Anthropic expects to spend at least $518 billion on cloud computing, chips and data-center infrastructure over the next decade. The figure comes straight from its confidential IPO prospectus. And roughly 80% of those commitments cannot be canceled or must be paid regardless of actual usage.

The numbers stun. Google stands to receive at least $111.1 billion from April 2026 through July 2033. Amazon gets $110 billion between May 2026 and April 2036. Microsoft claims $31.4 billion from November 2026 to May 2033. Add in $161.2 billion in largely non-cancelable equipment leases tied to Broadcom. The sums add up fast. These take-or-pay contracts turn variable costs into near-fixed obligations.

Compute scarcity drives the strategy. The prospectus states future demand for advanced AI systems will likely exceed available supply. Development will be “limited principally by the availability of compute.” No hedging. No speculation. Just a blunt assessment that chips and power now form the real bottleneck.

Last year the company spent $7.33 billion on compute and infrastructure. Revenue reached $4.6 billion in 2025, up twelvefold from the prior year. By July 2026 its annualized revenue run rate had climbed above $65 billion. Growth looks explosive. Yet the infrastructure bill dwarfs everything that came before.

Anthropic has locked in at least 14.8 gigawatts of capacity since October 2025. That tally comes from Reuters, which reviewed the draft filing. The deals span hyperscalers, specialized providers and chip suppliers. Fluidstack. Nscale. Lambda. SpaceX. Each adds another piece to a sprawling bet that AI models will keep scaling.

But the contracts carry real bite. “If our actual spend falls short, we must pay Google the difference,” the prospectus reads. Similar language applies to Amazon. The Microsoft commitment is non-cancelable except in the event of an uncured material breach. Flexibility exists only in certain arrangements with xAI. Those could reach $84.5 billion through 2029 yet remain largely cancelable with 90 days’ notice.

Suppliers double as competitors. Amazon, Google and Microsoft invest in Anthropic, provide its cloud capacity, and field their own frontier models. The filing flags the risk. Commercial tensions could limit access to compute at critical moments. Dependence runs deep. Nearly half of 2025 sales flowed through Amazon and Google channels.

Wall Street has taken notice. A $60 billion financing package tied to Anthropic’s chip needs surfaced in recent discussions on X. Banks appear ready to back the buildout with debt structures that blend senior and junior tranches. The move signals confidence. It also highlights how capital markets now finance AI infrastructure at sovereign scale.

Today Anthropic expanded its Cyber Verification Program, giving vetted U.S. security teams broader access to its most powerful models with reduced safeguards. The initiative, reported by Bloomberg and Reuters, builds on Project Glasswing. Partners uncovered more than 129,000 software vulnerabilities in just a few months. The timing underscores a dual reality. While the company locks in massive compute for training, it also deploys those models to harden critical systems.

Revenue concentration raises another flag. Two unnamed customers accounted for nearly a quarter of 2025 sales. Many large customers lack long-term contracts and could reduce spending. Losses mounted. The firm posted more than $8 billion in operating losses last year. Cash and short-term investments stood at $20.3 billion at year-end. The $518 billion commitment exceeds that buffer by more than 25 times.

Yet the bet looks rational inside the current arms race. Dario Amodei once cautioned against “crazy deals” and premature scaling. Demand changed his calculus. Claude’s rapid adoption forced a shift from caution to aggressive capacity securing. Similar dynamics play out at OpenAI. Industry executives now view compute as the primary moat.

AMD committed to supply more than $20 billion in capacity and agreed to buy up to $5 billion of Anthropic stock. The tie-in deepens the silicon supply chain. Memory partners such as Micron, Samsung and SK hynix also appear in strategic announcements. The infrastructure web grows denser by the quarter.

Power constraints loom larger than chips alone. Securing 14.8 gigawatts demands grid capacity equivalent to multiple nuclear reactors. Permitting, transmission and local opposition could delay delivery. Contracts guarantee payment. They do not guarantee electrons flow on schedule. That mismatch creates execution risk few balance sheets have faced.

Investors eyeing the eventual IPO confront a simple equation. Anthropic wagers that AI-generated value will outrun the fixed costs now locked in. Revenue run-rate growth offers encouragement. Model performance gains keep coming. Enterprise adoption accelerates. But the math requires sustained, massive uptake to service obligations that average more than $50 billion annually when spread over a decade.

The prospectus paints a picture of disciplined aggression. Take-or-pay deals secure priority access in a seller’s market. Non-cancelable leases guarantee chip supply from Broadcom-linked sources. Diversification across providers reduces single-point failure. Still, the scale feels historic. Few companies outside energy or defense have shouldered infrastructure commitments this large this early in their public lives.

Recent X conversations reflect the tension. One analyst noted the commitments equal roughly 30% of a potential $2 trillion valuation. Others pointed to OpenAI’s even larger estimated obligations. The industry, it seems, has accepted that frontier AI demands bets few traditional metrics can justify today.

Anthropic’s shift toward owning more of its stack shows in newer deals. Direct leases, custom silicon exploration and data-center partnerships with Fluidstack signal a move beyond pure cloud reliance. That evolution may improve margins over time. It also adds capital intensity in the near term.

So the company hurtles forward. Massive locked-in spend. Rapid revenue ramp. Model releases that expand defensive capabilities even as they fuel commercial growth. The IPO, whenever it arrives, will test whether public markets endorse this vision at the implied valuation. For now the prospectus leaves little doubt. Anthropic believes compute scarcity defines the next decade of AI. It has placed the largest bet yet to make sure it never runs short.