TeraWulf touts $19B Anthropic lease to cement exit from bitcoin mining
The Kentucky AI campus tied to the win is expected to begin coming online in 2028, extending the timeline risk for traders.
TeraWulf CEO Paul Prager is framing a 20-year AI hosting lease with Anthropic, valued at roughly $19 billion over its life, as the clearest signal the company is no longer a bitcoin miner. The first capacity for the Kentucky project that won the bid is expected to come online beginning in 2028, leaving a multi-year gap between contract narrative and delivered megawatts.
Key Takeaways
- A 20-year AI hosting lease with Anthropic is valued at roughly $19 billion over its life, and management says the figure exceeds TeraWulf’s current market capitalization.
- The Kentucky project secured Anthropic through a competitive bid where grid power access and long-term infrastructure were central.
- Initial Kentucky capacity is expected to come online beginning in 2028, with Fluor hired to help construct the facility.
- TeraWulf says bitcoin mining is no longer part of its long-term strategy as it sells non-core assets to fund wholly owned AI infrastructure, including additional eastern Kentucky sites.
A $19B Anthropic Lease as the Proof Point for TeraWulf’s AI Pivot
TeraWulf is using the headline size of its Anthropic agreement as a valuation anchor. CEO Paul Prager described the AI hosting agreement as a 20-year lease valued at roughly $19 billion over its life, and said that figure exceeds the company’s current market capitalization.
That framing matters for how the stock gets traded. A miner is typically priced like a cyclical exposure to a commodity-driven revenue model. A long-duration hosting lease invites a different mental model: contract-backed infrastructure with the potential for more predictable long-term cash flows.
Prager made the pivot explicit. “We’re not involved in Bitcoin,” he said, adding that bitcoin mining is no longer part of TeraWulf’s long-term strategy. He contrasted bitcoin’s price-linked economics with AI infrastructure as a more natural fit for the business.
How Kentucky Won the Bid: Grid Power and Long-Term Infrastructure
Prager said the Kentucky project won Anthropic through a competitive bidding process centered on access to grid power and long-term infrastructure. The subtext is that the bottleneck is not land or branding, it is power that can be delivered reliably at scale.
He warned investors that “not all megawatts are created equally.” In Prager’s telling, hyperscale AI facilities are constrained by power quality and reliability, not just raw site size. He said the U.S. faces a shortage of electricity and argued that successful AI campuses require reliable generation, redundant transmission, favorable regulation, and strong community relationships.
For traders comparing AI-hosting contenders, that shifts the moat discussion from who has the biggest megawatt headline to who can actually clear interconnection, deliver uptime, and operate inside local regulatory and community constraints.
From Lake Mariner to Kentucky: The Existing Anthropic and Google Relationship
TeraWulf is not pitching Kentucky as a cold start. Prager said the company already works with Anthropic and Google at its Lake Mariner campus in New York, giving the parties an established operating relationship ahead of the Kentucky buildout.
That existing footprint can help credibility, but it does not solve the core modeling problem. The excerpt does not specify whether the roughly $19 billion figure reflects contracted revenue, includes pass-through power costs, or represents another measure of contract value.
2028 Go-Live and Build Execution: Fluor’s Role and the Multi-Year Timeline
The Kentucky AI facility is expected to come online beginning in 2028, and TeraWulf has hired Fluor to help construct the project. Prager described AI data center development as multi-year and said securing skilled labor and contractors is a bigger challenge than equipment procurement as hyperscale builds become more specialized.
That timeline is the market structure issue. With first capacity not expected until 2028, near-term price action is more likely to be narrative-driven than cash-flow-driven until investors get clearer milestones on permitting, interconnection progress, and construction execution.
Marcus Hale’s Take: Contract Narrative vs. Timeline Risk for Traders
I see TeraWulf trying to re-rate itself from “miner beta” to “infrastructure multiple” by pointing to a 20-year, roughly $19B lease that management says is bigger than the company’s market cap. That can work as a framing device, but only if the market gets enough contract detail to underwrite economics instead of just duration.
The threshold that matters is whether the company can turn the Kentucky win into a sequence of verifiable build milestones well ahead of 2028, because until then this looks more like a sentiment catalyst than a fundamental shift, and the practical difference is whether traders can model cash flows rather than trade headlines.
