First-order effects

• Anthropic is committing to compute through leases rather than placing comparable hardware spending on its own capex line, leaving it with $782 billion of long-term contractual obligations while retaining a relatively asset-light balance sheet.
• Culpan's review identifies Broadcom, Google and SpaceX among major lease counterparties, concentrating the immediate benefit of Anthropic's demand in infrastructure providers rather than in Anthropic-owned equipment.

Second-order effects

• The lease structure makes utilization and revenue conversion central to Anthropic's economics: large contracted capacity must be monetized while the company is also reducing LLM prices.
• Infrastructure providers that can finance and supply long-duration capacity gain an advantage over vendors selling equipment outright, because customers can shift spending from upfront capex into contractual commitments.

Third-order effects

• If comparable arrangements spread, AI investment will be judged less by a model developer's reported capex than by the lease, credit and capacity commitments supporting its compute access.
• The pattern points to AI infrastructure becoming a financing market as much as a hardware market, with risk migrating from model developers' owned assets to long-term obligations shared with suppliers and capital providers.