How much AI revenue do hyperscalers need to break even?
Core argument: Goldman Sachs forecasts the largest US hyperscalers will spend $1.2tn on capex in 2027 and $1.4tn in 2028, $130bn and $100bn above consensus, with capex growth decelerating from roughly 100% in 2026 to 54% in 2027 and 12% in 2028.
The above estimates reflect expectations for average annual AI capex in 2026 and 2027 for the US hyperscalers, including AMZN, GOOGL, META, MSFT, ORCL, and SPCX. The breakeven revenues required for this capex represent the sum of annual depreciation and operating expenses. Exhibit 8 provides an estimated sensitivity of required AI revenues the hyperscalers must generate in different return on invested capital ("ROIC") scenarios. These required revenue estimates are anchored on average annual capex in 2026 and 2027 and would argue for even larger estimates based on cumulative capex. Realized hyperscaler return on invested capital has declined from 40% to 25%. Year/year cloud revenue growth for AMZN, GOOGL, MSFT, and ORCL has accelerated from 25% in 2024 to 48% in Q2 2026. In dollar terms, annualized cloud revenues in excess of their 2024 growth rate registered nearly $70 billion in Q2.

