Axe Compute signs $1.5B dedicated Nvidia cluster deal, lifts 2026 contracts past $3B
Neocloud provider Axe Compute signed a $1.5 billion, five-year contract for a dedicated Nvidia Blackwell B300 GPU cluster, bringing its total 2026 signed contract value past $3 billion. The deal includes $534 million in customer prepayments and is expected to nearly double ARR to $696 million upon deployment.
Context from: Investors | Axecompute
The decision it puts on your desk
If your company spends more than $5 million annually on GPU cloud, benchmark a dedicated neocloud build against your hyperscaler contract by Q4 2026. Dedicated Blackwell clusters are now available with five-year terms, hardware refresh guarantees, and sub-20ms latency to your data. Your hyperscaler is a shared pool with noisy neighbors. The dedicated alternative is no longer theoretical. It is public, it is prepaid, and it is scaling fast. One loose thread: Axe Compute did not disclose the customer. If it is a single enterprise committing $1.5 billion, that is a concentrated bet. If it is a GPU aggregator or broker reselling capacity, the revenue quality is different. The name matters.
Axe Compute signed a $1.5 billion, five-year contract for a dedicated Nvidia Blackwell B300 GPU cluster, the neocloud provider said Sunday. The deal includes 9,200 GPUs and lifts 2026 signed contracts past $3 billion.
The customer name was not disclosed. Axe Compute will design, deploy, own, and operate the full cluster through its Build program. The company expects over $534 million in customer prepayments within the next 30 days.
This is the second major announcement in a week. On July 22, Axe Compute disclosed $1.3 billion in new contracts across the United States and Europe, blowing past its $1 billion annual target with five months to spare.
The combined contracts are expected to push annual recurring revenue to $696 million upon deployment, nearly double the $385 million run rate reported earlier this month.
"We are focused on continuing that pace through the end of 2026 and into 2027," CEO Christopher Miglino said. The company could close another $2 billion in signed contracts this year, he said.
The Build model is the key differentiator. Axe Compute selects GPU types, locations, and configurations for each customer. It owns the hardware. The customer gets dedicated single-tenant infrastructure without putting data center assets on its balance sheet.
Gartner projects worldwide AI spending will pass $2.5 trillion in 2026, with hyperscaler capex alone reaching $660 billion to $690 billion, according to the Futurum Group. The spending is there. The question is who captures it.
Axe Compute trades under NASDAQ: AGPU. The stock has followed the broader AI infrastructure rally. The $3 billion contracted backlog, with prepayments funding the buildout rather than equity dilution, gives it a capital-efficient growth story that the larger hyperscalers cannot match.
One loose thread: Axe Compute did not disclose the customer. If it is a single enterprise committing $1.5 billion, that is a concentrated bet. If it is a GPU aggregator or broker reselling capacity, the revenue quality is different. The name matters.
Source
← Previous
Nvidia launches Open Secure AI Alliance with 50+ members. OpenAI, Google, and Anthropic sit out.
Next →
OiiOii AI launches multi-agent animation studio in the US. 100,000 makers wanted in. 100 got in.
The Morning Recap
We sift the day. You make the calls.
The decision on every story, in your inbox before your first call.