US GDP Q1 2026 Without AI: How Much of the 2.0% Growth Came From the Data Center Buildout

The Bureau of Economic Analysis put real GDP growth at 2.0% annualized in Q1 2026, below the 2.3% economists had forecast. That headline number understates what is actually happening underneath. Strip out the categories carrying most of the AI buildout (computer and peripheral equipment, plus software), and the picture changes substantially. By one measure, AI-related capital formation contributed roughly as much to GDP growth as all of household consumption combined. By another, AI capex now represents about 5% of US GDP, a level last seen during the late-1990s technology boom.

The question of how much of GDP growth is AI spending, and what real GDP would look like without it, is not a hypothetical. Big Tech's Q1 2026 earnings disclosed roughly $700-$725 billion of planned 2026 capex across the five largest names. That capex is showing up directly in the official statistics, on the same lines and at roughly the same scale.

What the BEA Numbers Actually Say

In the BEA's Q1 2026 advance estimate, computer-and-peripheral-equipment investment grew at a 67.4% annualized rate. Software investment grew at 22.6%. Together, those two line items contributed approximately 1.09 percentage points to the 2.0% headline growth figure. Personal consumption (the entire household consumption category, accounting for about 68% of GDP) contributed 1.08 percentage points. The two categories therefore made roughly equal contributions to Q1 growth, even though the consumption base is approximately twenty times larger by share of GDP (roughly 68% versus 3.7%).

The arithmetic is the point. A small base growing at high rates can move the headline number as much as the entire consumer economy growing at trend. In Q1 2026, that is what happened.

GDP Without AI Spending: A Rough Counterfactual

A clean counterfactual is impossible; pulling AI investment out of GDP also affects employment, supplier output, and downstream services. But the direct contribution is straightforward to bound. If computer-and-peripheral-equipment plus software investment had grown at the rate of the rest of the economy in Q1 2026 (roughly the consumer-spending pace, in the 1-2% range), the contribution from those lines would have been around 0.05-0.10 percentage points instead of 1.09 percentage points.

Substituting the trend rate for the actual rate implies Q1 GDP growth without the AI investment surge would have been on the order of 1.0% annualized, give or take, instead of 2.0%. That is a coarse estimate (it does not account for the construction spending, power generation investment, and equipment supply chain employment that data center buildouts pull along), but it sets the order of magnitude. Roughly half of the headline GDP growth in Q1 2026 traces back to spending categories where the AI buildout is the dominant source of incremental demand.

Goods imports rose sharply in Q1, driven (the BEA noted) in large part by computers and related equipment associated with the data center boom. Imports subtract from GDP in the national accounts. So the gross contribution of AI investment to domestic activity is even larger than the +1.09 ppt net figure: a meaningful share of the equipment is being purchased abroad (NVIDIA chips manufactured by TSMC in Taiwan, for instance) and shipped to US data centers, where it then shows up as domestic investment.

AI Capex as a Share of GDP: 5%, and Climbing

Combined hyperscaler 2026 capex is approximately $725 billion, with about 75% of that (roughly $545 billion) directed at AI-specific infrastructure: GPUs, servers, data center construction, power systems, and cooling. Against a US GDP of approximately $30 trillion, AI-related capital formation now sits at roughly 5% of GDP. That is the highest reading since the late 1990s telecom and dot-com investment cycle, when fiber-optic and networking capex briefly reached comparable levels.

The composition is different from the late-1990s cycle in one important respect: today's AI capex is concentrated among five companies (Microsoft, Alphabet, Meta, Amazon, plus Apple at smaller scale), all of which are profitable, cash-generating businesses. The 1990s telecom buildout was distributed across dozens of carriers, many of which were operating at a loss and financing capex with debt that ultimately defaulted. The financing source for the 2026 buildout is mostly internally generated cash flow plus relatively conservative incremental debt; Microsoft's, Alphabet's, and Meta's existing balance sheets absorb most of the spending. That changes the systemic risk profile, even if the gross capital intensity is similar.

