Hook: The Earnings Miss Was Not the Entire Story
Galaxy Digital stock fell from $22.14 to $19.07 after the company reported a second-quarter 2025 net loss of $85 million. The decline was approximately 14 percent in one session. That number was immediately interpreted as another verdict on crypto exposure. The interpretation is incomplete.
The same report contained a second data point. Galaxy’s Helios data center generated revenue for the first time. The initial 133 megawatts had been delivered, and management expects the data center business to add approximately $80 million of revenue per quarter from the third quarter onward. The company also secured $3.5 billion in senior secured notes due in 2031 to accelerate construction.
This produces a balance sheet with two opposing clocks. The digital asset business is marked to market every day. The data center business requires years of construction, financing, power delivery, and contract execution before its full economic value can be recognized. The stock market marked the first clock immediately. It has not yet accepted the second.
The ledger does not lie, only the auditors do. In this case, the ledger is not a single ledger. It is a collection of operating statements, asset values, power contracts, debt obligations, and market expectations. Combining them without separating realized performance from projected cash flow creates a misleading picture of Galaxy Digital.
Context: A Financial Company Becoming an Infrastructure Owner
Galaxy Digital is not a token issuer and does not have a conventional token economy. There is no supply cap, staking schedule, burn mechanism, or protocol treasury to audit. GLXY is an equity security in a publicly traded company. Its value is determined by the company’s assets, liabilities, earnings, capital allocation, and market expectations.
The company’s historical identity has been built around digital asset trading, investing, asset management, and market making. That business remains economically sensitive to the price of Bitcoin and Ether, the level of exchange activity, institutional demand, and the amount of liquidity available to crypto markets. In the second quarter, the digital asset division produced adjusted gross profit of approximately $66 million, a 34 percent increase from the previous quarter. This improvement occurred while broader crypto trading volume declined by roughly 7 percent.
That divergence matters. Lower industry volume did not eliminate Galaxy’s ability to monetize institutional flow. However, gross profit is not the same as net income. Trading results, investment revaluations, compensation, financing expenses, depreciation, and construction costs still determine the consolidated result. A profitable operating segment can coexist with a group-level loss.
The second business is Helios, a Texas-based power and data center development. Galaxy has delivered the first 133 megawatts and controls three additional Texas sites. The combined power pipeline is reported to exceed 5.7 gigawatts. The company has entered a long-term arrangement with CoreWeave, an artificial intelligence cloud provider, and expects the data center operation to produce high margins after the relevant facilities are in service.
The structure is not a blockchain protocol. It is a physical infrastructure project. Its technical bottlenecks are land, interconnection rights, transmission capacity, cooling systems, construction execution, customer credit, and reliable power. Software remains important to the eventual tenant, but it is not Galaxy’s principal source of differentiation. Galaxy is attempting to convert infrastructure originally associated with crypto mining into an asset capable of serving artificial intelligence workloads.
My audit experience with early crypto contracts makes the distinction important. In 2017, I reviewed fifteen ICO contracts and found that the strongest marketing claims often depended on assumptions that were not present in the code. A data center has a similar verification problem, although the failure modes are physical rather than contractual. A promised gigawatt is not an operating gigawatt. A signed lease is not the same as collected rent. A power pipeline is not the same as delivered capacity.
The relevant question is therefore not whether the AI narrative is attractive. It is whether Galaxy can convert power access into contracted, operating, and cash-generating capacity without allowing debt service to outrun the construction schedule.
Core Insight: Separate the Two Earnings Engines
Galaxy’s second-quarter loss is a real financial result, but it does not measure the same economic object as the Helios pipeline. The market is currently compressing a volatile trading business and a long-duration infrastructure project into one crypto-equity multiple.
This compression explains much of the stock reaction. Investors saw revenue fall from approximately $10.2 billion in the first quarter to $8.7 billion in the second quarter. They saw a net loss of $85 million. They saw EBITDA of negative $77 million. They saw an increase in leverage through $3.5 billion of senior secured notes. These figures are immediate, comparable, and easy to price.
