Why Anthropic Is Committing $44.6 Billion to Nscale Before Its Nasdaq Listing

Nscale's reported $44.6 billion compute deal with Anthropic in West Virginia is the latest and largest node in a global infrastructure buildout backed by $518 billion in reported aggregate commitments, as the company readies a $2 trillion Nasdaq listing.

The scale of Anthropic's infrastructure ambitions came into sharper focus this month when Nscale was reported to be finalizing a $44.6 billion compute agreement with the company for a facility in West Virginia — and that may represent only part of the picture. A second report described Nscale shopping an additional deal, reportedly worth approximately $45 billion, anchored at its Monarch data center campus. Taken individually, either arrangement would rank among the largest single compute procurement commitments on record; together they underscore a capital intensity that few AI-native companies have attempted in a single deployment cycle, and that sets the material context for Anthropic's widely anticipated push toward a public listing.

Those two reported Nscale deals are the largest disclosed nodes in what has become a genuinely global infrastructure programme. According to reports, NTT is in negotiations described as having entered a “final stage” to construct a $6.3 billion data center for Anthropic in Japan. Akamai reportedly secured a seven-year, $11.6 billion cloud contract with Anthropic, with options to scale the arrangement to $20 billion. Anthropic signed its first Canadian data center lease with EverWind, committing to 95 megawatts at the developer's Nova Scotia campus — an agreement described as the largest private investment in that province's history. A $366 million data center linked to Anthropic was filed for permitting in Cedar Creek, Texas. On the silicon side, Broadcom unveiled what it described as a major AI-chip financing plan for Anthropic — Broadcom shares rose 5.8 percent following that announcement, according to Yahoo Finance — with one analysis placing the potential financing figure at $42 billion; separately, reports described Anthropic planning custom AI accelerators on Samsung's 2-nanometer process node. TipRanks noted that Anthropic has disclosed $518 billion in aggregate AI infrastructure commitments, a figure that, even accounting for multi-year optionality and reporting conventions, situates the company's buildout in a category of its own.

That infrastructure push is the physical foundation of an IPO that multiple reports characterize as imminent. Anthropic has reportedly selected Nasdaq as its listing venue, with Morgan Stanley, Goldman Sachs, and JPMorgan named as lead underwriters, and a valuation target described variously across sources as $1 trillion to $2 trillion. A prospectus that at least one outlet characterized as having “leaked” depicted growing revenue alongside steep losses but also, notably, recent operating profits — a datapoint that, if confirmed in the final filing, would materially distinguish Anthropic from most of its frontier-model peers at the time of listing. The Information analyzed the economics of Anthropic's data center sprint as “attractive,” and SemiAnalysis estimated that the inference business could reach 88 percent gross margins — a figure that, if the buildout achieves the intended scale, would justify the magnitude of upfront commitment. OpenAI is reportedly racing to raise $30 billion ahead of the Anthropic listing, a dynamic that itself speaks to how consequential the offering is expected to be for the broader funding landscape.

Not all signals point in one direction, and scrutiny is warranted on several fronts. Financial Newswire raised circular financing concerns tied to the $2 trillion IPO target, questioning whether related-party capital structures could complicate the valuation. One analyst was quoted describing it as the “most ridiculous IPO” of the year. Michael Burry cautioned publicly this month that much AI capital spending may ultimately prove to be sunk cost — a warning that carries particular weight when the company in question has reportedly declared $518 billion in infrastructure commitments. The IREN episode illustrated the volatility of rumor-driven dynamics in Anthropic's supplier ecosystem: shares rose on reports of a multi-year cloud deal, then fell 5 percent as the contract remained unconfirmed — investor confidence shaken, per one account, precisely by the absence of official corroboration. On the competitive side, DeepSeek released its open-source DeepInfer inference engine this month, cited in at least one report as directly pressuring OpenAI and Anthropic by lowering inference cost and energy consumption; Nikkei Asia further noted that China's frontier-model developers launched 16 new base models in a single month, demonstrating a supply-chain resilience that complicates simple assumptions about durable Western compute advantages.

Anthropicís operational posture is not static in the face of these pressures. Claude's orchestration layer was extended this month to support up to 1,000 parallel sub-agents within a single dynamic workflow — a capability with direct relevance to enterprise deployments where throughput compounds pricing power. Hitachi joined Anthropic's critical-infrastructure defence programme, broadening its industrial partnership base. Broadcom, in an analyst note reported by Stocktwits, forecast $230 billion in AI revenue by 2028 driven substantially by Anthropic and OpenAI workloads, while acknowledging that certain risks remain unresolved. Three concrete signals will determine how the next chapter reads: whether the reported Nscale agreements receive formal contract and financing confirmation, closing the gap between headline and executed deal; how the IPO book-build ultimately prices the tension between SemiAnalysis's 88 percent inference-margin thesis and the $518 billion in declared infrastructure obligations flagged by TipRanks; and the rate at which open-source inference tools such as DeepSeek's DeepInfer compress industry cost floors — because if inference economics erode faster than Anthropic's compute contracts amortize, the financial logic underpinning this buildout will face its most rigorous test not in a data center, but in the quarterly reporting cycle that follows a public listing.