Applied Digital’s $2.4 billion steam-power bet for AI campuses | Magica
Applied Digital’s $2.4 billion bet on steam power is a wager on the turbine queue
Editorial Team
••📖6 min read
A Base Electron project backed by Applied Digital has authorization to build 1.2 GW of gas-fired boiler and steam-turbine generation for AI campuses. The equipment is established, not novel; the unproven part is whether the related power developer can permit, finance and deliver it soon enough to make the alternative to gas turbines matter.
Base Electron has authorized a $2.4 billion, 1.2-GW gas-fired boiler-and-steam-turbine project intended to supply Applied Digital AI campuses.
The strategy is to use available boiler and steam-turbine supply chains while combustion-turbine delivery slots are tight—not to introduce a new generating technology.
Full notice to proceed starts engineering and manufacturing; it does not settle the project’s permitting, financing, power-contract or delivery risks.
Applied Digital is backing an effort to turn a gas-turbine bottleneck into a data-center power option. Base Electron, an independent power producer founded by the Applied Digital team, has given Babcock & Wilcox full notice to proceed on a $2.4 billion design-build agreement for 1.2 gigawatts of generation: four 300-MW natural-gas-fired boilers with steam-turbine generators. The capacity is intended for Applied Digital’s AI Factory campuses under separate power-supply agreements, the contractor’s announcement says.
Wes Cummins, Applied Digital’s chairman and chief executive, is advancing the plan while his Dallas-based company builds and operates high-performance-computing data centers. Its stakes are unusually direct: Base Electron says it was founded to develop dedicated generation for Applied Digital, has an exclusive customer-provider relationship with it, and Applied Digital says its shareholders own roughly 10% of the power producer through the company’s investment. That makes the arrangement more than an arms-length equipment purchase, even though Base Electron is legally independent.
The consequence is not yet lower-cost electricity or a completed power plant. It is a bet that a conventional steam configuration can be ordered and assembled before a more familiar gas-turbine project, allowing Applied Digital’s campus development to keep moving.
B&W’s company-reported presentation identifies the Base Electron project as four 300 MW natural-gas boilers with Siemens Energy steam-turbine generator sets. Source: www.sec.gov.
Internet scans and leaked databases expose the spread of ARTEX, a security-testing framework used in cyberattacks. Researchers documented largely autonomous compromises, but the records also show failed cashouts and uncertain counts of independent deployments.
Editorial Team
The hardware is established Rankine-cycle generation: a boiler burns natural gas, makes steam and drives a steam turbine. Babcock & Wilcox, the Akron, Ohio-headquartered public power-equipment company, will engineer, procure and construct the facility; Siemens Energy has been released to design and supply the turbine-generator sets.
The claimed advantage is manufacturing lead time. Cummins told an industry publication that gas-turbine orders placed now might not be delivered until 2032 and that the first boiler-and-steam installation is scheduled to begin operating late in 2028. Those are a customer’s timing expectations, not an independently verified construction schedule.
B&W’s chief technology officer, Brandy Johnson, described the alternative more narrowly in an interview: the company can offer designs it has already built and standardize configurations rather than begin with a fully custom plant. B&W’s investor material says it has more than 5,000 industrial water-tube package boilers and other waste-heat-recovery products installed worldwide, evidence of an established equipment base rather than a new AI-specific invention.
That comparison has limits. The same industry analysis says very large steam turbines are facing rising demand from combined-cycle and nuclear projects, while turbines in the 150- to 300-MW range are also seeing stronger demand. A boiler and steam turbine may offer a different procurement path; the sources do not establish that it will remain a faster one once orders scale.
B&W’s company-reported diagram shows its proposed steam-first configuration and the later gas-turbine addition. Source: www.sec.gov.
The eventual target is still a gas-turbine plant
Applied Digital does not present the steam plant as the endpoint. Cummins said the company intends to add gas turbines when they are available, allowing the existing boilers to operate as fired heat-recovery steam generators. B&W calls that second stage an “enhanced combined cycle” configuration in its investor overview.
That distinction matters. It means the first phase is designed to generate power while a gas turbine is still being manufactured; it is not evidence that steam-only equipment displaces gas turbines on efficiency or economics. B&W illustrates the later configuration with a 50-MW gas turbine plus a 50-MW boiler producing about 100 MW, versus about 75 MW for a traditional 50-MW combustion-turbine-plus-HRSG configuration. That is a company illustration of a particular configuration, not a cost or performance estimate for the 1.2-GW North Dakota project.
The $2.4 billion headline also needs a careful denominator. Dividing the announced design-build value by the stated 1,200 MW yields about $2 million per MW of planned capacity. It is a project-value ratio, not the plant’s eventual electricity price, delivered cost or financing need. The sources do not disclose Base Electron’s financing terms or the separate power-supply agreements.
Applied Digital has more exposure than the structure suggests
Applied Digital’s March 4 8-K filing says the company provided a guarantee. Its maximum potential exposure is effectively limited by Base Electron’s obligations under the design-build agreement—an anticipated total contract price of about $2.4 billion—subject to Applied Digital’s right to terminate the guarantee in specified circumstances.
The financing context is substantial. In fiscal 2026, Applied Digital reported $4.2 billion in cash, cash equivalents and restricted cash and $5.0 billion in debt as of May 31. It also reported a $249.2 million net loss from continuing operations attributable to common stockholders for the year. Those figures describe the company, not a project financing package, but they show why the terms and timing of new capital remain material.
