According to a recent U.S. technology industry research report from Goldman Sachs, the firm constructed quantitative estimates to demonstrate the scale of AI economy needed to justify the ROIC on hyperscaler AI capital expenditures, focusing on the market's core debate: whether the massive AI compute investments by leading U.S. hyperscalers will deliver reasonable returns on their enormous capital outlays in the future.
The report established a quantitative analytical framework to conduct return stress tests on the 2026-2027 compute investments of six companies 鈥?Google, Amazon, Microsoft, Meta, Oracle, and SpaceX 鈥?providing a reference benchmark for the profitability outlook of AI infrastructure investment.
Goldman Sachs believes that amid the explosive growth in Token consumption driven by large language models, the capital intensity of major U.S. hyperscalers has undergone dramatic changes to adapt to the industrial transformation at the compute level. This round of massive investment by these companies stems partly from real demand signals in the current market (which is experiencing supply-demand imbalance) and partly from the need to meet the compute demands of core customers for years to come.
Looking at this year's market, especially in light of the recently concluded second-quarter earnings season, the market's understanding of this theme has clearly deepened. Based on earnings data and management commentary, investors have gradually formed two points of consensus: (a) hyperscalers can leverage prior capital expenditures (2023-2025) to achieve substantial, even better-than-expected investment returns through incremental revenue and operating cash flow; and (b) driven by tight compute supply and continuously accelerating end demand, cloud providers are initiating a new capital investment cycle (2026-2027), with current compute service pricing significantly above the firm's assumed long-term benchmark price.
As a result, the focus of market debate over the past few months has shifted to a longer-term question: how much return on investment can this round of capital expenditure in 2026-2027 generate in 2028-2030? And what kind of ROIC achieved from earlier capital expenditures can serve as a reference benchmark?
Goldman Sachs built an analytical framework in its report to estimate the scale of market incremental output the AI economy needs to generate for U.S. hyperscalers to achieve a baseline return on invested capital (ROIC) on capital expenditures of the current magnitude. The core of the analysis is to estimate the revenue threshold required for the second phase (2026-2027) of AI compute investment by leading U.S. hyperscalers to achieve a 15% annualized ROIC.
Based on a series of assumptions: average upfront investment of approximately $42 billion per gigawatt (GW) of compute; 70% of capital expenditure allocated to compute hardware and 30% to data center shells; conservative depreciation rules applied alongside ongoing operating cost assumptions, among others. The calculation results show that the six major U.S. hyperscalers (Alphabet, Amazon, Microsoft, Meta, Oracle, and SpaceX) need to collectively generate approximately $1.42 trillion in cumulative revenue during 2028-2030 (equivalent to approximately $11.6 billion in annual revenue per gigawatt of compute) to cross the 15% ROIC threshold.
Although significant short-term capital expenditure expansion will weigh on near-term returns (short-term performance will come under pressure), the firm believes that the attractive ROIC level of this capital expenditure cycle will gradually materialize in medium-to-long-term operating statements. In other words, short-term return pressure is a natural consequence of a large-scale front-loaded investment cycle and does not indicate a structural flaw in the AI business model itself.
In several previous reports on the Token economy, consumer AI development landscape, and enterprise AI sector, the firm has already elaborated that market share shifts and declining Token pricing will actually increase AI penetration rates, driving continuous expansion of compute application scenarios on both consumer and enterprise sides.
Current market focus is concentrated on the evolution of frontier foundation model providers 鈥?including market share and the pricing power of compute segments relative to open-source models. However, from the firm's perspective, it is the cost-performance landscape of compute that determines the overall scale and development boundaries of the AI economy. Multiple models with different capability tiers and agents spread along the cost-performance curve will unlock a rich range of application scenarios 鈥?from "commoditized affordable intelligence (under significant price deflation pressure)" to frontier cutting-edge intelligence (with stronger pricing resilience) 鈥?which collectively will support the current capital investment by compute infrastructure providers.
In short, not every type of Token carries identical commercial value, and not every dollar of capital expenditure will yield uniform returns. But taken together, considering the vast market space over the next 3-5 years, the firm still judges that capital investments deployed over the next 18 months can generally achieve decent return levels.
This was also corroborated at Goldman Sachs' Communacopia Technology Conference, where participating companies conveyed three signals: (a) the industry has moved from the AI experimentation and exploration phase to the implementation phase; (b) returns from efficiency gains and product iteration cycles are accelerating; and (c) corporate thinking is shifting from purely pursuing Token consumption scale to optimizing Token input-output efficiency, improving companies' own return levels, which also signals that enterprise-side AI penetration will rise further (enterprise AI adoption will directly drive cloud providers' revenue and incremental operating profit corresponding to earlier capital expenditures).
In addition, several consumer-facing agentic AI products have recently been officially launched (most notably Meta's Muse), marking a paradigm shift in consumer AI: from conversational interaction products to agent-driven action-execution product forms. The scaling of such agent platforms (with other leading tech giants likely to follow suit in building consumer AI platform layers) will drive medium-to-long-term compute demand growth; corresponding monetization models will also gradually materialize, spanning subscriptions, advertising, e-commerce, and multiple other pathways. The boundaries of these business models are themselves continuously converging, and they will become core drivers of future revenue growth.