TeraWulf Inc. (NASDAQ: WULF) priced an upsized common stock offering of 47.4 million shares at $19.00 per share on April 15, raising approximately $900 million to fund construction of its Hawesville, Kentucky AI data center campus and retire a bridge credit facility. The offering, which was increased from an initial $800 million target, includes an additional 7.11 million share option available to underwriters for 30 days, and is expected to close on April 16. Morgan Stanley led the deal as bookrunner, with Cantor Fitzgerald serving as equity capital markets advisor. Despite the scale of the raise, shares fell in pre-market trading as the new issuance dilutes existing holders by expanding the company's float significantly.

The Hawesville site sits on 750 acres of former industrial land acquired in February 2026 from Century Aluminum for $200 million in cash plus a 6.8% minority equity stake in the holding entity. The former smelter already carries 480 megawatts of existing power capacity with an energized substation. TeraWulf plans to put that 480 MW online by the second half of 2027, with a total projected buildout cost of $3 billion to $4 billion. The $900 million raised this week funds an early phase of that broader capital plan.

At the company's Lake Mariner, New York campus, the CB-2 data center is already fully delivered and generating revenue for anchor tenant Core42, while Phase 1 capacity of 40 MW is targeted for completion in the first half of 2026, with full 200-plus MW capacity at the site expected by year-end.

TeraWulf's pivot away from Bitcoin mining is now reflected in its financial results. In preliminary Q1 2026 figures, the company said HPC (high-performance computing) hosting revenue exceeded 50% of total revenue for the first time, with total Q1 revenue estimated between $30 million and $35 million. CFO Patrick Fleury described the shift plainly: "Our preliminary results reflect a business that has effectively transitioned to long-term, credit-enhanced revenues." The company has locked in approximately $12.8 billion in total contracted HPC revenue across announced deals, including a 10-year, 200-plus megawatt hosting agreement at its Lake Mariner campus with Fluidstack, valued at $3.7 billion (rising to $8.7 billion if extension options are exercised). Google has provided approximately $3.2 billion in credit backing across these deals and now holds approximately 14% pro-forma equity in TeraWulf.

The financial structure carries notable weight. TeraWulf carries $5.8 billion in total debt, split across $2.5 billion in convertible notes, $3.2 billion in senior secured notes, and the $100 million bridge facility being retired with part of these proceeds. This debt load reflects the capital intensity of infrastructure buildouts of the projected scale, with Hawesville alone expected to cost $3 billion to $4 billion, though analyst commentary on debt serviceability was not available at publication time. A new $250 million revolving credit facility, maturing April 2030 and administered by Morgan Stanley, gives the company additional liquidity runway.

TeraWulf is not alone in this shift. Across the publicly listed Bitcoin mining sector, more than $70 billion in AI and HPC contracts have been signed. Comparable deals include CoreWeave's $10.2 billion, 12-year agreement with Core Scientific and a $7 billion, 15-year Google-backed deal with Hut 8. The economic logic is straightforward: miners producing Bitcoin at roughly $79,995 per coin in Q4 2025 while the asset traded between $68,000 and $70,000 were losing approximately $10,000 to $12,000 on every coin mined. AI infrastructure, by contrast, offers margins above 85% with contracted, multi-year revenue, according to industry analysis from CoinDesk.

The global Bitcoin hashrate has already dropped from around 1,160 exahashes per second in October 2025 to approximately 920 EH/s, with three consecutive negative difficulty adjustments recorded. Those difficulty drops reflect reduced competition, which can marginally benefit smaller miners outside the US, including operations such as Bitdeer's 40 MW facility in Ethiopia and HIVE's 300 MW operation in Paraguay.

For readers outside North America, the more consequential story is what this capital concentration means for global access to AI compute. More than 80% of all data centers are located in developed countries and China. Africa hosts less than 1% of global capacity despite representing 18% of the global population. The continent has 223 data centers across 38 countries, with a total market value of $3.49 billion in 2024, projected to reach $6.81 billion by 2030. That growth rate is meaningful in isolation, but it is a small fraction of the hundreds of billions being deployed in the United States and Europe over the same period. The access gap extends into South Asia as well, where India and neighboring markets face comparable structural constraints on AI compute availability. Developers and researchers in cities like Lagos, Nairobi, Karachi, or Mumbai face structurally higher inference costs, limited access to bare-metal GPU infrastructure, and latency penalties that widen as AI systems become more central to commercial activity.

There are functional counterexamples worth watching. Nigeria illustrates the infrastructure challenge starkly: the country operates 17 data centers on a national grid that delivers only around four hours of reliable electricity per day, forcing operators onto diesel generators. Kenya, where the national grid runs more than 60% on renewable energy, has attracted a $1 billion joint investment from Microsoft and G42 for a 100 MW green data center, supported by a 10% corporate tax exemption for qualifying investments. Ethiopia's hydropower, Senegal's wind resources, and similar untapped renewable assets across the continent represent viable foundations for AI infrastructure investment. But private capital requires reliable power, clear data sovereignty regulations, and competitive incentive structures before it moves. The message from this week's TeraWulf raise is that those conditions exist in Kentucky. For policymakers in the Global South, the question is whether they can be built fast enough to matter.