Road Town, British Virgin Islands, May 14, 2026, 15:05 AST
- Alpha Compute submitted a Form 6-K on May 13, detailing a $32.2 million lease for AI compute over two years.
- It’s a 504-unit deal for Nvidia B200 GPUs in Canada, projected to generate $16.1 million in recurring revenue each year.
- Shares jumped, though the client hasn’t been identified and execution risks are still on the table.
Shares of Alpha Compute Corp. jumped Thursday, following news that the Nasdaq-listed AI infrastructure player inked a two-year, $32.2 million lease. The agreement, disclosed in a Form 6-K dated May 13, gives an unnamed frontier artificial intelligence lab access to 504 Nvidia B200 graphics processors at a Canadian data center. According to the filing, the contract announcement went out the previous day.
The agreement locks in steady income for Alpha Compute, which only recently repositioned itself as an AI compute player. The company now counts on $16.1 million in annual recurring revenue—essentially, what it expects to pull in each year from contracts. It’s also lined up for a $7.5 million advance, money paid upfront to hold compute slots.
The deal also hands the company a tangible selling point as rivals in the AI and software space vie for chips, power and coveted data-center capacity. Back in April, Alpha Compute dropped its old AlphaTON Capital Corp name, shifting focus toward compute and AI infrastructure. Its shares started trading as ALP on Nasdaq starting April 21.
Alpha Compute isn’t naming the customer. What’s clear: the lease involves a cluster of Nvidia B200 Tensor Core GPUs—chips designed for demanding AI training and inference workloads. The company says its Canadian facility runs on entirely hydroelectric power.
Chief Executive Brittany Kaiser hailed the contract as a “defining moment.” Yury Mitin, chief business development officer, called the 504-GPU rollout a “critical inflection point” in the company’s commercial strategy. GlobeNewswire
Alpha Compute hasn’t limited its push to North America. This week, The Economic Times reported the company is rolling out confidential GPU-as-a-service in India. Kaiser linked that move to India’s sizable Telegram user base driving demand. GPU-as-a-service lets customers tap into graphics-chip power via the cloud, skipping outright purchases; confidential computing, meanwhile, is designed to keep data protected during processing.
The fight for market share is fierce. CoreWeave and Nebius are now standing out among AI cloud firms, with Reuters noting a spike in “neocloud” demand from tech companies chasing AI hardware. Nebius disclosed this week that supply can’t keep up. CoreWeave, for its part, has landed sizable AI cloud contracts with Meta, Anthropic, and more. Reuters
Shares of Alpha Compute surged roughly 28% to 34.3 cents in the afternoon, MarketScreener data showed. Still, the stock remains down over 40% for the year. That kind of jump highlights just how sharply micro-cap AI infrastructure stocks can swing on fresh contract headlines—particularly when volumes pick up.
Still, a few flags to note. The customer’s identity isn’t disclosed, the deal hinges on a single deployment, and Alpha Compute itself highlights risks in its filing—everything from outside vendors and partners, to funding needs, to shaky ground with legacy ventures.
Those aren’t just hypothetical risks. Bigger players are pouring billions into power, chips, and locations—CoreWeave, for example, just bumped up the lower end of its 2026 capex target to $31 billion, according to Reuters. Nebius, too, is now planning to spend between $20 billion and $25 billion a year. For Alpha Compute, the task is clear: parlay one signed lease into a real customer base, all while dodging the same expense traps.