The Helios figures are different. The first phase delivered 133 megawatts. The data center produced approximately $20 million of gross profit in the second quarter. Management expects approximately $80 million in quarterly revenue beginning in the third quarter. The long-term power pipeline exceeds 5.7 gigawatts. Each figure requires a different level of verification.
The first verification layer is physical delivery. The 133 megawatts is the most concrete number because it refers to installed and delivered capacity. It can be examined through construction milestones, energized equipment, tenant occupancy, and power consumption. It is not enough to establish the value of the entire pipeline, but it provides an initial operating reference.
The second layer is contractual delivery. A fifteen-year arrangement with CoreWeave can support project financing and create a predictable revenue profile. It also transfers part of the market risk from Galaxy to the tenant. The contract does not eliminate risk. It changes the risk. Galaxy becomes exposed to construction performance, power pricing, customer concentration, contract terms, and the tenant’s own financial position.
The third layer is accounting recognition. Revenue recorded in a quarterly report is more useful than management’s description of future capacity. Yet even recognized revenue must be separated into recurring lease income, construction-related income, power pass-throughs, and one-time items. A high-margin data center lease should not be valued in the same manner as a temporary construction payment.
The fourth layer is cash conversion. A data center can report revenue while consuming cash. Construction deposits, equipment purchases, interest expense, maintenance, taxes, and working capital requirements can reduce the amount available to equity holders. The company’s negative EBITDA indicates that consolidated cash generation remains under pressure. That does not invalidate the AI project. It establishes the financing burden that the project must overcome.
This is why the $3.5 billion financing is both an accelerator and a constraint. It provides capital for the second phase of Helios I and signals that Galaxy can access institutional debt markets. It also creates fixed obligations that do not decline merely because crypto trading volume declines. The notes mature in 2031, but the economic burden begins immediately through interest expense, covenants, collateral restrictions, and refinancing expectations.
The market should track interest coverage rather than headline financing size. If EBITDA remains negative while construction spending expands, the company may depend on asset sales, additional equity, or further debt. If data center cash flow arrives on schedule and operating expenses remain controlled, leverage can become an instrument for accelerating asset formation. The same debt can be constructive in one execution path and destabilizing in another.
A simple monitoring equation is useful:
Operating cash available for debt service equals data center rent collected, plus digital asset operating profit, minus corporate expenses, cash taxes, maintenance capital expenditure, and trading-related funding costs.
This is not a formal company guidance metric. It is an analytical filter. If the result remains negative after Helios revenue scales, the AI narrative has not yet become a balance sheet solution.
Liquidity flows are just money with a pulse. In Galaxy’s case, liquidity is moving from a trading-oriented business model into power and compute infrastructure. The movement is not automatically value creation. Capital becomes value only when the resulting asset earns an acceptable return after financing and operating costs.
The 34 percent quarterly increase in adjusted gross profit from the digital asset division should not be ignored. It indicates that Galaxy can still extract revenue from institutional markets during a period of lower aggregate activity. The company may have stronger market-making capabilities than the consolidated loss suggests. However, that improvement is exposed to volume, spreads, volatility, and asset prices. It is not a substitute for recurring infrastructure revenue.
The reverse is also true. The initial data center revenue should not be treated as proof that Galaxy has become an AI company. At 133 megawatts, the operating base is small compared with the stated 5.7 gigawatt pipeline. The difference is a factor of more than forty. The pipeline is an option on future development, not current production.
That gap is the central measurement problem. A 5.7 gigawatt pipeline can support an ambitious valuation model. It cannot support a current earnings model unless capacity is contracted, financed, built, energized, and occupied. Analysts who capitalize the full pipeline today are valuing a series of future projects. Analysts who ignore the pipeline entirely are treating the first phase as the final business. Both approaches are incomplete.
