Galaxy Bets Its Data Center Edge on Power, Not ‘Bragawatts’

Galaxy Digital is building a multi-gigawatt data center platform focused on securing reliable power and transmission infrastructure, rather than just headline megawatt figures, according to Austin Storms, co-head of data centers. The company is applying this strategy at its 1.63-GW Helios campus in West Texas and a 500-acre site in McGregor, Texas, as ERCOT tracks over 438 GW of large-load interconnection requests, with data centers accounting for nearly 89% of the total.

Galaxy Bets Its Data Center Edge on Power, Not ‘Bragawatts’ The company’s strategy emphasizes that “bragawatts” aren't all equal: a megawatt at the end of a weak radial line has far less value than one supported by redundant, resilient transmission systems. Galaxy Digital, a digital assets and data center infrastructure company, is betting that a megawatt is only as valuable as the grid behind it. The company is building a multi-gigawatt platform around securing reliable power, validating the transmission systems that deliver it, and building the mechanical and electrical infrastructure needed to turn that capacity into an AI computing space. Austin Storms, Galaxy’s co-head of data centers, told Data Center Knowledge the company wants to own and operate the physical infrastructure while leasing facilities to computing customers. “We want to be the owner and operator of the buildings, the mechanical and electrical infrastructure, as landlord to a variety of tenants,” Storms said. The strategy is unfolding as the Electric Reliability Council of Texas ERCOT tracked more than 438 GW of large-load interconnection requests https://www.ercot.com/news/release/06182026-puct-approves-ercots in June, with data centers accounting for nearly 89% of the total. That month, the Public Utility Commission of Texas PUCT approved ERCOT’s “Batch Zero” process /build-design/texas-creates-a-credibility-test-for-gigawatt-scale-data-center-demand to evaluate eligible large-load requests of 75 MW or more through a systemwide process. Galaxy is applying its approach at its 1.63-GW Helios campus in West Texas and a 500-acre site /data-center-construction/galaxy-s-500-acre-texas-ai-campus-reflects-new-grid-model in McGregor, Texas. Helios is operating with CoreWeave as its tenant. At McGregor, Galaxy plans an initial 74 MW phase, with a multi-hundred-megawatt buildout contingent on additional transmission infrastructure /energy-power-supply/how-ercot-s-post-crez-bet-is-reshaping-ai-infrastructure . Power Quality Comes First Power availability and reliability sit at the top of Galaxy’s site-selection process, Storms said. The company evaluates available capacity, energization schedules, ramp rates, and the transmission infrastructure serving each prospective campus, with a power systems engineer and former ERCOT transmission modeler on staff to assess those systems. Storms has a term for the industry’s abundance of headline-grabbing power figures. “Anybody can secure what we call ‘bragawatts’ across the industry,” he said. “Not all ‘bragawatts’ are created the same.” Neil Osnato, founder of Persistence Analytics Group, said the difference between headline capacity and usable power is critical for large AI campuses. A nominal or approved megawatt figure can describe what may be available under a particular set of assumptions without showing when power can be energized or how reliably it will serve a campus under system stress. For a gigawatt-scale facility, developers need to examine transmission topology, contingency performance, substation configuration, upstream reinforcements, equipment availability, and construction and energization schedules, he said. “A large AI campus does not buy a point of interconnection,” Osnato said. “It buys exposure to an entire upstream electrical system.” As an illustrative example, Storms cited a hypothetical West Texas site at the end of a 138-kilovolt radial system – an example of a power position Galaxy might reject. Osnato said such a site would prompt questions about contingency performance, independent electrical paths, upstream upgrades, restoration exposure, and whether onsite generation is compensating for a weaker grid position. “I would not value 500 MW at the end of a weak radial the same way I would value 500 MW supported by a robust, redundant transmission configuration,” Osnato said. “The number of megawatts may be identical. The infrastructure value is not.” Galaxy puts prospective campuses through a multi-criteria evaluation before they reach its investment committee. Its power, development, and data center teams review the infrastructure and development assumptions. Osnato said Galaxy’s transmission analysis is meaningful, although sophisticated developers increasingly treat that analysis as standard practice. The advantage comes from identifying weak assumptions early and revalidating a power position as utility commitments, required upgrades, equipment availability, neighboring loads, and energization timelines change. “Building data centers is a tough business,” Storms said. “But it’s fairly commoditized at this point.” Galaxy sees differentiation in sourcing highly available, reliable power and delivering physical infrastructure on schedule and on budget. The company can make smaller commitments before all major power approvals are complete, then increase its capital exposure as projects clear development milestones. Even so, Storms said Galaxy wants significant power approval before committing substantial capital to a campus. In West Texas, transmission capacity – not headline megawatts – defines how much AI compute can actually come online. Image: Getty Images Helios Provides the Template Galaxy’s conversion of Helios from bitcoin mining to AI infrastructure helped establish the company’s operating model. The original facility was designed around bitcoin mining, where computing loads can tolerate interruptions that would be unacceptable for an AI data center serving commercial customers. Galaxy upgraded the