AI productivity.

3. 03 Europe Has Entered the Same Equation

   The ECB raised its three key rates by 25 basis points on September 10, citing persistent inflation associated primarily with the Middle East energy shock. Its baseline now projects euro-area inflation of 3.0% in 2026, 2.5% in 2027 and 2.1% in 2028.

   Underneath the energy story, the ECB explicitly reported increased AI-related activity across digital services, business investment and exports. Lagarde noted that AI is becoming a major consumer of financing, increasingly extending from equity into bonds and private credit. signal: This is where the AI Supercycle begins interacting directly with the cost of capital.

4. 04 The Bottleneck Is Moving From Compute to Capital

   The first AI race was about GPUs. The next is increasingly about financing the physical system surrounding them: power and grid → data centers → infrastructure → debt financing → duration → cost of capital.

   AI therefore creates a second-order effect rarely incorporated into conventional technology analysis: AI competes for global savings. The enormous capital requirement can increase demand for long-duration financing at precisely the moment governments themselves are issuing large quantities of debt. signal: The AI race is migrating from the semiconductor market into the global capital market.

5. 05 The Great Bifurcation

   Between 2025 and 2027, the infrastructure buildout drives CapEx, energy demand, construction, skilled-labor scarcity and credit creation — all inflationary. From 2027 to 2030+, diffusion drives automation, productivity, lower marginal costs and supply expansion — all disinflationary.

   Central banks risk having to tighten against the very investment necessary to produce the productivity that could ultimately reduce inflation. That is the Great Bifurcation. signal: The demand shock arrives years before the productivity shock.

Quick Take

Five conclusions

• AI is no longer only a technology variable — it is becoming a macroeconomic variable.
• The first-order AI shock is increasingly inflationary: compute, electricity, infrastructure and capital are scarce.
• The second-order shock should be disinflationary, but productivity diffusion takes longer than infrastructure deployment.
• Central banks therefore face a timing mismatch, encountering the demand shock years before the productivity shock.
• The decisive variable becomes the cost of capital: the Supercycle must generate returns capable of surviving structurally higher hurdle rates.