Hyperscalers commit $1.1 trillion to AI data centers by 2027, requiring 10x+ earnings growth to break even by 2030
Wachter's capex model and OpenAI's biotech data acquisition both hinge on unproven productivity multipliers. Current filings show spend trajectories but lack corresponding revenue proof points. Operational success requires sustained utilization and token growth rates above 3x annually through 2028.
Wachter calculated required earnings trajectories for hyperscalers assuming capex peaks near $1.1 trillion by 2027. The model isolates capex from deployment uncertainty and shows revenue must rise at rates exceeding historical precedents by 2030. OpenAI Foundation separately funds acquisition of failed biotech regulatory filings and manufacturing data to expand training corpora.
NVIDIA's FY2025 10-K reports $26 billion in data center revenue tied to AI accelerators, while Microsoft and Google capex filings exceed $50 billion combined in 2024. These numbers align with Wachter's baseline but leave productivity multipliers unproven. OpenAI's dataset bid targets an estimated 10-20% increase in high-quality biological tokens.
The combined spend creates a narrow operational window. Data center utilization must exceed 70% sustained load with model inference margins above current benchmarks. Failure to hit earnings thresholds by 2028 triggers write-downs or slowed buildout.
Next milestone is 2026 earnings reports that will test whether inference revenue scales linearly with installed capacity.
Wachter model: hyperscaler AI-related earnings growth falls below 40% YoY by end of 2027 unless inference revenue per GPU exceeds $15k annualized.
Sources (3)
- [1]MIT Technology Review(https://www.technologyreview.com/2026/09/16/1144205/the-download-ai-trillion-dollar-build-openai-biological-data/)
- [2]NVIDIA FY2025 10-K(https://www.sec.gov/Archives/edgar/data/1045810/000104581025000012/nvda-20250126.htm)
- [3]Wachter AI productivity model(https://fnce.wharton.upenn.edu/profile/jwachter/)