Goldman Sachs and Apollo Economists Find AI Capex Growth Impulse Nearing Peak With No Measured Payoff Outside Big Tech

The macro impulse supporting modern economic growth is cresting before delivering measurable earnings or productivity gains to 98% of public corporations.

* Wall Street economists marked the ceiling: Goldman Sachs chief economist Jan Hatzius and Apollo chief economist Torsten Slok issued coordinated research concluding that the economic growth boost from AI capital spending is approaching its peak. * Domestic GDP contribution stalled at 0.1 percentage points: Despite approximately $600 billion deployed into AI hardware and data facilities, Goldman Sachs calculates that import leakage and foreign sourcing have capped the actual domestic growth contribution at a fraction of prior expectations. * Non-tech financial returns registered at zero: Apollo documented that while the Magnificent 7 captured higher margins, the remaining S&P 493 experienced flat revenue per worker and flat operating margins, treating AI token outlays strictly as margin-diluting expenses.

* Hyperscaler spending drove prior growth models: Prior to late 2026, tech infrastructure outlays anchored corporate investment benchmarks, with Goldman Sachs previously modeling potential long-term GDP expansions reaching up to 1.5 percentage points. * Projections point toward an expenditure peak: Total AI infrastructure commitments are projected to reach $725 billion to $800 billion in 2026, but the rate of investment growth is set to crest between late 2026 and early 2027. * Historical infrastructure parallels repeat: Classical technology rollouts like railroads and electrification saw years of frontloaded physical capital deployment before operational workflows caught up to generate actual bottom-line gains.

The economic mechanics of artificial intelligence have decoupled into two opposing realities: the balance sheets of hardware suppliers and the income statements of everyone else. Over the past year, capital expenditure from hyperscalers served as an anchor for domestic economic activity. However, data from Goldman Sachs and Apollo indicates that this multi-hundred-billion-dollar impulse is delivering diminishing aggregate economic expansion.

Jan Hatzius calculated that AI-related capital expenditure—running at an estimated rate of $600 billion—has generated just a 0.1 percentage point addition to recent United States GDP growth. The structural friction behind this gap is import leakage. Because a high share of critical hardware components, semiconductors, and server assemblies are sourced from international supply chains, the domestic GDP formula accounts for substantial equipment imports as an offset to gross private domestic investment. Consequently, the domestic economic output retained from each dollar spent is structurally muted. Hatzius projects that this capital expenditure impulse will hit its absolute growth peak in late 2026 or early 2027.

Simultaneously, Torsten Slok analyzed performance divergence within the benchmark equities index. His findings confirm an asymmetric distribution: revenue per worker and operating profit margins expanded for the Magnificent 7 mega-cap technology firms, but remained flat across the remaining S&P 493 companies. For non-technology enterprises—spanning retail, healthcare, logistics, and industrials—enterprise software subscriptions and model API token consumption have functioned as margin-diluting cost centers.

Instead of unlocking labor efficiencies or expanding corporate margins, integration outlays have added software overhead without shifting baseline worker throughput. This dynamic creates an economic friction point. As hyperscalers expand annual infrastructure outlays toward $725 billion to $800 billion, their corporate customers are failing to demonstrate the revenue expansion required to justify ongoing commercial pricing tiers. The incentive structure now forces corporate leadership to defend unreturned software investments while infrastructure providers face a mathematical flattening in incremental expenditure growth.

* Enterprise Chief Financial Officers: Face mounting scrutiny over recurring software line items as board directors audit AI pilot programs that fail to lift operating margins. * S&P 493 Corporate Operators: Absorb higher technology overheads without realizing improvements in output or reductions in total labor expenses. * Institutional Equity Allocators: Re-evaluate portfolio risk profiles tied to mega-cap technology concentration as hyperscaler capex growth guidance slows. * Infrastructure and Procurement Leads: Confront potential spending freezes on multi-year hardware and data center leases if enterprise client subscriptions decelerate.

* Hyperscaler forward capex guidance: Quarterly earnings calls will signal whether mega-cap tech firms begin moderating multi-billion-dollar hardware budgets in early 2027. * S&P 493 operating margin prints: Subsequent earnings cycles will reveal whether non-tech enterprises can convert digital pilot programs into measurable reductions in cost per unit. * Hardware import data and customs tallies: US trade balance metrics will illustrate whether domestic component manufacturing captures more value or continues leaking capex abroad.