Applied Digital’s own results release also says it had executed long-term leases for 1.41 GW of contracted critical IT load across five campuses as of May 31, with $36 billion of total contracted revenue over the initial 15-year base terms. At Polaris Forge 1, its first 100-MW data center became operational in October 2025 and a further 75 MW was ready for service on June 30. The record demonstrates operating progress at one campus; it does not establish that the proposed generating plant will be completed on the same timetable. The company’s release separately lists construction, customer approvals, additional capital, power disruptions, equipment failures and financing on acceptable terms among its risks.
The next proof is outside the equipment order
B&W says engineering and manufacturing have begun, and its investor presentation says the plant would be near Center, North Dakota, where a conditional-use permit was submitted in August 2026. That is progress, but a submitted permit is not an approval.
The next evidence to watch is therefore concrete: a permit decision, disclosed project financing and the terms and counterparties of the power-supply agreements. Those milestones would show whether Base Electron can convert a fully authorized equipment contract into usable capacity. Until then, the deal is best understood as a serious attempt to buy time around a turbine shortage—not proof that Applied Digital has removed the power, capital or supply-chain constraints facing its AI-campus buildout. A contemporaneous project account likewise describes the work as power to be delivered under separate agreements, underscoring that the equipment contract and the commercial delivery of power are distinct steps.
Meta's strategy chief and former CFO is joining Fireworks AI ahead of a reported fundraising push targeting at least a $30 billion valuation. The inference company announced a $17.5 billion valuation in July, making the hire part of a much larger expansion than its older funding figures suggest.
Sand.ai says its $200 million Series A+ round lifts its valuation above $1 billion. Co-lead investor Wangsu plans inference and deployment cooperation, while its filing shows substantial losses behind the startup’s video AI ambitions.
An advocacy group estimates Interior Secretary Doug Burgum’s AI holdings at $18.1 million–$64.4 million as his department advances data centers on public land. His ethics agreement allows retained investments with recusals, while pending permits will determine both the application of those safeguards and the public’s return. Interior denies his decisions are driven by personal investments.
Microsoft is evaluating fewer neocloud partners while expanding its own infrastructure. Delivery-dependent contracts underpin supplier financing, while IREN reports rising revenue alongside a $6 billion funding target for its broader expansion.
Physical Superintelligence says its AI system found a proof of a conjecture limiting quantum query advantages on random inputs. Public proof files enable scrutiny, but do not establish the private system’s cost or repeatability.
OpenAI is expanding its relationship with Hacktron, whose researchers previously reported an attack chain into employee accounts, to test third-party connections with greater access to cybersecurity models.
Google has acquired Crusoe’s southeast Wyoming data center site, where it plans a long-term $17 billion investment. Crusoe describes completed pre-development work; independent reporting points to customer setbacks. Power delivery and carbon capture remain key tests.
The team behind document-parsing service SoMark has released OXR-1.0 code and weights. A benchmark retest supports a strong result, but speed and accuracy were measured separately, and users above funding or revenue thresholds need a commercial license.
OpenAI is reportedly developing a Business Hub with daily mention and click charts, product updates and business agents. The unannounced feature could give verified businesses direct visibility into ChatGPT responses, without revealing prompts or competitor rankings.
Palantir and Nvidia have specified an integrated stack for running AI in customer-controlled environments. The architecture builds on existing software and infrastructure offerings, with published configurations spanning 32 to 640 GPUs, while pricing and workload economics remain unproven in the launch materials.
Democrats’ incoming congressional majorities put hearings and investigations within reach, but AI rules and crypto legislation still face bipartisan bargaining and a Republican president.
Microsoft reportedly declined to renew Oracle server rentals as older chips became less useful to it. The decision comes despite continuing capacity shortages, while Oracle says demand for earlier-generation GPUs remains strong.
An investigation puts Nvidia’s future commitments and guarantees at $530.5 billion. Its disclosures show rapidly expanding supplier commitments and phased support for customers whose infrastructure could generate further chip sales, with cash payments depending on timing, demand and contract terms.
TSMC is reportedly planning a possible move into 3D DRAM at Nvidia’s request. The foundry already helps integrate AI processors with memory, but a manufacturing expansion has no disclosed factory or production date.
Physicl has raised a $10 million seed round and introduced Studio, a beta app for generating targeted physical AI data. Its reported customer gains concern 3D reconstruction, leaving broader robot reliability and commercial economics to be demonstrated.
New cloud partners in Southeast Asia must enroll in Nvidia’s Fleet Intelligence system, according to a report, linking preferred chip access to monitoring while leaving data sharing and existing partners’ obligations unresolved.
Airbnb sold intellectual property to a private AI lab controlled by CEO Brian Chesky in a $7.1 million arrangement that also reimburses expenses. It obtained a license to certain lab technology and future investment rights, but disclosed no asset valuation or detailed license terms.
Chip shares recovered part of Thursday’s losses in early Friday trading. Fears about data-centre construction face a test against AI-chip supplier Nvidia’s strong reported sales, growth forecast and conditional infrastructure-financing plans.
SoftBank shares fell 10.8% as default-insurance costs reached a four-year high. Delayed listings and unfinished data centers complicate its AI investment outlook, while a $40 billion bridge facility has a March 2027 maturity and a repayment plan based on existing assets and other financing.