My experience building liquidity dashboards during the 2020 DeFi cycle reinforces this point. I once tracked roughly 5,000 ETH entering newly launched liquidity pairs and found that about 60 percent of apparent volume was associated with wash trading from a small number of wallets. The headline metric was real, but its economic meaning was overstated. Data center capacity requires the same forensic discipline. The number can be accurate while the conclusion remains wrong.
For Helios, the equivalent of wash trading is underutilization. Delivered power does not guarantee productive compute. Compute capacity does not guarantee customer demand. Customer demand does not guarantee rent collection. Rent collection does not guarantee free cash flow. Each transition must be verified separately.
The Power Pipeline Is the Real Technology Story
The most important technical asset in Galaxy’s pivot is not a proprietary model, chip architecture, or software platform. It is the ability to obtain and deliver power at a location that can support high-density computing.
This changes the competitive comparison. CoreWeave is a specialized AI cloud provider. It operates closer to the compute customer and manages the hardware and cloud layer. Galaxy is moving toward the role of infrastructure owner and landlord. Traditional Bitcoin miners such as Riot Platforms and Marathon Digital have relevant experience in site selection, power procurement, facility operations, and high-density electrical systems. Their mining fleets are not interchangeable with AI clusters, but their physical infrastructure experience is transferable.
The 5.7 gigawatt figure therefore has strategic importance only if the company controls the rights required to develop that capacity. The potential moat is not the headline number. It is the combination of land, grid interconnection, permits, transmission access, local relationships, construction capability, and long-term tenant demand.
A power pipeline may also resemble a real-world asset structure. The economic asset is a contracted right to transform electricity and physical facilities into predictable lease income. In that sense, Galaxy is converting a mining-oriented infrastructure footprint into an infrastructure rental asset. This resembles a power-and-property real-world asset model more than a decentralized network.
The risk is that the value of these rights depends on local market conditions. Texas power markets are governed by ERCOT, and extreme weather can affect availability and price. Heat events can increase demand. Winter storms can constrain supply. Transmission congestion can create local price differences. A long-term tenant contract can stabilize revenue while leaving Galaxy exposed to the cost and reliability of the underlying power system.
The company must also manage cooling, substations, backup generation, networking, and equipment density. AI workloads can impose different operating requirements from Bitcoin mining. Training clusters require low-latency networking and stable performance. Inference workloads can have different demand patterns. A facility designed for one workload may require expensive modifications for another.
This is a complex engineering system without the benefit of a cryptographic consensus layer. There is no on-chain finality to verify. There is no public mempool showing utilization. There is no permissionless audit trail for every maintenance event. The verification process depends on construction reports, power records, tenant disclosures, engineering certifications, and audited financial statements.
Tracing the ghost funds from the genesis block is useful as a metaphor for corporate infrastructure analysis. The equivalent investigation begins with financing proceeds and follows them through deposits, construction contracts, equipment purchases, energized capacity, lease commencement, invoicing, and collected cash. A financing announcement is merely the starting block.
The Valuation Problem: Crypto Beta Versus Infrastructure Duration
Galaxy Digital has historically offered investors a high-beta way to obtain exposure to crypto markets through a public equity. It is not a pure Bitcoin treasury vehicle like MicroStrategy. It is not a regulated exchange with the direct transaction-volume profile of Coinbase. Its earnings can be influenced by trading spreads, investment gains and losses, market-making performance, asset prices, and corporate expenses.
This makes GLXY an imperfect proxy for crypto prices. When Bitcoin and Ether rise, the company can benefit through asset appreciation, trading demand, and investor interest. But its own investment and trading losses can offset part of that benefit. When crypto activity declines, the company may still generate gross profit through institutional flow, but the fixed cost base remains.
The AI data center business introduces a different valuation framework. Long-term leases are often valued according to contracted revenue, operating margin, tenant credit, remaining contract duration, capital intensity, and discount rates. A fifteen-year lease can resemble a bond-like cash flow stream, but only after construction and commencement conditions are satisfied. The market may assign a higher multiple to predictable infrastructure income than to volatile trading income.
