A $32.2 million contract for just 504 GPUs highlights the intense scarcity and high value of AI-ready data center capacity outside the hyperscaler ecosystem.
A $32.2 million contract for just 504 GPUs highlights the intense scarcity and high value of AI-ready data center capacity outside the hyperscaler ecosystem.

Alpha Compute Corp. has secured a $32.2 million, two-year contract to supply a frontier AI laboratory with 504 of Nvidia’s newest B200 GPUs, a deal that shows the surging demand for specialized, high-performance computing infrastructure beyond traditional cloud providers.
"This contract is a defining moment for Alpha Compute, confirming our belief that there is a massive global appetite for sovereign, specialized AI infrastructure," Brittany Kaiser, CEO of Alpha Compute, said. "Finalizing a $32.2 million deal with a leading AI research firm is a testament to our technical excellence, our robust Canadian data center presence, and the trust our clients have in our enterprise-grade delivery."
The agreement, announced May 12, carries a total contract value of $32.2 million, translating to $16.1 million in annual recurring revenue (ARR). The deal includes a $7.5 million upfront payment to reserve the dedicated cluster of 504 NVIDIA B200 Tensor Core GPUs, which are housed in the company’s Canadian data center powered entirely by hydroelectric energy.
For Alpha Compute (Nasdaq: ALP), the contract is a critical commercial inflection point, supporting its strategy of providing sovereign, privacy-focused GPU-as-a-Service (GPUaaS). It also signals a robust, high-value market for NVIDIA's latest hardware among advanced AI researchers who operate outside the massive data centers of hyperscalers like Amazon and Google.
The deal arrives as demand for artificial intelligence hardware, particularly the advanced GPUs designed by Nvidia, continues to far exceed supply. Nvidia, now the world's largest company by market capitalization, has seen its valuation soar on the back of its data center business. The B200 GPU, based on the new Blackwell architecture, offers a significant performance leap over its predecessors, making it the preferred tool for training the next generation of large-scale AI models. This intense demand has created a significant opportunity for a new class of infrastructure providers who can secure and lease this scarce hardware.
Alpha Compute is one of several emerging companies building businesses around providing this high-demand infrastructure. Its focus is on "sovereign" and "confidential" computing, appealing to clients who require privacy or need infrastructure outside of traditional hyperscaler environments. While this first major deal is significant, it also highlights the current scale of such independent providers. For context, competitor Bitdeer, which is also transitioning from crypto mining to AI infrastructure, reported its AI cloud business reached an annualized revenue run-rate of approximately $69 million in April with 4,184 GPUs deployed.
Alpha Compute's contract secures a per-GPU revenue stream that is multiples higher than the broader market, reflecting the premium value of both the new B200 hardware and the specialized, private access model. As both Alpha Compute and peers like Bitdeer continue to build out their AI data center capacity—with Bitdeer converting a site in Tydal, Norway, and Alpha Compute expanding in North America and Europe—they are establishing a new, high-margin tier in the AI infrastructure market.
This article is for informational purposes only and does not constitute investment advice.