“More capital. Less incremental labor.”
What if the next economic regime is neither recession nor soft landing, but accelerating capital formation inside an economy that requires progressively less incremental labor?
That is the question markets should carry into Monday. Friday delivered one of the clearest manifestations yet of The Great Split.
The United States lost 23,000 jobs in July, against expectations for an 80,000 increase, with May and June revised down by another 103,000 jobs. Yet the S&P 500 closed at a record 7,757.64. This morning's Chinese inflation data slowed to a six-month low even as exports surged 23.9%. Two economies are increasingly visible inside both the United States and China — one driven by employment and household demand, the other by AI, industrial investment, exports and capital intensity.
Key Numbers to Watch
the data behind the divergence
The Five Overnight Signals
how the split is surfacing
-
The U.S. lost 23,000 jobs in July. Unemployment nevertheless fell from 4.2% to 4.1% because 264,000 people exited the labor force. Yet equities rallied because weaker employment reduced the probability of Fed tightening.
Markets are treating labor weakness as a lower discount rate rather than a lower earnings event — for now. The traditional relationship between Main Street momentum and asset prices is weakening. signal: employment ↓ → rate expectations ↓ → equity valuations ↑
-
July exports increased 23.9%, high-tech exports are up 40.7% year to date, and semiconductor exports have nearly doubled. Yet CPI slowed to 0.5%, core to 0.9%, and food prices declined 1.5%.
China is developing an increasingly clear K-shaped productive structure: AI infrastructure, semiconductors and industrial exports on the upper arm; household consumption, property and domestic pricing power on the lower. China can become more powerful industrially while remaining weaker domestically. signal: the wider Export Velocity / Domestic Demand Velocity becomes, the more pressure on international industrial markets.
-
Nvidia is reportedly preparing to invest as much as $3 billion in Lancium, the developer behind the first operational Stargate data-center site in Abilene, Texas. The phrase that matters is not $3 billion — it is grid hookups.
Yesterday's scarcity was GPU capacity. Today's is HBM, networking and electricity. Tomorrow's is grid connection, permitted land, firm power, cooling and financing. This would represent something larger than vertical integration: bottleneck integration. signal: the chip company becomes an infrastructure company becomes an energy company becomes part of the intelligence supply chain.
-
Microsoft, Meta, Oracle, Amazon and Alphabet have disclosed approximately $1.09 trillion in payments under leases not yet commenced, mainly for data centers. Including Meta agreements signed after quarter-end lifts the pipeline to around $1.16 trillion.
The question is no longer how much hyperscalers are spending, but how much future cash flow has already been pre-committed to infrastructure whose utilization is unproven. AI is industrializing the technology balance sheet: high fixed cost, low marginal cost, huge economies of scale — and brutal economics if utilization falls. signal: watch Incremental AI Cash Flow / Fixed AI Commitments.
-
Iran said Saturday an agreement with Oman on the Strait of Hormuz was close, but that it alone would not reopen the waterway. Markets trade the probability of diplomacy; industrial companies must manage the probability of physical disruption. Roughly one-fifth of global oil and gas shipments transited Hormuz before the current conflict.
This matters especially for Europe, where Q2 earnings are projected to rise 22.4%, ex-energy up 11.5%, but energy profits up 135.8% — higher energy profits against higher consumer energy costs. signal: record equities / greater macroeconomic fragility.
Hidden Patterns Beyond the News
the production function is shifting
The Split has moved into the production function
Capital formation is accelerating across AI factories, fabs, data centers, grid infrastructure, energy generation and memory, while U.S. job creation averaged only 20,000 over three months. The defining question shifts from how many workers additional GDP requires to how much capital it requires without additional workers.
AI deflation is producing infrastructure inflation
Marginal intelligence cost falls while fixed infrastructure cost rises. Cheaper intelligence makes more applications viable, generating more tokens, demanding more compute, more electricity, more generation and grid connections. This is Jevons' paradox applied to intelligence.
Sovereign capital is moving toward productive capacity
Sovereign capital is migrating from passive portfolio ownership toward direct ownership of compute, energy, land, data centers and semiconductors. This increasingly resembles national positioning inside the productive architecture of the Next Economy.
Europe's Industrial AI opportunity exceeds its software rankings
Europe need not defeat every frontier model to matter. Its opportunity lies in the physical layer — power equipment, grid technologies, industrial automation, cooling, optics, semiconductor equipment. Industrial AI requires an industrial base; Europe already owns parts of it. The problem is orchestration, scale and capital formation.
The Fundamental Analytics
three ratios for every dashboard
what to carry into Monday
- The market should distinguish AI exposure from AI economic quality — scarcity deserves a premium, capacity alone does not.
- For CEOs, AI strategy is becoming capital strategy: track AI revenue/CapEx, revenue per MW, intelligence per watt and infrastructure utilization at board level.
- Policymakers face an unconventional combination: weak labor formation, persistent inflation and extraordinary capital formation.
- Critical assets increasingly include grid-connected land, firm power, transformers, memory, optical infrastructure and contracted data-center capacity.
- The Great Split is no longer merely inequality or market polarization — it is becoming a capital regime.