Goldman Sachs and Apollo Economists Flag Peak in AI Capex Growth Impulse With Gains Confined to Tech Sector

A projected plateau in capital expenditure alongside stagnant productivity outside major technology firms threatens to remove the primary economic buffer supporting broader market valuations and output growth.

• Top Wall Street economists signaled a peak: Goldman Sachs chief economist Jan Hatzius and Apollo chief economist Torsten Slok released separate analyses on October 24 and October 26, 2026, warning that the macroeconomic boost from artificial intelligence investment is cresting.
• Capital expenditure delivered minimal growth: Hatzius calculated that roughly $600 billion in artificial intelligence capital expenditure contributed only 0.1 percentage points to recent United States gross domestic product growth, largely due to import leakage.
• The rest of the market saw flat returns: Slok reported that while Magnificent 7 tech companies expanded profit margins and revenue per worker, the remaining S&P 493 experienced flat productivity and recorded artificial intelligence token usage as a direct cost center.

• Hyperscalers carried the broader macroeconomic expansion: Over the preceding year, substantial corporate deployment of hardware, chips, and data center facilities counterbalanced wider macroeconomic headwinds and elevated government deficits.
• Earlier projections outpaced near-term realities: Goldman Sachs had previously modeled potential long-term growth gains reaching up to 1.5 percentage points, which helped fuel corporate willingness to accept rising bond yields linked to infrastructure borrowing.
• Capital waves follow historical adoption cycles: Previous major infrastructure deployments, including historical rollouts of railroads and electrification, exhibited similar front-loaded infrastructure outlays accompanied by delayed widespread commercial productivity.

The economic mechanics governing corporate artificial intelligence have bifurcated into two distinct tracks: the capital infrastructure suppliers and the end-user enterprises attempting to deploy the technology. For the past year, capital expenditure led by hyperscale technology providers served as the foundational pillar supporting United States economic metrics. However, the transmission mechanism from infrastructure buildout to macroeconomic output has run into structural constraints.

Jan Hatzius at Goldman Sachs identified that approximately $600 billion in artificial intelligence investments translated into an addition of just 0.1 percentage points to recent United States gross domestic product. A significant portion of this outturn stems from import leakage: because extensive shares of specialized components, materials, and specialized processing hardware are imported, the capital deployed fails to generate a proportional multiplier effect within the domestic economy. Hatzius projected that this growth impulse as a share of gross domestic product will reach its upper limit in the fourth quarter of 2026 or early 2027, reflecting structural saturation points seen in earlier industrial deployment phases such as rail networks and municipal electrification.

Simultaneously, financial metrics published by Apollo’s Torsten Slok reveal an incentive disconnect across public equities. While the Magnificent 7 technology firms continue to capture revenue expansion and higher revenue per worker—driven by the sale of foundational compute and cloud access—the remaining S&P 493 enterprises have absorbed these products strictly as input expenses. Non-technology corporations have not realized measurable improvements in employee output or profit margins. Instead, expenditure on inference tokens, model subscriptions, and enterprise software add-ons is registered as an operational margin drag.

Consequently, enterprise leadership outside technology faces a shifting incentive structure. Because corporate balance sheets outside the dominant platform providers must account for compute usage without offsetting revenue gains, spending patterns face impending rationalization. When artificial intelligence infrastructure spending crests without delivering measurable bottom-line expansion to end-user corporations, the reliance on concentrated technology investment to offset broad macroeconomic pressures diminishes.

• Enterprise Chief Financial Officers and Procurement Heads: Corporate finance leaders in non-technology sectors must audit active artificial intelligence software licenses and inference token budgets that have functioned as margin costs without yielding higher worker output.
• Technology Infrastructure Providers and Hyperscalers: Cloud platforms and hardware suppliers face scrutiny from capital markets as corporate clients reassess compute allocations and demand demonstrated return on investment.
• Fixed-Income and Equity Allocators: Institutional investors managing portfolios weighted heavily toward large-cap equities must account for concentration risk as the growth impulse from the Magnificent 7 decouples further from broader S&P 493 corporate fundamentals.

• S&P 493 Corporate Margins: Track upcoming quarterly reports from non-tech firms to observe whether enterprise software integration moves from a cost drag into demonstrable margin improvement.
• Hyperscaler Forward Guidance: Monitor guidance on future capital spending schedules from top technology operators to confirm the projected late-2026 or early-2027 investment plateau.
• Bond Market Yield Trajectories: Observe high-yield and investment-grade debt spreads connected to digital infrastructure projects to evaluate whether borrowing costs adjust as capital expenditure crests.