Galaxy Digital is a digital assets brokerage and asset manager that has grafted a West Texas AI data center franchise onto its balance sheet. The equity is now a wager on which of the two engines actually produces the return.
The most consequential recent event is a 9.875 percent senior secured note offering by its Helios data center subsidiary, completed in late July 2026. The raise funds two buildings of the Helios campus that lease to CoreWeave under a long term term, and it converts a construction cost into a debt contract that isolates the build from the parent trading losses.
The central tension is that the profitable, revenue heavy Digital Assets segment is a spread business exposed to a falling bitcoin price. Meanwhile the Data Centers segment that carries the long term growth story produced only a small amount of leasing income in the first half of 2026.
The timing trigger is the delivery of the Phase II and Phase III Helios load to CoreWeave in 2027. That is the point at which the data center thesis stops being a promise in a lease document and starts showing up in segment revenue.
Galaxy Digital runs two operating segments, Digital Assets and Data Centers, and a third Treasury and Corporate sleeve that holds its own portfolio of digital assets and venture and private equity investments. The Digital Assets segment is the economic center of gravity today, generating the overwhelming share of revenue through Global Markets and Asset Management and Infrastructure Solutions, while Data Centers is a young franchise that recorded its first leasing revenue only in the second quarter of 2026. The strategic logic of the pairing is deliberate. The brokerage and asset management business earns cash flow that is tightly correlated to the price and volume of digital assets, and that flow was negative in the first half of 2026 because bitcoin and ether fell sharply. The data center business exists to add a contracted, inflation resistant revenue stream that has no direct exposure to the crypto cycle, and that is why management has been willing to fund it with expensive subsidiary debt rather than dilute the parent. The consequence of that choice is a capital structure that treats the two businesses as separate entities for debt, even though the market prices them as a single equity.
The company describes itself as a global financial services and infrastructure firm that facilitates access to digital assets for institutional clients through Global Markets and Asset Management and Infrastructure Solutions, and that develops high performance computing facilities to meet demand for power and compute in the AI build out. It reports relationships with roughly 1,700 counterparties spanning crypto native and traditional finance, and about 7.6 billion in assets across the platform at the end of the second quarter. That counterparty base is the durable asset of the brokerage. It is a distribution network that does not exist on the balance sheet, and it is the reason the spread business can be profitable in a normal market even though the current market is punishing it. The firm is also positioning itself around the broader regulatory shift, having noted the executive order that established a strategic bitcoin reserve and a digital asset stockpile, a development that it frames as a long term positive for institutional adoption even as the near term price effect remains uncertain. The strategic implication of the pairing is that the company is trying to build a business where one half of the income statement is cyclical and the other is contracted, and the market has to decide how much to pay for the contracted half before it produces a meaningful share of the total.
The Data Centers segment anchors on the Helios campus in the West Texas panhandle, where the Electric Reliability Council of Texas has approved over 1.6 gigawatts of gross power capacity. The first building holds 133 megawatts of critical IT load and serves CoreWeave under a 15 year lease. An additional 393 megawatts of IT load is leased to CoreWeave under the Phase II and Phase III agreements, expected to be delivered beginning in the second quarter of 2027, and the remaining approved power stays available for future tenants. The strategic point is that Galaxy has contracted most of the campus to a single named hyperscaler before the buildings are finished, which is the rare position of owning a real estate and power asset with an anchor tenant locked in by contract. The mechanism behind the position is that grid interconnection in Texas is allocated by queue, and a developer that has already cleared that queue for a multi gigawatt site holds an asset that a late entrant can only reach by waiting behind the existing approvals.
The argument that runs through this report is that GLXY is not one company but two, and that the market is forced to assign one price to both. The brokerage is a mature, cyclical spread and asset management business whose earnings are being suppressed by a falling bitcoin price. The data center is a pre scale growth asset with a contracted anchor tenant and a dedicated 3.5 billion dollar debt structure that is secured at the subsidiary level, not the parent. Whether the stock is a value or a trap depends on which of the two stories a holder is actually buying, and the rest of the analysis separates them. The two businesses also sit on different points of the credit cycle. The brokerage earns more when the market is active and prices are stable, while the data center earns from a contracted schedule that is indifferent to the state of the crypto market, and the combination is intended to soften the cyclicality that has defined the brokerage alone.
The Digital Assets segment earns its fee and spread income from Global Markets, which runs over the counter spot and derivatives trading, lending, and structured products, and from Asset Management and Infrastructure Solutions, which manages exchange traded fund and alternatives strategies and provides staking, tokenization, and custodial technology. The moat here is counterparty access and trust. The firm reports roughly 1,700 counterparties spanning crypto native and traditional finance, and that network is the product. It is not something a new entrant can replicate in a quarter, because institutional counterparties take years to qualify and route volume to a broker they believe holds up through a volatile tape. The asset management layer adds a second, stickier fee base, because external allocators and individuals invest through Galaxy managed ETF and alternatives strategies rather than transacting one off. The consequence is a fee stream that is less sensitive to the daily price of any single token than the headline revenue suggests, because fees are earned on flow and on assets under management rather than on the level of the market.
