Inside Goldman Sachs' $5.3 Trillion AI Infrastructure Forecast

The scale of capital pouring into artificial intelligence infrastructure has entered uncharted territory. On June 3, 2026, Goldman Sachs published a research note that significantly raised its long-term spending forecasts for the industry's biggest players, signaling that the AI buildout is accelerating faster than even optimistic analysts had predicted . The report details not just eye-popping dollar figures, but a fundamental shift in how these massive projects will be financed and powered.

The $5.3 Trillion Forecast: A Six-Year Spending Surge

The centerpiece of Goldman's revised outlook is a cumulative capital expenditure forecast of $5.3 trillion for the four largest hyperscalers—Meta, Microsoft, Amazon, and Alphabet—spanning fiscal years 2025 through 2030. This represents a substantial upgrade from the bank's prior estimate of $4.5 trillion for the same period, made before the start of first-quarter earnings .

The upward revision reflects a reality where each earnings cycle brings higher spending guidance. Goldman Sachs had previously projected that hyperscaler capex could exceed $500 billion in a single year, but actual company plans have raced past those estimates . This trend of consistent upward revisions has been a defining feature of the AI investment cycle, with analyst consensus for 2026 spending climbing sharply over several quarters .

The 2026 Landmark: ~$725 Billion in a Single Year

The year 2026 stands out as the moment when AI infrastructure spending reaches a new plateau. The four hyperscalers plan roughly $725 billion in combined capital expenditures, a staggering 77% jump from an estimated $410 billion in 2025 . Company earnings releases from late April 2026 placed the combined guidance midpoint at approximately $710 billion, with the upper end of guidance reaching $725 billion .

This single-year figure dwarfs historical comparisons. Total hyperscaler capex just three years prior, in 2023, was approximately $160 billion . The near-term acceleration is even sharper than it appears: Wall Street consensus for 2026 capex had already risen to $527 billion by late 2025, but updated guidance revealed growth projections surging to around 70% year-over-year .

Approximately 75% of 2026 spending, or nearly $500 billion, is tied directly to AI infrastructure—including GPUs, servers, networking equipment, and data centers—rather than traditional cloud computing .

The Financing Revolution: Private Capital Steps In

Goldman's note highlights a crucial structural change in how the data-center boom will be funded. The sheer scale of required investment is pushing companies beyond traditional balance-sheet and debt financing. Private infrastructure and real estate capital are expected to play a much larger role .

This shift is already underway. Hyperscalers raised $108 billion in debt during 2025 alone, with projections suggesting up to $1.5 trillion in total debt issuance over the coming years to fund the buildout . The bank's report projects that total infrastructure assets under management could surpass $3 trillion by 2030, reflecting the scale of capital flowing into AI-related power, data-center, and grid infrastructure .

This financing evolution matters because it broadens the pool of investors exposed to AI infrastructure beyond public equity markets. It also signals that the capital demands of the AI buildout are now competing with—and in some cases exceeding—the scale of traditional infrastructure asset classes.

Power Demand: A 220% Surge by 2030

The physical footprint of this spending is immense, and nowhere is that more visible than in electricity consumption. Goldman Sachs Research has raised its projection for global data-center power demand growth to 220% by 2030 compared to 2023 levels, up significantly from earlier forecasts of roughly 160%–165% .

An analyst at Goldman Sachs Global Investment Research stated in a February 2026 report: "We have raised our forecast for global data center electricity demand growth from 175% to 220% between 2023 and 2030" . The United States is expected to account for approximately 60% of that incremental demand .

This upward revision has profound implications. Data centers are projected to consume 8% of total U.S. power demand by 2030, up from about 3% currently, driving the first sustained growth in U.S. electricity demand in a generation . Goldman estimates that about 47 gigawatts of incremental power generation capacity will be required in the U.S. alone .

What This Means for the Broader AI Buildout

The Goldman Sachs Global Institute separately models an even wider industry picture. Its "Tracking Trillions" baseline projects roughly $7.6 trillion in total AI capital expenditure between 2026 and 2031 across compute, data centers, and power, with spending scaling from $765 billion in 2026 to $1.6 trillion annually by 2031 .

While those figures include a broader set of spenders beyond the Big Four hyperscalers, the direction is consistent: the AI infrastructure cycle is accelerating, not plateauing. The question that increasingly preoccupies markets is not whether the spending will continue, but whether the revenue returns will justify it. As Goldman noted in earlier research, maintaining historical returns on capital would require the hyperscalers to realize an annual profit run-rate of over $1 trillion—more than double the 2026 consensus income estimate .

The June 2026 research note makes clear that for now, the spending commitments are only growing. And as the financing models evolve and the power grid strains to keep up, the AI buildout is reshaping not just the technology sector, but the infrastructure and energy landscape for the rest of the decade.