Google Backs Anthropic's $5B Texas Data Center in AI Arms Race

Google is weeks away from confirming construction financing for a multibillion-dollar data center in Texas currently leased to Anthropic, according to Financial Times reporting. The deal could exceed $5 billion and includes construction loans for developer Nexus Data Centers. A consortium of banks is competing to finalize the debt package by mid-2026, sources told the FT.

Nexus is developing a 2,800-acre campus with a target of 500 megawatts of capacity by late 2026. The campus could expand to 7.7 gigawatts long-term, according to project documents. That initial 500MW phase would rank among the larger AI-focused data center builds in North America.

Meta's AI Research SuperCluster runs at around 200MW. This is more than double that scale, and it's just the first phase.

Part of Anthropic's $50B US Infrastructure Push

Anthropic revealed March 18 it would build custom data centers with Fluidstack in Texas and New York. The announcement outlined a $50 billion domestic infrastructure plan creating approximately 800 permanent jobs and 2,400 construction positions. The Google-backed Texas facility is the largest component of that program.

Dario Amodei, Anthropic's CEO and co-founder, told reporters in the March announcement: "We're getting closer to AI that can accelerate scientific discovery and help solve complex problems in ways that weren't possible before. Realizing that potential requires infrastructure that can support continued development at the frontier."

We're getting closer to AI that can accelerate scientific discovery and help solve complex problems in ways that weren't possible before. Realizing that potential requires infrastructure that can support continued development at the frontier.

— Dario Amodei, CEO and co-founder of Anthropic

Anthropic told investors it serves more than 300,000 business customers. The company said its number of large accounts, customers representing over $100,000 in annual revenue, "has grown nearly sevenfold in the past year," according to the statement. That user growth is driving the infrastructure buildout.

Bypassing Grid Delays with Direct Gas Supply

Nexus is sidestepping grid interconnection delays by using direct gas supplies for on-site power generation. This behind-the-meter approach has become increasingly common for AI data centers facing multi-year wait times for utility grid connections.

ERCOT, Texas's grid operator, told a state legislative committee in February that "data center power demand could double by 2030," straining existing capacity. Direct gas generation allows Anthropic to avoid that queue entirely.

Gary Wu, co-founder and CEO of Fluidstack, told the Financial Times: "Fluidstack was built for this moment. We're proud to partner with frontier AI leaders like Anthropic to accelerate and deploy the infrastructure necessary to realize their vision."

Fluidstack was built for this moment. We're proud to partner with frontier AI leaders like Anthropic to accelerate and deploy the infrastructure necessary to realize their vision.

— Gary Wu, co-founder and CEO of Fluidstack

Behind-the-meter gas generation bypasses the interconnection queue but raises emissions questions. Natural gas peaker plants typically run at 50-60% thermal efficiency. The carbon intensity is higher than grid power in regions with substantial renewable penetration. Texas isn't one of those regions.

Google's Deepening Financial Ties

Google has invested $5 billion in Anthropic across multiple rounds since 2023, according to regulatory filings. The tech giant uses Claude models in some of its enterprise products and has cloud infrastructure agreements with the company. This data center financing deepens that relationship further.

Sundar Pichai, Google's CEO, told analysts in the company's February earnings call: "We're pursuing multiple pathways to AI leadership, including investments in external partners." While developing its own Gemini models, Google maintains close ties with Claude's creator as a strategic alternative.

Microsoft has similar arrangements with OpenAI. Amazon Web Services invested $4 billion in Anthropic and, according to AWS's Andy Jassy, "hosts much of Anthropic's training workloads." The pattern across the AI industry is cloud providers and tech giants offering preferential infrastructure terms to frontier AI labs they've backed financially.

AI Infrastructure Arms Race Accelerates

Meta told shareholders in January it would spend "$60 billion on infrastructure in 2026." Microsoft and OpenAI announced plans for a "$100 billion Stargate supercomputer" in Texas, according to a January press release. Oracle, xAI, and Amazon are each pursuing gigawatt-scale AI facilities, according to commercial real estate data from CBRE.

Anthropic's Texas campus is part of that broader wave. Training frontier models like GPT-5, Claude Mythos, or Gemini 3 requires thousands of high-end GPUs running for months, according to published technical papers from OpenAI and Anthropic. Inference, serving those models to users, adds continuous compute load.

Gartner analysts told clients in a February briefing that "current infrastructure won't meet projected 2027-2028 demand." Texas offers deregulated energy markets, gas pipeline access, and land availability that California and Virginia lack. CBRE's data center leasing reports show the state is attracting a disproportionate share of new AI data center announcements.

Infrastructure as Competitive Moat

Google isn't just investing in Anthropic. It's underwriting the physical assets that enable the company to compete long-term. Path dependency kicks in when your primary investor finances the data center you operate in. Switching costs increase substantially.

Pure cloud partnerships offer more flexibility. Startups leasing AWS capacity can theoretically migrate to GCP or Azure. The switching cost is high but possible. Google financing your bespoke 500MW data center campus in Texas removes that option entirely.

Anthropic, OpenAI, and xAI have tech giant sponsors willing to commit billions to infrastructure. Smaller AI labs without deep-pocketed backers face widening gaps. Independent AI companies must compete for cloud capacity or raise infrastructure-specific capital at higher costs.

Benchmark Capital's Sarah Guo told TechCrunch in March: "The gap between frontier labs with infrastructure partners and everyone else is now unbridgeable. You can't catch up to GPT-5 or Claude Mythos without comparable compute resources. And you can't secure those resources without financial backing most startups don't have."

The Texas campus, if it reaches its 7.7GW expansion target, would consume more power than Austin's entire residential sector. That scale separates frontier AI labs from the rest of the market.

Google's financing of Anthropic's Texas data center is another data point in a clear trend: the companies that control compute will shape which AI models dominate the next decade. Infrastructure is the new moat.