Alpha Compute Corp closes 504-GPU NVIDIA B200 lease with frontier AI lab — Canadian compute capacity expands as Stargate rollout continues

Alpha Compute Corp announced a definitive lease agreement with a leading frontier AI laboratory for an enterprise-scale 504-GPU NVIDIA B200 deployment in its Canadian data center. The deal extends NVIDIA B200 inference capacity outside the major Stargate sites and signals continued enterprise demand for Hopper-successor compute.

The cell size (504 GPUs) is the detail to read. B200 cells of that size are deployed for production inference workloads rather than training — at typical FP4-precision throughput, 504 B200s service roughly 10-15M tokens/second for a 70B-class model, suitable for high-traffic API endpoints. The Canadian geographic placement (specific site not disclosed) gives the buyer redundancy outside US Stargate concentration and aligns with Canadian data-residency requirements for regulated enterprise customers.

Macro context: NVIDIA's FY2026 data-center revenue hit $194 billion (up 68% YoY), with B200 ramp continuing alongside Vera Rubin general availability targeted for H2 2026. Enterprise lease deals at the 500-GPU scale represent the demand-floor that determines NVIDIA's full-year shipping cadence — the segment that's growing fastest, not the marquee 100k+ Stargate deployments.

SEC EDGAR — Alpha Compute Corp Form 6-K FY2026 → · HeyGoTrade — NVIDIA vs TSMC vs Broadcom AI Chip Stocks 2026 →