A monthly view of the compute economy through one question: can cheaper intelligence create enough durable demand to pay for increasingly expensive capacity?
Read the July 2026 editionJuly 2026 · 2026-07-01 to 2026-07-31
July shifted the central AI question from whether models can do more to whether hyperscalers can turn cheaper intelligence and accelerating cloud demand into cash before capital expenditure, depreciation, power and operational risk overwhelm the economics.
Signature signal
The marginal cost of intelligence moved down while the capital needed to supply it moved up. The distance between the blades is the industry’s economic test.
Cost of intelligence vs cost of capacity
Marginal cost index
Launch baseline → month end
50
Disclosed quarterly capex
Cumulative July disclosures
$111.8bn
Source · OpenAI list-price announcements; Alphabet, Microsoft and Meta quarterly filings.
Limit · Inaugural edition. No prior-month baseline is backfilled under a different methodology.
Four lenses · twelve instruments
Capability, price, reliability and the path from launch to production.
Cloud demand against capex, depreciation, power and physical bottlenecks.
The Solow test, enterprise outcomes, funding and expectations.
Regulation, sovereignty, security and falsifiable forward claims.
Continuity
Each month carries the same methodology forward, so movement becomes visible over time.