Google expects AI budget to rise amid capacity shortages, with capex doubling to $190B

Alphabet just told the world it plans to spend between $175 billion and $190 billion on capital expenditures in 2026. For context, that’s roughly double the $90 billion it spent in 2025. The reason is almost comically simple: Google literally cannot build AI infrastructure fast enough to keep up with demand.

CEO Sundar Pichai has flagged persistent compute constraints throughout 2026, meaning the company’s AI ambitions are currently bottlenecked by physical reality.

The numbers behind the AI arms race

Alphabet isn’t just dipping into its existing cash pile to fund this expansion. The company has initiated equity raises totaling between $80 billion and $85 billion, a financial maneuver that underscores how capital-intensive this race has become.

One particularly notable piece of that funding puzzle: a $10 billion commitment from Berkshire Hathaway in June 2026.

Google isn’t alone in this spending frenzy. The combined AI-related capital expenditure guidance for major hyperscalers, including Amazon, Microsoft, and Meta, is projected at $650 billion to $725 billion for 2026. That represents a 36% to 77% year-over-year increase depending on where the final numbers land.

The capacity crunch is already creating real-world consequences. In March 2026, Google reportedly could not fulfill all of Meta’s Gemini inference capacity requests, disrupting Meta’s project timelines.

What this means for the compute supply chain

The infrastructure buildout Alphabet is pursuing spans data centers, tensor processing units (TPUs), and the sprawling supporting ecosystem required to train and serve AI models at scale. Google DeepMind and cloud-based AI services are the primary drivers of demand.

Alphabet has indicated it expects compute shortages to persist throughout all of 2026, which means even after doubling its spending, the gap between demand and supply likely won’t close this year. That raises legitimate questions about growth trajectories beyond 2027 and whether the returns on these massive investments will materialize on a timeline that satisfies increasingly impatient shareholders.

The crypto and market angle

The most direct link is energy. Bitcoin mining and AI data centers compete for many of the same resources: cheap electricity, favorable regulatory environments, and physical space with adequate cooling. Companies like Core Scientific and Iris Energy have explored or executed deals to repurpose mining capacity for AI workloads.

There’s also a decentralized compute angle. Projects building distributed GPU networks, think Render, Akash, and similar protocols, position themselves as alternatives to centralized cloud providers. When Google can’t even fulfill requests from Meta, the value proposition for decentralized compute networks becomes easier to articulate.

The risk, as always, is execution. Alphabet’s investor base is already scrutinizing whether these expenditures will translate into proportional revenue growth. If the AI revenue story stumbles, the knock-on effects could dampen enthusiasm across the entire compute sector, including the crypto projects riding the same wave.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.