The AI Trade Has Split Into Cash Flow and Capex

“Technology determines how fast the system can accelerate. The balance sheet determines who can afford to industrialize it.”

The morning session has delivered the clearest expression yet of The Great Split inside Artificial Intelligence.

Microsoft rose almost 8% before the U.S. open after Azure growth, contracted cloud demand and cash generation exceeded expectations. Meta fell more than 8% after its quarterly free cash flow collapsed 91% to $784 million, despite 28% revenue growth. European equities advanced modestly, but the U.S. 30-year Treasury yield reached 5.239%, its highest level since 2007, while Brent crude traded around $91.

This is no longer a binary debate between an AI boom and an AI bust. The market is separating AI that produces contracted revenue; AI that improves productivity; AI that consumes cash; and AI whose liabilities have merely been transferred into leases, debt and infrastructure vehicles.

The market is not questioning the industrialization of intelligence. It is demanding proof that intelligence can earn more than the capital required to produce it. Data cut: 12:55 CEST.

1. 01 Microsoft and Meta have defined the new AI hierarchy

   Microsoft reported quarterly revenue of $90 billion, with Azure growth of 43%. It forecast 45% constant-currency Azure growth for the current quarter, while paid Microsoft 365 Copilot seats exceeded 30 million. Cloud backlog reached $678 billion, with the quarterly increase driven by commitments outside the leading U.S. model developers. Microsoft generated $19.6 billion of free cash flow — 23% below the prior year but substantially above expectations — after investing $41 billion in the quarter. It expects $175 billion of reported capital expenditure during calendar 2026 and has $329.1 billion of data-center leases yet to commence. Meta produced 28% revenue growth to $60.8 billion, but free cash flow fell from $8.55 billion to $784 million; it raised the lower end of its capital-expenditure forecast and plans to expand computing capacity from roughly seven gigawatts in 2026 to 14 gigawatts in 2027. The market is applying a new three-part test: demand visibility × operating leverage × capital productivity. Microsoft passed because expenditure is matched by accelerating Azure demand, contracted backlog, paid enterprise adoption and meaningful current cash. Meta failed the morning test because its advertising engine remains powerful but the link between AI infrastructure and new economic value remains less visible. Microsoft is monetizing the infrastructure it already built; Meta is asking the market to finance the economics it still expects to create. signal: The difference is not ambition — it is the distance between capital committed and cash returned.

2. 02 The bond market has become the real central bank

   The Federal Reserve held its target rate at 3.50%–3.75%, with three policymakers dissenting in favor of an increase. Chair Kevin Warsh provided limited forward guidance, while the 30-year Treasury yield rose to 5.239% and market-implied odds of a September increase climbed to approximately 65%. The U.S. ten-year yield traded near 4.69%, while German ten-year yields stood around 3.17%. The Fed held short rates unchanged; the market raised the cost of long-duration capital anyway. This matters because the AI Supercycle is increasingly financed through corporate bonds, long-term leases, private infrastructure equity, project debt and power-purchase commitments. A 5.2% long-bond yield changes the valuation of data centers, grids, semiconductor plants and every asset whose returns lie years into the future. The chain now runs: energy risk → inflation uncertainty → higher long yields → higher AI hurdle rates → greater pressure on capital productivity. signal: The Fed is no longer the only institution tightening — the bond market is doing it directly.

3. 03 Europe has surprised on growth — but not escaped its energy constraint

   The eurozone expanded 0.4% quarter on quarter and 1% year on year, exceeding expectations, as AI investment, government expenditure, resilient consumption and stronger industrial activity contributed; unemployment remained at 6.3%. Spain grew 0.7%; the Netherlands 0.4%; Germany, France and Italy each 0.2%. The IBEX rose approximately 1% during the morning, against a 0.6% gain in France's CAC 40 and a 0.2% decline in Germany's DAX. Inflation accelerated across four major German states; economists expect Germany's harmonized rate to reach 2.8% in July, up from 2.4% in June, with eurozone inflation seen at 2.9%. The data reveal three Europes: Spain, where investment, services, infrastructure and international exposure reinforce one another; northern industrial Europe, where AI and defense spending support demand but energy costs constrain margins; and legacy corporate Europe, where automotive complexity and weak productivity dilute the improvement. The equation is becoming: AI investment + public infrastructure + resilient consumers − energy inflation − legacy industrial drag. The upside surprise is real; the second-half durability is unproven. signal: Europe has shown resilience, not yet escape velocity.