The Three Pieces of the AI GDP Story

First, equipment investment. This is the cleanest line: computer-and-peripheral-equipment investment grew at 67.4% annualized in Q1 2026 and contributed 0.58 percentage points directly. The hyperscaler capex breakdown matches: Microsoft disclosed that roughly two-thirds of its capex went to short-lived assets, primarily GPUs and CPUs. The macro and the company-level numbers are aligned.

Second, software investment. Up 22.6% annualized, contributing 0.51 percentage points. This bucket captures both purchased software and own-account software development. AI model training, foundation model R&D, and AI-driven internal tools all contribute. The 33% R&D growth rate at Apple in fiscal Q2 2026 is a microcosm of the same pattern: large companies are spending more on software development, including AI software, faster than revenue growth.

Third, structures investment. This is harder to isolate because data center construction does not have its own BEA line; it sits inside non-residential structures. But Q1 2026 data shows non-residential structures investment was a contributor to growth, and ConstructConnect data points to data centers as a leading subcategory of construction starts. Some industry estimates put US data center construction spending near $80 billion at recent run rates, an approximately 200% increase over the prior couple of years. Power generation investment, both gas peaker plants and nuclear restarts, is being driven by the same demand and shows up in private fixed investment as well.

What Real GDP Would Look Like Without AI Spending

• Reported Q1 2026 real GDP growth: +2.0% annualized.
• Direct contribution from computer/peripheral + software investment: ~+1.09 percentage points.
• Contribution at a trend-rate counterfactual: ~+0.05-0.10 percentage points.
• Estimated Q1 2026 GDP growth without the AI investment surge: roughly +1.0% annualized.
• AI-related capital formation as share of GDP: ~5% (highest since late 1990s).
• Combined 2026 hyperscaler AI capex: ~$545B (out of ~$725B total tech capex).
• Data center construction starts: leading subcategory of non-residential construction in 2026.

What This Means for the Rest of 2026

The arithmetic implies the US economy in 2026 has two real growth rates: one with the AI buildout running flat-out, and a slower one underneath it. If the buildout pace holds, headline GDP growth keeps a 1+ percentage-point tailwind from equipment and software investment. If hyperscaler capex slows (for instance, because customer commitment coverage at AWS softens, or because component prices ease), that tailwind compresses fast.

The forward signal that matters is the next-quarter capex disclosures from Microsoft, Alphabet, Meta, and Amazon. Microsoft already guided fiscal Q4 2026 capex above $40 billion. Alphabet projected 2027 capex "significantly higher" than 2026. Meta raised its 2026 range twice in successive quarters. Amazon held its $200 billion plan. As long as those numbers keep rising, the GDP contribution from AI capex compounds. The day they stop rising (or start falling) is the day the underlying growth rate of the rest of the economy becomes the headline number again.

For the labor market, the corollary is that data center construction, electrical equipment manufacturing, semiconductor packaging, and AI software development are concentrated sources of marginal demand. The geographic concentration of those activities (Texas, Virginia, Ohio, Arizona) is shifting which regional economies grow fastest. The Q1 2026 GDP report did not separate that out, but the underlying dynamics are visible in the BEA's regional series and in Bureau of Labor Statistics employment data.

The Comparison That Matters

The closest historical analogy is the late-1990s telecom and tech investment boom. From 1996 to 2000, computer-and-peripheral-equipment investment plus software investment contributed materially to US GDP growth, peaking at a similar share of GDP to today's reading. When the cycle turned in 2001, equipment investment fell sharply and dragged GDP growth into recession. The shape of the eventual reversal in 2026 will likely depend on how much of the current capex is supported by recurring cloud revenue commitments versus speculative anticipation of future demand.

On the supportive side, Microsoft's $627 billion of remaining performance obligations and Amazon's disclosed customer commitments suggest a meaningful share of the 2026 capex has already been pre-sold. On the cautionary side, those commitments do not cover all of the capex, and the depreciation curve on $545 billion of AI infrastructure will eventually hit the income statements regardless of revenue ramp.

For now, the Q1 2026 GDP print is what it is: a headline number where roughly half the growth came from a single investment theme. Without the data center buildout, US growth in Q1 would have looked considerably weaker. With it, the economy is running hot in one specific direction and at trend in most others.