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

Galaxy Digital has put 133 megawatts of critical IT load into service, giving its latest google news moment a measurable foundation. The capacity sits at Helios, a former Bitcoin mining campus in West Texas that Galaxy is converting for artificial intelligence workloads. Its new pitch is simple: available power matters more than the biggest projected campus number.

That distinction challenges a familiar data center marketing habit. Developers routinely announce multigigawatt pipelines, even when much of that capacity lacks completed grid studies, firm utility arrangements, financing, tenants, or delivery dates. Industry shorthand sometimes calls these speculative claims “bragawatts.”

Galaxy still publishes large future numbers of its own. Helios has 1.63 gigawatts of approved capacity, and the company says the campus has a potential path to 3.6 gigawatts. However, Galaxy now has something more persuasive than a development map. It has delivered the first 133 megawatts of usable IT capacity to CoreWeave under a long-term lease.

The milestone does not settle the company’s data center case. Most of its contracted capacity remains under construction, and the expansion carries substantial financing and execution risk. Galaxy also depends heavily on CoreWeave, whose own growth rests on concentrated customers and sustained demand for AI computing.

The real contest is therefore not Galaxy against one conventional data center operator. It is deliverable power against speculative scale. Helios gives Galaxy credible evidence for the first side of that contest, but the next two phases will determine whether that evidence becomes a repeatable business.

Galaxy Has Moved Helios From a Promise to an Operating Asset

The important change is not another capacity announcement. Galaxy has started delivering contracted computing infrastructure and collecting rent.

On July 6, 2026, Galaxy said it had completed Phase I at Helios on schedule. The company delivered about 200 megawatts of gross power, meaning total electricity supplied to the facility, for 133 megawatts of critical IT load. Critical IT load measures the electricity available directly to computing equipment rather than cooling, electrical conversion, and other supporting systems.

CoreWeave receives that capacity through a 15-year lease. Galaxy said rent commencement began during the second quarter of 2026. That detail turns Helios from a construction story into an operating revenue asset, even though later phases remain unfinished.

The gap between gross power and critical IT load also matters. A headline campus figure does not equal the electricity available to servers. Phase I’s 200 megawatts of gross supply supports 133 megawatts of IT equipment, illustrating why comparisons based only on top-line power can mislead readers.

Galaxy’s Phase I delivery covers a converted portion of a campus originally built for Bitcoin mining. The company shut down its proprietary mining operations at Helios while preparing the site for AI and high-performance computing, or HPC, which uses clustered systems to process demanding workloads.

That conversion is central to Galaxy’s advantage. The company did not begin with an empty parcel and an aspirational utility request. Helios already had land, transmission access, power infrastructure, and experience supporting electricity-intensive computing.

A former mining site is not automatically ready for modern AI clusters. AI facilities require different cooling systems, network capacity, backup infrastructure, power quality, and operational practices. Galaxy still had to retrofit the campus and meet CoreWeave’s technical requirements.

Completing that work gives the company an execution reference it lacked when the first lease was announced in March 2025. At that point, Galaxy had a tenant and a valuable power position, but it had not demonstrated that it could convert a mining campus into hyperscale AI infrastructure.

Phase I now provides that proof point, within the limits of the company’s own reporting. Galaxy says the work finished on budget and on schedule. Investors and prospective tenants can compare that claim with a completed delivery rather than relying entirely on a target date.

CoreWeave’s role gives the milestone additional weight. The company operates a cloud platform built around graphics processing units, or GPUs, and leases large blocks of capacity for AI training and inference. It needs facilities that can support dense clusters and sustained electrical demand.

The first phase also creates pressure for Galaxy itself. Once a developer delivers one stage successfully, customers expect the same performance at a larger scale. Phase II is almost twice the critical IT load of Phase I and uses newly constructed buildings rather than only a retrofit.

Google news coverage can amplify the 133-megawatt headline, but the operating distinction deserves more attention. Galaxy has crossed the boundary between controlling a promising power site and running a leased AI data center.

Why Google News Is Focusing on Power Instead of Bragawatts

AI infrastructure has shifted from a race for announced capacity to a race for capacity that can clear grid, construction, and financing constraints.

A developer can describe a future campus in gigawatts long before the underlying power becomes usable. The number might represent a utility application, land suitable for development, an early study request, or a long-term expansion scenario. Each stage carries a different probability of completion.

Galaxy’s argument is that buyers should separate those stages. At Helios, the company says 1.63 gigawatts has completed the required ERCOT interconnection process and is covered by utility arrangements. ERCOT is the grid operator responsible for most of Texas, coordinating electricity supply and transmission across its market.

In January 2026, Galaxy announced that Helios had completed a large-load interconnection study for another 830 megawatts. It also said it had signed a service agreement with AEP Texas for that capacity. The ERCOT approval doubled the campus’s approved total to more than 1.6 gigawatts.

