Artificial intelligence developer Anthropic is reportedly negotiating another massive cloud computing contract with infrastructure provider Nscale, a transaction that could match the staggering 45 billion dollar price tag of their previous agreement signed just weeks ago.
According to a report from Bloomberg citing people familiar with the discussions, the two companies are exploring an additional 45 billion dollar compute commitment. The sources emphasized that negotiations remain underway and that terms could still shift before any formal announcement.
If finalized, the agreement would mark one of the most aggressive capital commitments in artificial intelligence history, coming barely a month after Anthropic first engaged the British data center startup for a multi-year compute expansion.
What happened?
In late August, Anthropic finalized an initial six-year, 45 billion dollar agreement to secure computing capacity from Nscale. That transaction granted Anthropic access to roughly 460 megawatts of power at Nscale's flagship Monarch data center campus in Mason County, West Virginia. The site, which was previously slated for Microsoft before the tech giant stepped away, is designed to run cutting-edge Nvidia Vera Rubin systems scheduled to come online in late 2027.
Now, Anthropic appears ready to double down on that relationship. While representatives for both companies have not officially commented on the ongoing talks, the reported secondary deal indicates Anthropic's relentless push to reserve every available megawatt of future computing power.
Over the past year, Anthropic has orchestrated an unprecedented series of infrastructure agreements. The company's expansion roadmap includes:
- A multi-gigawatt commitment spanning Amazon Web Services and Google Cloud infrastructure
- A 50 billion dollar computing arrangement with cloud provider Fluidstack
- A 10 billion dollar agreement with European AI cloud startup Volta for Norwegian data center capacity
- A 5 billion dollar hardware and compute deal with Advanced Micro Devices
- Specialized capacity agreements with SpaceX
Should this second transaction with Nscale close, Anthropic's total committed compute obligations would climb even further into historic territory.
The problem the company is trying to solve
Anthropic's massive spending is driven by two intersecting pressures: intense competitive rivalry and sheer infrastructure bottlenecks. The creator of the Claude AI model family has seen surging demand from enterprise and consumer clients over the past year. Earlier in the year, company executives acknowledged that rapid usage growth had led to severe strains on its infrastructure, occasionally degrading reliability and response times during peak demand hours.
Training frontier foundation models requires colossal clusters of graphics processors running continuously for months. Inference, the process of serving model responses to millions of active users and integrated enterprise software suites, consumes even more continuous electrical power.
To prevent rivals like OpenAI from outpacing its capabilities, Anthropic has chosen to contract compute years in advance. Rather than building data centers entirely on its own balance sheet, Anthropic has turned to specialized infrastructure builders like Nscale to construct the physical sites, procure the high-voltage electrical grid connections, install liquid cooling infrastructure, and secure allocations of next-generation hardware.
IPO ambitions meet investor skepticism
The timing of these negotiations is critical for both firms, as each company prepares for public market debuts.
Anthropic is widely expected to pursue an initial public offering in the coming months, seeking public market capital to support its ballooning operational footprint. Demonstrating secured access to power and next-generation silicon is crucial for convincing potential public shareholders that Anthropic can continue scaling its foundation models without running out of capacity.
For Nscale, which was founded only in 2024, securing back-to-back mega-deals validates its positioning as a Tier 1 AI cloud provider. However, the company faces its own headwinds. Reports surfaced recently indicating that Nscale has met with investor pushback regarding its proposed 30 billion dollar valuation for an upcoming listing. Skeptical investors have questioned the gap between early-stage actual cash revenue and the massive multi-billion-dollar contracted order backlogs that AI infrastructure startups are using to justify sky-high valuations.
More broadly, Wall Street has begun questioning the broader economics of the generative AI boom. While semiconductor manufacturers and infrastructure builders have booked record revenue, market analysts increasingly worry that foundational AI developers are committing hundreds of billions of dollars to capital expenditures without a clear, near-term path to commensurate profitability.
What happens next
If negotiations between Anthropic and Nscale culminate in a signed contract, the agreement will further cement the trend of foundation model labs behaving like industrial energy utilities. Securing nearly a gigawatt of combined power at dedicated campuses represents an enormous operational bet on the trajectory of AI adoption through the end of the decade.
Industry observers will watch closely to see whether Nscale can satisfy investor scrutiny ahead of its listing, and whether Anthropic's enormous capital commitments reassure or alarm public market investors as the company approaches its own IPO.
Filed under: AI, TechNews, Startups, Software, CloudComputing