OpenAI and Anthropic are signing artificial intelligence infrastructure agreements worth hundreds of billions of dollars, even though neither company needs to hand over the entire amount when
OpenAI and Anthropic are signing artificial intelligence infrastructure agreements worth hundreds of billions of dollars, even though neither company needs to hand over the entire amount when the contracts are announced.
The explanation lies in how long-term cloud computing contracts work. Instead of purchasing every GPU and data center outright, AI developers agree to buy computing services over multiple years, allowing infrastructure providers to spread construction costs and payments across time.
But these enormous commitments raise an important question: what happens if future AI revenue cannot cover the bills?
How Do $100 Billion AI Contracts Actually Work?In April 2026, Anthropic announced a commitment to spend more than $100 billion on Amazon Web Services over ten years, securing access to as much as five gigawatts of computing capacity.
Microsoft separately disclosed that OpenAI had contracted to purchase an additional $250 billion in Azure services. Neither announcement means the full headline amount was immediately transferred in cash.
These agreements generally establish future spending obligations, capacity reservations and commercial terms. Payments may occur as infrastructure becomes available, services are delivered or contractual milestones are reached.
For example, a hypothetical $100 billion contract covering ten years would average $10 billion annually if spending were evenly distributed. Actual payment schedules can differ substantially, and some contracts may require deposits, prepayments or minimum purchase commitments.
AI agreement Announced value What it represents Anthropic–Amazon Over $100B AWS technology spending over ten years OpenAI–Microsoft $250B additional Contracted future Azure services CoreWeave customer backlog About $104B Contracted future revenue, subject to fulfillment Who Pays to Build the AI Data Centers?The cloud provider or infrastructure operator often finances equipment and construction before collecting most of the expected customer payments.
Companies such as CoreWeave use debt, equity and GPU-backed financing to expand capacity. Long-term customer agreements can strengthen the case for lending because they provide evidence of future demand.
However, a signed contract does not eliminate financial risk. CoreWeave's approximately $104 billion revenue backlog accompanied $640 million in quarterly net interest expense, illustrating how financing costs can remain substantial even when contracted demand is strong.
The wider question of data-center ownership also matters. The company building a facility, the institution lending against it and the AI developer buying its computing power may be three different businesses.
What Happens if an AI Company Cannot Pay?The consequences depend on the contract. Some agreements contain minimum spending requirements or take-or-pay provisions, meaning customers may owe money even when they use less computing capacity than expected.
Others include termination rights, performance conditions or negotiated remedies. A headline contract value alone does not reveal how much spending is unconditional or legally enforceable.
If an AI customer fails to pay, infrastructure operators could face revenue shortfalls while still owing lenders and equipment suppliers. Their GPU collateral may then become central to restructuring negotiations.
For investors, the distinction is crucial: contract value is not revenue, revenue is not cash flow, and future commitments are not guaranteed profits.
Ultimately, $100 billion AI deals are possible because payment obligations can extend across years. Whether they become profitable depends on how much computing capacity customers actually use, when payments arrive and whether AI businesses generate enough income to support their commitments.