facility to meet higher availability and reliability requirements. Storms described bitcoin mining facilities as effectively “Tier 0” data centers and said Helios was redesigned for concurrent maintainability and a materially higher availability standard than the mining operation it replaced. At Helios, Galaxy distinguishes site-level approved utility capacity from gross facility power and the smaller critical IT load ultimately available to servers. According to Galaxy, the campus has 1.63 GW of capacity that has completed ERCOT’s LLIS Large Load Interconnection Service process and is under a utility contract. CoreWeave has committed to 800 MW of gross power capacity, corresponding to 526 MW of critical IT load across three development phases. Phase I delivered about 200 MW of gross power capacity and 133 MW of critical IT load to CoreWeave under a 15-year lease. Galaxy says it is in discussions with prospective tenants for the additional 830 MW of approved Helios capacity not yet under lease. Storms described Galaxy’s broader business as multi-gigawatt and multi-tenant, although he prefers a single tenant for each campus. That structure simplifies power allocation and reduces the complications that can arise when multiple customers depend on the same upstream utility infrastructure. “We want a multi-tenant business,” Storms said. “But the best and easiest way to scale this today for our own growth is single-tenant campuses within a multi-tenant business.” Labor Now Shapes Where Galaxy Builds Power availability gets Galaxy’s attention first. Construction labor increasingly determines whether a site can make it through the rest of the process. Storms said skilled electrical and mechanical trades are now the company’s biggest broad constraint as developers pursue multiple gigawatts of data center construction across the country. Supply-chain pressure has eased from conditions 18 to 24 months ago, he said, but skilled labor remains difficult to secure /data-center-construction/ai-data-center-boom-strains-global-construction-capacity . Galaxy built a workforce-development hub next to Helios with capacity for up to 1,600 beds, according to Storms. That experience affects where Galaxy looks next. A hypothetical gigawatt-scale campus in Wyoming could offer attractive power and environmental characteristics, Storms said, while still failing Galaxy’s screening process if the company cannot mobilize enough skilled workers to build it on schedule. Power availability, power reliability, and access to construction labor are major factors for Galaxy when evaluating markets outside Texas. Water and fiber, Storms said, are generally easier problems to solve. Texas Remains a Core Market, but Not the Only One Galaxy continues to favor Texas because of the state’s underlying energy resources. Storms cited natural gas generation, solar development, and utility-scale battery storage as advantages for long-term data center development. Galaxy is also looking outside ERCOT. Storms said the company has a strong origination pipeline beyond Texas, including a prospective site in the continental Midwest /data-center-site-selection/hyperscale-growth-shifts-inland-as-ai-drives-power-demand with roughly 300 MW of utility capacity. The screening framework remains consistent across markets: start with power availability and reliability, then examine whether Galaxy can assemble the labor and infrastructure required to build the campus. The company also evaluates the AI workload a site can support. West Texas can accommodate very large training clusters, Storms said, while sites closer to major economic corridors can offer latency advantages for inference workloads. Galaxy Wants a Common Rulebook for Large Loads Galaxy’s experience across power markets has shaped what Storms wants from grid operators: a standardized framework for connecting large computational loads across regional transmission organizations and independent system operators. ERCOT, PJM, MISO, and SPP have all been developing approaches for large electronic or computational loads, and Galaxy wants greater consistency across markets. “The biggest thing that Galaxy needs and the industry needs more broadly, from utilities and independent system operators or regional transmission operators, is a fully formed framework and rule set for how these types of facilities interconnect to the existing grid,” Storms said. A consistent framework, he added, would allow developers to apply established high-voltage infrastructure and data center designs across multiple markets. ERCOT’s Batch Zero process illustrates the direction of that change by evaluating eligible large-load requests on a systemwide basis and identifying their combined reliability impacts and required transmission upgrades. For Galaxy, the process reinforces the value of testing those assumptions before committing capital. Galaxy Saw the Power Bottleneck Early Galaxy began studying the AI infrastructure opportunity in mid-2024, Storms said, when a small team traveled to Silicon Valley to assess the scale of expected AI demand. The company concluded that AI developers wanted enormous amounts of compute and that many lacked sufficient power pipelines to support their deployments. Galaxy identified a bottleneck spanning generation, transmission, and interconnection and began pursuing ERCOT interconnections before Batch Zero took shape. In July, Galaxy priced $3.507 billion of senior secured notes to finance two Helios buildings with eight data halls, 400 MW of utility capacity, and 260 MW of critical IT capacity. The financing illustrates the capital required to turn an approved power position into an operating data center. The strategy does not eliminate execution risk: transmission upgrades, equipment delivery, construction labor, utility milestones, and tenant timing can still delay the conversion of approved capacity into revenue. Galaxy’s bet is that rigorous power analysis can identify those risks early enough to distinguish a viable AI campus from a headline capacity claim.