Hyperscalers' $3 trillion AI spend is double the US defense budget

Hyperscalers are expected to spend roughly $3 trillion in cumulative capital expenditures over the next three years, a figure Everpure CFO Tarek Robbiati said is twice the US defense budget, confirming an unprecedented buildout of AI compute infrastructure.

"The hyperscaler category is spending at a rate that dwarfs government investment," Robbiati said at the Everpure Accelerate 2026 summit in Las Vegas. "We are looking at $3 trillion over three years — 2 times the defense budget."

The FY2026 enacted Department of War budget came in at roughly $1.01 trillion, with the FY2027 request totaling about $1.45 trillion. Alphabet alone spent $35.7 billion on capex in Q1 2026 and raised full-year guidance to $180 billion to $190 billion, with CFO Anat Ashkenazi telling investors 2027 spending "will significantly increase compared to 2026." Microsoft guided to roughly $190 billion in calendar 2026 capex, including approximately $25 billion from higher component pricing, with CFO Amy Hood saying the company expects to "remain constrained at least through 2026." Amazon reported $43.2 billion in cash capex in Q1 2026 and disclosed over $225 billion in revenue commitments for Trainium, with CEO Andy Jassy noting the AWS backlog of $364 billion does not include the recent Anthropic deal of over $100 billion.

The spending wave flows directly to suppliers feeding the buildout. Nvidia, trading at 23 times forward earnings with an analyst target of $298.93, stands to capture the largest share of GPU procurement. Broadcom's XPV project, targeting roughly 20 gigawatts of custom chips delivered through 2028, offers investors a diversification play beyond Nvidia's dominant position.

$3 Trillion in Capex Reshapes the Hyperscaler Club

The hyperscaler category has widened beyond the traditional quartet of Meta, Alphabet, Amazon and Microsoft. SpaceX, following its IPO and early-2026 acquisition of xAI, now qualifies after securing roughly 20 percent of Nvidia's next-generation Vera Rubin chip allocation, pushing its AI infrastructure spend into the same tier as the incumbents. The company's entry adds a fifth major buyer to a market that was already straining supply chains.

Meta raised its 2026 capex envelope to $125 billion to $145 billion, with Mark Zuckerberg flagging a $107 billion step-up in contractual commitments this quarter. Google Cloud's backlog stands at $462 billion, per the company's Q1 filing. Amazon's AWS backlog of $364 billion excludes the Anthropic deal, which alone exceeds $100 billion. Together, these commitments represent multi-year revenue visibility that justifies the upfront infrastructure spending.

Everpure Targets Enterprise Data as the Next AI Bottleneck

Everpure, the company formerly known as Pure Storage, is marketing its Universal Data Intelligence platform as the software layer that makes enterprise data usable for AI agents. VP of Product Development Chad Kenny said most corporate data architectures were "built in a very different computing era" and that companies are "taking agents and connecting them to all their apps, which has no security or governance really built around it."

The platform can "rationalize data outside of Everpure" without requiring customers to migrate everything onto Everpure systems, Kenny said. Robbiati noted the return on investment and total cost of ownership conversation has become central to every major infrastructure decision over the past 6 to 12 months. Product managers are now rated anywhere from 2 times to 5 times more productive depending on AI tool adoption, helping move corporate buyers from pilot budgets to multi-year infrastructure commitments.

For Everpure, the strategic shift means selling into chief data officers — a buying center it had not previously engaged. If the $3 trillion thesis holds, the spend flows through suppliers feeding the buildout, and Everpure's storage and data management products stand to benefit as enterprises build the data layer to support AI agents.

Investor Impact

The number to watch is whether 2027 capex guidance from the hyperscaler cohort steps up in line with Robbiati's framing. Alphabet's Ashkenazi has already said that 2027 capex will significantly exceed 2026 levels. If the other hyperscalers follow, the $3 trillion figure may prove conservative, and the comparison to the US defense budget will only grow larger. Nvidia shares, trading at 23 times forward earnings, and Broadcom, with its 20-gigawatt custom chip pipeline, remain the two primary beneficiaries of the buildout.

This article is for informational purposes only and does not constitute investment advice.