Approval does not mean every data hall already exists. It means the project has advanced further through the power process than a site supported only by a preliminary request. Construction, equipment procurement, local permitting, financing, and customer contracting remain separate hurdles.

Those distinctions are becoming more important as grid operators confront unprecedented large-load requests. Data centers can seek hundreds of megawatts at one location, while the transmission system may need new substations, lines, generation, or reliability controls to support them.

The problem encourages inflated comparisons. One developer might cite an operating capacity figure, while another cites a utility queue, and a third cites a site’s eventual physical potential. All three numbers can appear beside “GW,” despite representing different commercial realities.

Galaxy is not innocent of ambitious presentation. Its public materials highlight Helios’s potential to reach 3.6 gigawatts, well above today’s operating load. That figure should not be treated as delivered capacity or a guaranteed expansion.

However, the company can now place four separate categories around its numbers. It has 133 megawatts of operating critical IT load, 526 megawatts contracted across three phases, 1.63 gigawatts of approved gross capacity, and a larger potential campus buildout. Keeping those categories separate makes the Helios story more credible.

The phrase “bragawatts” captures the temptation to collapse those stages into one promotional figure. It also explains why Galaxy’s Phase I delivery matters more than another announcement about theoretical scale.

For enterprise buyers, the difference affects deployment schedules. An AI company cannot train models on power expected after an uncertain transmission upgrade. It needs a date when energized racks, cooling, networking, and backup systems will be available.

For investors, the difference affects valuation. Operating capacity can generate revenue, while contracted capacity supports forecasts with construction risk. Approved but unleased capacity represents an option, and theoretical capacity remains an earlier development possibility.

Google news headlines tend to compress these categories into a single number. The better reading is that Galaxy has improved the quality of its capacity claim, not merely increased its size.

CoreWeave Turns Power Availability Into Galaxy’s Competitive Test

Galaxy’s primary opponent is the industry’s pipeline-first development model, but CoreWeave determines whether its delivery-first alternative works.

Across Phases I through III, CoreWeave has committed to 526 megawatts of critical IT load at Helios. Supporting that computing capacity requires 800 megawatts of gross power. The leases have initial 15-year terms and include extension options.

Phase I accounts for 133 megawatts of critical IT load. Phase II adds 260 megawatts, while Phase III adds another 133 megawatts. Galaxy expects Phase II data hall deliveries to begin during the first half of 2027, with Phase III following later.

This structure gives Galaxy a contracted path beyond the first delivery. It also reduces near-term leasing uncertainty because CoreWeave has already committed to all 800 megawatts of gross capacity assigned to the three phases.

The arrangement does not eliminate risk. It concentrates Galaxy’s initial data center expansion around one tenant. A long lease can strengthen revenue visibility, but the value of that lease still depends on the tenant’s financial health, customer demand, and ability to fill the deployed infrastructure.

CoreWeave has expanded rapidly by acquiring GPUs, leasing data center capacity, and supplying computing resources to AI companies. That model requires large continuing capital commitments. It also leaves the company exposed if customer demand, hardware economics, or financing conditions change.

Galaxy is therefore building around a tenant that faces many of the same industry pressures. Both companies benefit when AI developers seek more computing capacity than established cloud providers can immediately supply. Both face pressure if utilization weakens or customers delay deployments.

The relationship also shows why power has become a competitive weapon. CoreWeave could obtain servers from multiple suppliers, but those servers create no value without energized buildings. A campus with secured power and a firm delivery plan can become more strategically important than the hardware inside it.

Helios competes indirectly with large projects backed by hyperscalers, specialist cloud providers, real estate companies, utilities, and private infrastructure funds. Many rivals have deeper operating histories in data centers than Galaxy. Some have broader tenant bases, lower financing costs, or established construction supply chains.

Galaxy’s counterargument rests on its starting position. The company acquired an energy-intensive computing site, controlled significant land, and inherited grid infrastructure from the mining operation. It then redirected those assets toward a market willing to pay for dense AI capacity.

That history resembles CoreWeave’s own transition. CoreWeave began with cryptocurrency mining before turning its GPU fleet toward cloud computing. The two companies reflect a wider shift in which assets created for crypto workloads find new value in AI infrastructure.

The parallel should not be overstated. Converting a GPU business differs from converting a power campus. Yet both cases show how crypto-era investments can become inputs for the AI buildout when operators possess useful equipment, power access, or technical expertise.

Galaxy’s CoreWeave expansion began with an additional 260-megawatt commitment in 2025. Later agreements brought contracted critical IT load to 526 megawatts.

The contract gives Galaxy a clearer demand signal than an unleased campus pipeline. Still, it does not prove that every later phase will arrive on time or produce the company’s projected economics.

The company has won the first round against the bragawatt model because it delivered usable capacity. The harder test is repeating that result while construction scale, debt exposure, and customer concentration all increase.

