⚪
New
Jul 30, 2026
The Price of Selling AI Is Collapsing
OpenAI cut inference pricing up to 80% on the same day hyperscalers reported record capex. The cost of building AI infrastructure is accelerating while the revenue per unit of inference sold is deflating — a scissors mechanism at the monetization layer that no earnings call has yet addressed.
THE SIGNAL
OpenAI API pricing: Luna (GPT 5.6) at $0.20 per million input tokens and $1.20 per million output tokens; Terra at $2 per million input tokens and $12 per million output tokens. Luna cut 80%, Terra cut 20%. OpenAI. 2026-07-30.
View source ↗
2026-07-30
THESIS CONNECTION
Layer 3 monetization compression from a direction the framework had not fully priced: not hardware commoditization (the defined V9 falsifier) but software-layer competitive deflation. Hyperscalers building infrastructure at record capex — MSFT $255-260B FY2027, META $130-145B 2026 — are selling inference into a market where the frontier pricing benchmark dropped 80% in under a year. Revenue equals volume multiplied by price, and price is collapsing. MSFT Azure AI, GOOGL Vertex, AMZN Bedrock must all price against $0.20/M input tokens. META's Q2 is the first print where this scissors is visible in the income statement: revenue +28%, net income -14%, FCF collapsed 91%. The capex is locked in at inflated hardware costs (Samsung: memory shortage through 2028). The revenue is being sold at deflating software prices. The shale parallel applies: massive upfront capital deployed into infrastructure whose per-unit economics are deflating in real time.
WHAT TO WATCH
- Falsifier: inference pricing stabilizes or rises as demand outstrips competitive supply
- Catalyst: AMZN earnings tonight Jul 30 — AWS inference pricing vs capex; Anthropic pricing moves
- Monitor: frontier model pricing trends quarterly; hyperscaler revenue-per-compute-unit disclosures; GPU/memory spot vs contract pricing divergence