Key terms of the SpaceX arrangement and the Meta exchange are not public, and the reported figures have not been officially confirmed by either company. Relief is also not expected soon. Micron's CEO said on its October 1 earnings call there is "no line of sight" to balancing memory supply, which he expects to stay much tighter in fiscal 2027 and 2028 than in 2026. One analysis puts relief no earlier than late 2027.
The squeeze affects every buyer, not just the largest. Hyperscalers are spending nearly $700 billion on AI infrastructure this year and placing open-ended memory orders. H100 rental prices have risen nearly 40% in six months, and cloud providers are demanding multi-year contracts with large upfront payments. Non-priority buyers face data center GPU lead times of 36 to 52 weeks. Attorneys advising AI companies, cloud customers, or investors should watch for compute-access terms in contracts, cross-company infrastructure arrangements, and any regulatory interest in how these alliances allocate scarce capacity.