A successful reclassification would require evidence. The first evidence is recurring quarterly revenue. The second is stable or improving gross margin. The third is cash conversion after maintenance capital expenditure. The fourth is diversification beyond a single tenant. The fifth is a transparent reconciliation between data center earnings and corporate financing obligations.
The company’s $27 billion total equity figure provides a reference point, but price-to-book cannot be calculated accurately without the share count and a clear treatment of digital asset positions, deferred taxes, and infrastructure development assets. A low apparent price-to-book multiple would not automatically make the stock inexpensive. Book value may include volatile assets. Conversely, a market price that appears close to book value may undervalue future contracted infrastructure cash flow.
The earnings decline therefore creates two possible interpretations. It may be a normal mark-to-market loss during a crypto slowdown, with an emerging infrastructure business hidden inside the consolidated accounts. Or it may demonstrate that the new business is arriving too slowly to offset the old business and the additional debt. The next two or three reporting periods will distinguish these outcomes.
Contrarian Angle: The AI Pivot May Increase Risk Before It Reduces It
The conventional reading is straightforward. Crypto earnings weakened, the stock sold off, and Helios provides a new growth engine. The contrarian reading is less comfortable. Galaxy’s move into AI infrastructure may increase financial risk before it produces diversification.
A business is not diversified merely because it has two labels. Diversification requires different revenue drivers and sufficiently independent cash flow. At present, Galaxy’s digital asset business is volatile and its data center business is still in a construction and ramp-up phase. Both require capital. Both face institutional counterparties. Both are exposed to changes in financing conditions. The apparent hedge may be weaker than the narrative suggests.
The debt structure is the most direct reason. The $3.5 billion senior secured notes create a fixed claim against a company whose consolidated EBITDA is negative. If Helios construction is delayed, the debt remains. If the tenant renegotiates, the debt remains. If power costs rise, the debt remains. If crypto prices fall, the trading division may contribute less cash while the infrastructure obligations continue.
Tenant concentration is another blind spot. A long-term CoreWeave contract reduces near-term demand risk, but it also concentrates counterparty risk. CoreWeave is itself exposed to the capital expenditure cycle of AI customers. If a small number of large model developers reduce spending, delay deployments, or shift workloads, the effect can move through the chain from end user to cloud provider to data center owner.
The high-margin forecast also requires qualification. Lease margins can look attractive before corporate overhead, interest, depreciation, taxes, and expansion capital expenditure are assigned. A facility may produce strong site-level gross profit and weak equity-level returns. Investors should ask whether the reported margin is calculated before or after power costs, tenant improvements, maintenance, and required reserves.
The 5.7 gigawatt pipeline may create another form of narrative risk. Large future capacity numbers are useful for strategic positioning and fundraising. They can also encourage the market to capitalize projects that remain subject to permits, interconnection, financing, procurement, and tenant commitments. The distance between 133 megawatts delivered and 5.7 gigawatts planned is not a footnote. It is the execution program.
There is also a competitive threat from direct arrangements between technology companies and energy suppliers. Large cloud providers may sign power purchase agreements, acquire land, or build facilities without an intermediary. Traditional data center operators, specialized AI cloud companies, real estate investment trusts, and former mining companies are competing for the same constrained resources.
Regulation is less likely to be the immediate failure point because Galaxy is a public company subject to disclosure, audit, and securities oversight. The more material uncertainties involve operational execution, customer concentration, capital structure, and policy changes affecting digital asset activity. The chain holds the knife when the oracle bleeds, but this is not primarily an oracle problem. It is a contract, power, and balance sheet problem.
The public-company structure also has limits. Transparent reporting does not guarantee good capital allocation. A board can disclose a project accurately while still overpaying for expansion. A secured note can be fully documented while still imposing an unfavorable return requirement. Governance lowers anonymity risk. It does not remove execution risk.