Helios Works Because Galaxy Controlled the Scarce Input First

Galaxy’s mechanism is not a secret data center design. It is control of land and grid access before AI demand made those assets harder to secure.

Helios sits in Dickens County in the Texas Panhandle. The site provides ample land for expansion and access to electricity infrastructure originally intended for large, continuous mining loads. Galaxy says the campus now spans more than 2,200 acres.

The company acquired Helios from Argo Blockchain in late 2022. At the time, the transaction centered on Bitcoin mining rather than AI cloud infrastructure. That timing gave Galaxy control of the site before the current surge in large AI data center proposals.

Power has since become one of the industry’s main limiting inputs. Chips receive more attention, but a GPU order alone does not produce computing capacity. Operators need utility supply, substations, backup systems, cooling, water or alternative heat-management systems, and high-bandwidth network connections.

These requirements interact. Higher rack densities place more demand on cooling and electrical distribution. Galaxy says Helios is designed to support rack densities between 130 and 140 kilowatts, meaning each equipment rack can consume far more power than a conventional enterprise server rack.

The campus also needs reliable connectivity. Galaxy reports round-trip network latency below 15 milliseconds between Helios and Dallas. Latency measures the time data needs to travel across a network and return, and lower figures help distributed systems exchange information faster.

Those specifications are company claims, not independent performance audits. They nevertheless show why the conversion required more than redirecting electricity from mining machines to AI servers.

The Phase I ratio also reveals the infrastructure overhead. About 200 megawatts of gross power supports 133 megawatts of computing load. The remaining electricity serves cooling, power conversion, lighting, pumps, controls, and other facility functions.

A developer advertising only gross power can therefore create an incomplete impression of available computing capacity. Buyers need to know the critical IT load, expected rack density, cooling design, redundancy level, and delivery schedule.

Galaxy says Helios includes uninterruptible power supply systems and battery backup. An uninterruptible power supply bridges brief disturbances and supports orderly transitions when primary electricity fails. It does not replace every form of long-duration backup generation.

The company’s strategy also depends on Texas’s distinctive market. ERCOT offers a large, relatively isolated electricity system with substantial wind and solar generation, flexible market pricing, and significant industrial development. It also faces reliability and transmission challenges created by rapid load growth.

Large data centers can provide predictable utility demand, but they can also strain local infrastructure. Their consumption may require new transmission investment and can intensify debates over who pays for grid upgrades.

Galaxy’s completed large-load study is therefore valuable because it places Helios beyond an early request. Yet ERCOT approval does not immunize the campus from congestion, changing grid rules, weather events, or future reliability requirements.

The company’s mechanism remains more defensible than a simple scale claim. It controlled a suitable site, advanced the interconnection process, contracted an anchor tenant, financed construction, and delivered the first operating phase.

That sequence is difficult to reproduce quickly. A competitor cannot create transmission capacity by announcing a larger campus. It must work through studies, equipment lead times, utility coordination, and local construction.

The conversion also gives Galaxy a potential portfolio lesson. Mining infrastructure can serve as a development bridge, but only when the site meets AI customers’ stricter requirements. Many mining locations lack enough fiber, cooling options, redundancy, or proximity to supporting infrastructure.

Helios appears to have cleared the first practical version of that test. The next question is whether its initial advantage survives a much larger greenfield build.

The Numbers Still Carry Construction and Financing Risk

Phase I validates delivery, but it does not validate Galaxy’s entire multigigawatt narrative.

The first uncertainty concerns scale. Galaxy must add 260 megawatts of critical IT load in Phase II, nearly twice the operating capacity delivered in Phase I. The work includes two buildings and eight data halls on roughly 260 acres.

Galaxy priced senior secured notes in July 2026 to finance part of that construction. The notes carry a 9.875 percent annual interest rate and mature in 2031, according to the company’s financing disclosure.

That rate illustrates the cost of turning approved power into usable capacity. Secured electricity creates development value, but the buildings, cooling equipment, electrical systems, and network infrastructure still require large capital commitments.

Higher financing costs reduce the room for delays or budget overruns. Galaxy says Phase I finished on budget, which supports management’s execution case. Phase II remains a separate project with different construction demands.

The second uncertainty is concentration. CoreWeave anchors all three contracted Helios phases. Galaxy expects the leases to produce substantial revenue when fully utilized, but those forecasts depend on one customer meeting long-term obligations.

CoreWeave can spread its capacity among end customers, which provides some demand diversification beneath the lease. However, Galaxy’s direct exposure remains tied to CoreWeave as the contractual tenant.

The third uncertainty involves the difference between contracted and operating load. Only 133 megawatts of the 526-megawatt commitment has reached the completed Phase I milestone. The remaining capacity still faces scheduled delivery work through 2027 and beyond.

