The Big AI Question Is Who Ultimately Pays for All That Compute

The biggest question facing AI investors may no longer be how much compute can be built, but who ultimately pays for it.
Eva Ados says Anthropic has committed roughly $500 billion toward infrastructure while some of its largest customers are not signing comparable long-term contracts.
That creates a potential duration mismatch.
AI companies are making enormous multi-year commitments to data centers, chips and cloud capacity, but customer demand may remain much shorter-term and less predictable.
If utilization and revenue scale fast enough, that infrastructure can generate enormous returns.
But if enterprise demand grows more slowly than expected, companies could be left carrying very large fixed commitments before the corresponding revenue fully materializes.
Ados says this was a key reason $XOVR passed on investing in both Anthropic and OpenAI.
The concern is not whether AI demand exists.
It is whether future customer revenue will be durable enough to support the extraordinary amount of capital already being committed to compute.