This distinction matters for investors searching for undervalued projects during a sideways market. A depressed share price can represent a delayed recognition of future cash flow. It can also represent an accurate discount for leverage and uncertainty. The correct signal is not the stock’s 14 percent fall. The correct signal is whether operational evidence improves while the price remains compressed.
What the Next Reports Must Prove
The third-quarter report is the immediate test. Management’s expectation of approximately $80 million in quarterly data center revenue creates a measurable threshold. Revenue close to that level would not prove the entire thesis, but it would confirm that the first phase is progressing from development into operation. A materially lower figure would force investors to examine timing, utilization, contract conditions, and construction delays.
The quality of that revenue matters more than the headline. Investors should distinguish rent received from accounting revenue, recurring lease income from one-time payments, and site-level gross profit from consolidated free cash flow. The report should also identify data center capital expenditure, remaining construction commitments, power costs, and the treatment of tenant-related investments.
Interest coverage must be monitored at the same time. If operating earnings remain insufficient to cover financing costs, the company is still in the capital formation stage. That may be acceptable if contracted revenue is growing and the funding plan is credible. It becomes dangerous if expansion depends on continuous access to favorable debt or equity markets.
CoreWeave’s disclosures are also relevant. Payment performance, contract duration, capacity commitments, customer concentration, and any renegotiation would directly affect Galaxy’s infrastructure valuation. A fifteen-year lease provides duration. It does not provide immunity from counterparty stress.
The crypto segment should be measured independently. Trading volume has declined, but Galaxy’s adjusted gross profit increased by 34 percent quarter over quarter. If volume stabilizes or recovers, the trading business may provide a second source of cash. If volume continues to decline, investors should determine whether the margin improvement came from repeatable market-making activity or from a temporary mix of transactions.
The power pipeline requires a milestone framework. Future capacity should be classified as proposed, permitted, interconnected, under construction, energized, contracted, and operational. Without this classification, 5.7 gigawatts is a promotional aggregate rather than an investable operating metric.
A useful data architecture would separate the company into three ledgers. The first ledger would contain digital asset positions and mark-to-market exposure. The second would contain trading and asset management revenue, expenses, and capital requirements. The third would contain data center assets, contracted capacity, construction spending, debt allocated to the project, and cash collected from tenants.
That separation would make the thesis reproducible. It would also prevent a rising crypto market from masking weak infrastructure execution, or a growing data center segment from masking excessive digital asset risk. Based on my work analyzing Terra’s collapse in 2022, timeline reconstruction is often more valuable than a single headline ratio. The sequence of events reveals whether a system is strengthening or merely postponing recognition of stress.
For Galaxy, the sequence is clear. Financing comes before construction. Construction comes before energization. Energization comes before tenant utilization. Utilization comes before recurring rent. Recurring rent comes before free cash flow. Free cash flow comes before meaningful deleveraging. Each stage has a separate failure mode.
Takeaway: Watch the Conversion Rate, Not the Narrative
Galaxy Digital’s 14 percent stock decline reflects a real loss, weaker consolidated revenue, negative EBITDA, and a market still assigning dominant weight to crypto exposure. It does not settle the value of Helios.
The decisive metric is the conversion rate from power pipeline to collected, recurring, post-financing cash flow. The company has delivered 133 megawatts and secured major financing. The next test is whether quarterly data center revenue approaches the expected $80 million while interest coverage, tenant performance, and construction commitments remain controlled.
If that conversion occurs, GLXY may gradually acquire an infrastructure valuation alongside its crypto beta. If it does not, the $3.5 billion financing will look less like an accelerator and more like a fixed claim attached to an unfinished project.
Fact-checking the hype with cold, hard chain data is useful even when the asset is no longer on-chain. The next quarter will show whether Galaxy is building a second earnings engine or simply adding leverage to the first one. The price will react before the answer is fully obvious. The ledger will take longer.