4. 04 Oil has reclaimed control of the inflation narrative

   Brent traded near $91 and WTI around $84.29 after the United States struck Iranian military infrastructure following missile attacks on American forces. Hormuz, which normally handles roughly one-fifth of global oil and gas flows, remains exposed to uncertain transit. The Houthis are considering fees on commercial ships using the southern Red Sea, while tankers scheduled to load at the Caspian Pipeline Consortium terminal changed direction after a vessel was struck during loading. Heidelberg Materials cut the upper end of its 2026 operating-profit forecast, citing higher oil, gas and electricity costs linked to the Iran war, and introduced fuel surcharges and price increases in Europe and North America. The world is no longer managing one energy chokepoint but a correlated corridor system — Hormuz, Bab el-Mandeb, the Black Sea, and pipelines and alternative export routes. This strengthens the case for domestic and diversified generation, nuclear power, grids and storage, pipelines and ports, maritime security and strategic inventories; and weakens the economics of airlines, unhedged transport, chemicals, cement, logistics and energy-intensive European manufacturing. signal: Oil at $91 is a tax on every business model that assumed geopolitics would stay outside its income statement.

5. 05 Europe's corporate market is rewarding direction — not size

   London's FTSE 100 reached an intraday record of 10,950.32, supported by miners, industrials and defense. Rolls-Royce gained 5.6% after raising its annual outlook, while British aerospace and defense shares advanced 3.6%. BBVA reported an 11.4% increase in quarterly net profit to €3.06 billion, lifted its profitability target to approximately 21% and announced a €2 billion buyback, with quarterly return on tangible equity of 22.2%. Stellantis reported €773 million of adjusted operating income, below the €914 million consensus; its margin stayed at 1.8%, its shares fell approximately 4%, and positive industrial free cash flow is not expected until 2027. The European market applies the same rule to banks, industrials and mobility that it applies to Big Tech: capital must produce visible conversion. BBVA is rewarded for profitability, loan growth and distribution; Rolls-Royce for converting aerospace scarcity into earnings; Stellantis is penalized because higher revenue has not yet produced sufficient margins, pricing power or free cash flow. signal: The split runs between firms shrinking the distance between investment and return — and firms widening it.

1. 01 Capital-Productivity Pressure 100
2. 02 AI and Accelerated Computing 98
3. 03 Energy and Power Infrastructure 98
4. 04 Enterprise AI Monetization 97
5. 05 Banking and Credit Formation 97
6. 06 AI Infrastructure Capital Formation 96
7. 07 Sovereign Capital 96
8. 08 Geopolitical Friction 96
9. 09 European Macroeconomic Velocity 91
10. 10 European Corporate Execution 87
11. 11 Mobility and Automotive Velocity 72
12. 12 AI Free-Cash-Flow Resilience 63

What to watch during the afternoon

The tests that decide the session

• The Wall Street open: whether Microsoft's gain survives the cash open and Meta's decline deepens — a sustained divergence formalizes the new law that AI revenue growth supported by free cash flow matters more; and whether Microsoft's strength lifts semiconductors, data-center suppliers and electrical infrastructure or stays company-specific.
• U.S. inflation data: the June PCE reading will decide whether the bond market extends its challenge to the Fed — a firm print plus Brent above $90 strengthens the case for higher long yields and tighter AI-financing conditions.
• The Bank of England: the decision and vote distribution test whether Britain joins the Fed in a hawkish hold; markets already price one increase and roughly a 50% chance of another by year-end.
• Amazon and Apple after the close: Amazon must show AWS growth and AI spending reinforce one another; Apple must show its ecosystem can convert device scale into an economically credible intelligence platform.
• Oil and strategic corridors: Brent around $90–$92, passage through Hormuz, any Houthi shipping-fee mechanism, and the Caspian Pipeline Consortium terminal — a synchronized disruption across two or more corridors would move energy from a sector issue into a global financial-conditions shock.

Implications

Investors · executives · policymakers

• Investors: strongest positioning remains in enterprise AI with contracted adoption; cloud platforms showing operating leverage; power generation, grids and electrical equipment; aerospace, defense and maritime security; banks with global fee income and high capital productivity; critical minerals and industrial infrastructure; and Spain and selected European firms converting investment into earnings. Greatest scrutiny: hyperscalers with deteriorating free cash flow, long-duration assets exposed to rising yields, energy-intensive European industry, legacy automotive, and projects with weak utilization guarantees. Own the conversion — not merely the exposure.
• Executives: every major AI and infrastructure plan should now disclose four separate figures — capital committed, contractual liabilities, contracted revenue, and free cash flow produced. The market is no longer impressed by the first without visibility on the other three.
• Policymakers: a stronger GDP reading should not create complacency. AI investment must be connected to lower-cost electricity, faster permitting, deeper capital markets, industrial skills, defense and infrastructure procurement, and reduced dependence on vulnerable trade corridors. Investment-led growth is durable only when it expands future capacity faster than it raises current costs.

The market has reached the most important phase of the AI Supercycle — not the phase in which intelligence becomes more powerful, but the phase in which investors discover who can afford to industrialize it. Follow the cash flow. Audit the leases. Price the energy. Measure the contracted demand. The Great Split is no longer between the companies participating in Artificial Intelligence and those left behind — it is between those that convert intelligence into economic power, and those that convert economic power into an ever-larger claim on the future.