Anthropic Locks In $19B TeraWulf Lease as Infrastructure Push and IPO Path Converge
A 20-year, $19 billion data center commitment in Kentucky anchors Anthropic's vertical integration strategy even as a looming Nasdaq debut, unresolved geopolitical exposure, and warnings of an AI infrastructure bubble add complexity to the picture.
On July 7, 2026, Anthropic signed a 20-year, $19 billion lease with TeraWulf (NASDAQ: WULF) for a 401-megawatt dedicated AI data center campus in Hawesville, Kentucky, sending WULF shares up more than 10 percent in a single session. The facility is slated to come online in 2027, and multiple sources characterize it as the largest dedicated cloud partnership for AI model training ever executed — a commitment measured in decades and tens of billions rather than the rolling annual contracts that have defined prior hyperscaler relationships. TeraWulf, a company that pivoted from Bitcoin mining to high-performance compute infrastructure, announced plans to raise $3.5 billion in debt through a Morgan Stanley-led offering to fund construction. At least one financial analysis outlet has flagged questions around the mechanics of that payment structure, a caveat worth holding onto ahead of any formal close.
The Kentucky lease is the most visible piece of a broader infrastructure accumulation that Anthropic has undertaken at considerable speed. The company announced a $50 billion domestic investment pledge encompassing data centers in Texas and New York, and separately disclosed plans for up to $15 billion in Australian data center capacity targeting as much as 1.4 gigawatts of power in that market. Concurrently, Anthropic is reportedly in advanced talks with Samsung to develop a custom 2-nanometer AI accelerator, and has entered a collaboration with Micron to co-design HBM, DRAM, and storage optimized for Claude-scale inference workloads. Taken together, the pattern reflects a deliberate push toward vertical integration: locking in the compute substrate rather than leasing capacity at market rates from third-party providers — and doing so at a scale that transforms long-term lease obligations into a fixed-cost structure with decade-long horizons.
These capital commitments arrive at a consequential inflection point. Anthropic has filed a draft S-1 and is targeting a Nasdaq listing as early as October 2026. According to reporting by Fortune, the company has reached an annualized revenue run rate of approximately $47 billion — a figure that would, if accurate, mark the first time it has overtaken OpenAI on that metric. Anthropic closed a $65 billion Series H at a valuation of roughly $965 billion, placing it within reach of a $1 trillion market cap; Polymarket prediction markets have assigned an 84.5 percent probability of that milestone being reached within 18 months. The credibility of that revenue figure matters enormously: UBS has maintained a buy rating on Broadcom partly on the strength of AI ASIC demand attributed to Anthropic and OpenAI, a signal that the company's procurement decisions now move public markets in adjacent sectors.
The past twelve months have also surfaced material risks that investors and counterparties will be required to price. In late June and early July, the U.S. Commerce Department imposed and then lifted export restrictions on Claude Fable 5 and Mythos — a three-week freeze that temporarily disrupted international access to the company's flagship models before being reversed. A separate and more structurally significant controversy erupted when hidden code in Claude Code was found to flag interactions originating from Chinese users; Alibaba subsequently banned the tool and directed employees to a domestic alternative, while Anthropic removed the offending code and acknowledged the episode. The incident exposed a durable tension: Anthropic's safety-first positioning and its deepening alignment with U.S. regulatory priorities sit uneasily alongside any aspirations in the Chinese market. Former White House economists have also cautioned publicly that the AI infrastructure buildout remains a bubble that is, in their words, "still inflating" — a systemic warning with particular relevance to long-dated fixed commitments like the TeraWulf deal.
The risk-opportunity calculus heading into Anthropic's public market debut is therefore unusually wide. On the opportunity side, a corroborated revenue trajectory, an expanding infrastructure footprint across three continents, and simultaneous custom silicon development suggest a company executing across multiple dimensions at once. An Elon Musk endorsement and deepened SpaceX compute partnership, while limited in public disclosure, add an unexpected dimension to the demand picture. On the risk side, the 20-year TeraWulf commitment is a long-dated fixed obligation being underwritten on current demand assumptions, the $3.5 billion construction financing has not yet closed, and the company carries unresolved geopolitical exposure following the China code episode. Three signals will define the near-term chapter: whether TeraWulf successfully closes its $3.5 billion debt raise and breaks ground on the Kentucky campus on schedule; how Anthropic prices its IPO relative to its $965 billion private valuation in a market still absorbing AI infrastructure risk premiums; and whether the Samsung chip partnership produces a committed fabrication agreement — a development that would mark a qualitative shift in the company's hardware independence and, by extension, its long-run cost structure.