Anthropic has committed to a six-year, $10 billion compute deal with Volta, a cloud infrastructure startup founded in early 2026 by former Brookfield Asset Management managers. The contract, reported by Bloomberg on August 4, is the latest sign that demand for AI compute is moving faster than traditional data center construction can keep pace.

Volta barely existed six months ago. The company raised $300 million in venture capital alongside the deal, led by Andreessen Horowitz and Altimeter Capital. Nvidia and Michael Dell also invested, giving Volta a $2.4 billion valuation before it has delivered a single rack of compute.

What the Deal Involves

The arrangement pairs Volta with Bitdeer, a crypto-mining-turned-data-center company, to build a 133-megawatt facility in Norway. The site will run on Nvidia’s next-generation Vera Rubin chip architecture. Delivery is split into two phases: target activation of December 31, 2026, and March 31, 2027.

J.P. Morgan and a second major financial institution are arranging $1.3 billion in credit support to back the contract. That level of structured finance, for a startup with no operating history, signals how urgently Anthropic needs to lock in future capacity.

Why AI Companies Are Signing With Startups

The Volta deal follows a now-familiar pattern. Anthropic previously signed large compute agreements with Google Cloud, AWS, and AMD. OpenAI committed billions to Microsoft Azure and to CoreWeave before its IPO. The difference here is the counterparty: Volta was not an established cloud provider but a purpose-built vehicle created specifically to capture AI demand.

That matters for two reasons. First, traditional hyperscalers are already constrained. Demand for GPU capacity from AI developers is running ahead of what Amazon, Microsoft, and Google can build fast enough. Second, startups like Volta, CoreWeave, and Lambda Labs have learned to move faster by focusing exclusively on AI workloads rather than managing legacy enterprise IT infrastructure.

Norway is a deliberate choice. The country has access to renewable hydroelectric power, cooler ambient temperatures that reduce cooling costs, and favorable permitting environments compared to the US and UK. Several European AI infrastructure projects have concentrated there for the same reasons.

What This Means for Business

For business leaders watching AI costs, the Volta deal points to two things happening at the same time: AI model providers are securing compute aggressively because they expect demand to keep rising, and the infrastructure layer supporting those models is becoming more distributed and competitive.

That competition should be good for enterprise customers over time. When compute capacity is scarce, model providers price accordingly. As infrastructure startups like Volta, together with expansion by established players, bring more capacity online through 2027, pricing pressure should ease.

In the near term, however, the supply crunch remains real. Enterprises running AI workloads at scale are still dealing with unpredictable costs and limited access to the most capable models. The compute deals being signed now are effectively forward contracts on capacity that will only start delivering in six months to a year.

For companies building AI-powered operations today, the practical implication is to design systems that can run on more than one model or provider. Single-vendor lock-in to an AI provider that itself depends on constrained infrastructure creates compounding risk. Organizations that treat model selection as a reversible decision, and build with abstraction layers between their applications and the underlying model, are better positioned to adapt as the infrastructure landscape changes.

The Bigger Picture

The Volta deal is not just about Anthropic. It signals how quickly the AI supply chain is being rebuilt from scratch. A company founded to meet demand from one AI lab, backed by venture capital and structured debt, building data centers on a different continent, using chips that shipped from Taiwan months ago. The whole chain has to execute on tight timelines simultaneously.

That infrastructure investment is what makes the current wave of AI capability sustainable. Enterprise DNA has written previously about the enterprise token cost crisis, where AI workloads became more expensive faster than budget cycles could adapt. Deals like Volta suggest the infrastructure response is underway, even if the timeline is 12 to 18 months out.

For Enterprise DNA clients deploying AI agents and voice employees inside their businesses, none of this changes what to build today. But it is useful context for why model pricing is volatile and why enterprise compute agreements, even at the provider level, look more like commodity futures than software licenses.

---

Omni by Enterprise DNA helps businesses build AI agent systems that are designed for resilience, not locked into a single provider. Learn more about Omni Ops.