Alpha Compute signs $81.7m GPU lease for Missouri B300 cluster
Alpha Compute Corp. (Nasdaq: ALP) has signed a binding four-year agreement worth $81.7 million to lease its ALPHA-03 GPU cluster to a venture-backed AI compute marketplace. The contract, executed on 9 October 2026, covers all 592 liquid-cooled NVIDIA B300 GPUs housed at a 1.3-megawatt Missouri data centre facility, with delivery targeted for early Q1 2027.
The unnamed customer, which Alpha Compute says was valued at more than $1 billion in a recent funding round and serves over 100 AI clients, has committed to the full 48-month term on a take-or-pay basis: fees are due whether or not the capacity is used. The total contract value is $81,732,160, generating approximately $20.2 million in annual recurring revenue. The customer has agreed to pay 25% of the contract value in advance, split as $16.1 million at signing and a further $5.1 million on delivery, on top of a $1 million booking deposit already received.
Deal structure and cluster details
ALPHA-03 is Alpha Compute's third contracted cluster and its first on US soil, following ALPHA-01 in Canada and ALPHA-02 in Sweden. The cluster comprises 576 GPUs covered by the commercial agreement plus 16 hot spares, all liquid-cooled under a signed data-centre capacity arrangement. Alpha Compute said the build will be funded through a combination of customer advance payments and asset-level, non-recourse GPU financing, limiting corporate balance-sheet exposure.
Chief executive Brittany Kaiser described ALPHA-03 as "the largest contract so far" for the business, while chief business development officer Yury Mitin noted that the 25% advance equates to a full year of contract fees, providing meaningful cash-flow certainty ahead of delivery.
Market context and competitive read-across
The deal sits within a rapidly expanding segment: turnkey GPU colocation and leasing, where infrastructure operators acquire accelerator hardware and sell compute access to AI developers at contractual rates. The model competes directly with hyperscaler on-demand capacity from AWS, Azure and Google Cloud, as well as pure-play GPU cloud providers such as CoreWeave and Lambda Labs, which have pursued similar take-or-pay structures to secure long-term revenue predictability.
NVIDIA's B300 architecture, part of the Blackwell generation, is among the highest-specification accelerators currently available, and securing 576-unit block allocations at this tier represents a meaningful supply-chain commitment. GPU availability has remained constrained across the industry, making advance deposits and binding off-take agreements a common mechanism for both suppliers and customers to manage risk.
Risks and regulatory considerations
Alpha Compute's forward-looking statements section flags concentration risk explicitly: the Missouri cluster is committed to a single unnamed customer, and any payment default or early termination would materially affect revenue. The customer's identity and its own financial health are not independently verifiable from the release, which describes it only by claimed valuation and approximate client count.
The company's use of the term "Sovereign Superintelligence" in its branding is positioned as appealing to finance, defence, intelligence and media clients seeking secure, confidential computing. That positioning places Alpha Compute adjacent to regulatory scrutiny: US government procurement of AI infrastructure and export-control rules under the Bureau of Industry and Security (BIS) remain active policy areas, particularly for advanced Blackwell-generation accelerators. Alpha Compute has not disclosed whether the Missouri facility has sought or holds any government security clearance relevant to its defence and intelligence market claims.
Investors will watch for confirmation of the advance payments being received, GPU delivery and commissioning milestones in Q1 2027, and whether Alpha Compute can sign additional clusters to diversify its customer base beyond the current three-site footprint.