The fourth uncertainty concerns the unused portion of Helios’s approved supply. CoreWeave’s three phases use 800 megawatts of gross power, leaving about 830 megawatts of the current approved total for other potential tenants.

That unused approved capacity is valuable, but it is not yet equivalent to a signed lease. Galaxy must secure additional customers and design later development around their requirements. A multitenant campus also introduces more operational complexity than a single anchor deployment.

The fifth uncertainty is the 3.6-gigawatt potential figure. Galaxy presents it as a possible long-term campus scale, not current operating capacity. Readers should keep it separate from the 1.63 gigawatts that the company describes as approved and utility-contracted.

This is where the anti-bragawatt message faces its own credibility test. Galaxy benefits from asking the market to discount competitors’ early pipeline claims. It should apply the same discipline when describing its potential expansion.

The company’s public filings acknowledge risks involving construction timing, financing obligations, counterparties, changing AI infrastructure needs, power availability, regulation, and technical incidents. Those are standard disclosures, but each connects directly to the Helios plan.

Power-market policy adds another layer. Texas authorities are reassessing how very large loads connect to the grid and support reliability. New requirements might affect interconnection timing, emergency curtailment, backup capability, or infrastructure contributions.

Environmental and community questions also remain relevant. Large AI campuses consume electricity continuously and can require significant water or other cooling resources. Local benefits such as construction activity and tax revenue must be weighed against infrastructure demands.

None of these concerns cancels the Phase I achievement. They define what the achievement does and does not establish. Galaxy has shown it can deliver one substantial phase. It has not yet shown that Helios can reach full contracted scale without material delays or economic pressure.

The most credible interpretation sits between celebration and dismissal. Helios is no longer a paper project, but much of its future value remains tied to work that has not been completed.

What the Next Google News Cycle Should Measure

The next Helios headlines should track delivery, tenant diversification, and operating economics rather than the largest available gigawatt figure.

The first signal is Phase II construction. Galaxy says deliveries should begin during the first half of 2027. Progress on civil work, structural completion, equipment installation, commissioning, and the first energized data hall will show whether Phase I’s schedule discipline can scale.

An on-time first Phase II delivery would strengthen Galaxy’s claim that it has become a repeatable hyperscale operator. A material delay would weaken that claim and increase attention on financing costs.

The second signal is a lease for the remaining approved Helios capacity. About 830 megawatts sits outside the 800 megawatts assigned to CoreWeave’s three phases. Signing another credible tenant would reduce concentration and confirm demand beyond the anchor relationship.

The identity of that customer would matter. A hyperscaler or well-capitalized AI platform could improve confidence in the campus. A smaller customer with uncertain funding would add less protection against concentration risk.

The contract structure matters as much as the tenant’s name. Readers should look for the committed critical IT load, lease duration, delivery schedule, extension rights, construction responsibility, and any conditions attached to the agreement.

The third signal is operating performance from Phase I. Galaxy says the entire 133 megawatts of IT load is generating rent. Future financial reports should clarify the contribution from data center operations, associated expenses, uptime, and progress toward management’s margin expectations.

Operating disclosure would help investors distinguish contracted projections from realized economics. It would also show whether Helios produces predictable infrastructure revenue alongside Galaxy’s more volatile digital asset businesses.

These three signals create a cleaner scorecard than campus potential alone. Phase II tests construction. A second tenant tests commercial breadth. Phase I financial performance tests whether delivered power converts into durable returns.

Other developments will still matter. CoreWeave’s customer demand, GPU deployment pace, debt position, and utilization can affect the tenant’s ability to absorb capacity. ERCOT rule changes can alter the operating environment for every large Texas data center.

Competitor behavior will provide another reference. Developers with approved power, financed buildings, and signed tenants deserve closer comparison than projects supported mainly by early pipeline figures. The market needs consistent vocabulary for operating, contracted, approved, and potential capacity.

That vocabulary would improve google news reporting as well. A headline describing a multigigawatt campus can imply more certainty than the project has earned. Identifying the stage of each number gives readers a better picture of risk.

Galaxy has earned attention because it moved one important number into the operating category. The 133-megawatt delivery is smaller than the campus’s multigigawatt potential, but it carries more immediate commercial meaning.

The company now needs to preserve that distinction. If management emphasizes delivered capacity, transparent milestones, and signed demand, Helios can stand apart from speculative development pipelines. If it leans too heavily on eventual scale, its anti-bragawatt argument will lose force.

For developers, enterprise buyers, and AI infrastructure customers, the practical question is no longer who can announce the largest site. Ask which operator has completed grid studies, secured utility service, financed construction, signed dependable tenants, and delivered energized computing load.

Follow those signals through Galaxy’s next construction update and financial report. That is where the Helios thesis will either gain strength or meet its limits, regardless of which number dominates the next google news headline.