The revenue mix shows how dominant the brokerage is. Sales of bitcoin, ether, and tether made up roughly 87 percent of Digital Assets sales in the second quarter and 84 percent over the first half, and that mix is why the segment swings with the underlying asset price. The firm runs a net long position predominantly in bitcoin, and bitcoin fell 14 percent in the second quarter, with ether down 25 percent over the same window. The mechanism is simple. When the house is long the asset and the asset is falling, the brokerage posts an unrealized loss that flows straight through the segment, and that is the source of much of the first half loss. The spread business is the moat, but the treasury and proprietary trading layer is the volatility. The distinction matters for a holder, because the fee and spread income is the part of the business that survives a bear market, while the mark to market loss on the long position is the part that can overwhelm it in a quarter of falling prices.
The Data Centers segment is a different product entirely. It is a power and real estate business that builds megawatt scale facilities and leases the critical IT load to a named hyperscaler. The technology is not proprietary in any secret sense, it is the ability to secure, build, and operate a 1.6 gigawatt ERCOT approved campus and deliver contracted megawatts on schedule. The moat is power access and contracting, not code. In a market where power availability is the binding constraint on AI compute, a developer that has already locked 1.6 gigawatts of grid capacity and most of it to CoreWeave under a 15 year lease has a position that is difficult for a competitor to take, because a new entrant has to clear the same grid interconnection queue before it can build a single megawatt. The economic structure of the product is also different from the brokerage. Leasing revenue is contracted over the term of the lease and is not marked to a market price each quarter, so the data center income should behave more like a real estate asset than like a trading book, even though it currently shows up as a small line inside a brokerage income statement.
The second quarter of 2026 anchored the first half print. Total revenue was 8.56 billion for the second quarter, essentially flat year over year. The first half revenue of 18.6 billion was down 14 percent from the prior year, and the headline is a loss, not a profit. The net loss was 85.3 million for the second quarter and 301.6 million for the first half. Adjusted EBITDA was negative 77.3 million for the second quarter and negative 264.8 million for the first half, against a positive print a year ago. The headline loss is not the spread business failing. It is the net long digital asset position taking a mark, and the segment detail shows exactly where. The revenue figure also overstates the cash earning power of the business, because the bulk of it is gross sales proceeds from digital assets that are largely offset by a corresponding cost of sales, leaving a thin margin for the spread and fee business underneath. The practical consequence is that a reader who anchors on the headline revenue and the headline loss can misjudge the business, because the revenue is a gross number and the loss is driven by the mark on the position, and the real operating spread is the number in between that matters for the long term.
The segment split makes the structure clear. Digital Assets generated 8.54 billion of the 8.56 billion second quarter revenue, and the other two segments together added a small remainder. Over the first half the brokerage again accounted for the vast majority of the revenue. The data center income is real but still negligible against the brokerage, and it is dominated by a small amount of leasing revenue plus a pass through revenue that came online only in the second quarter. The consequence for a shareholder is that the earnings power the market is paying up for in the data center story has not yet shown up in the segment income statement, and it cannot until the Phase II and Phase III load delivers. The practical implication is that a buyer of the current stock is underwriting a future segment on the strength of a present one, and the bridge between the two is a construction schedule that has not yet produced a single contracted megawatt of scale.
The balance sheet carries the weight of both stories. Total assets were 10.84 billion at the end of the second quarter, down from 11.35 billion a year earlier. Total liabilities stood at 8.12 billion, and total equity at 2.72 billion. Cash and stablecoin holdings were about 2.5 billion, which is the buffer that keeps the brokerage solvent through the drawdown. The equity of 2.72 billion against a market capitalization near 14 billion means the market is assigning a large premium over book, and that premium is almost entirely the data center and bitcoin option, not the current earnings. The segment split shows the brokerage carrying the most assets and liabilities, with the data center carrying a smaller share of the balance sheet. The leverage is also worth noting, because the capital structure is heavy on debt relative to book, and that is the condition that makes the equity more sensitive to both a bitcoin recovery and a data center delivery.
The forward case rests on a single variable: the delivery and monetization of the Helios campus. The Phase II and Phase III agreements commit CoreWeave to 393 megawatts of critical IT load with delivery expected to begin in the second quarter of 2027, and that is the quarter in which the Data Centers segment stops being a rounding error. Until then the segment revenue is a fraction of the brokerage and the growth story is a lease document, not a cash flow. The execution risk is the ordinary risk of a large construction project, and it is real. The firm has to fund two buildings, secure the power, meet the delivery date, and keep the anchor tenant in good standing, all while the parent posts a quarterly loss. The sequencing matters, because the lease revenue cannot flow until the buildings are complete and energized, and any slip in the power or construction timeline delays the first meaningful data center earnings by the same amount. There is also a funding risk embedded in the schedule, because the notes were sized to cover the build and the reserves, and any cost overrun would have to be financed on top of the existing debt or covered by the parent guarantee.
The funding of that build is the July 28, 2026 event that changes the risk profile. Galaxy Helios Data Centers II LLC completed a private offering of 3.507 billion of senior secured notes due 2031, with Morgan Stanley as representative. The notes carry a 9.875 percent coupon. The proceeds finance the two building project and debt service reserves. The notes are secured by first priority liens on substantially all assets of the data center issuer and its guarantor, and the parent provided an uncapped completion guarantee. The mechanism matters. Because the debt is ring fenced at the subsidiary level and secured by the data center assets, the build is financed without diluting the parent and without making the parent brokerage balance sheet bear the construction interest. A 9.875 percent coupon is expensive, but it is cheaper than equity for a company whose shares are priced at a multiple of book, and it keeps the parent free to defend the trading franchise through the crypto drawdown. The trade off is that the high coupon is a permanent drag on the data center cash flow for the life of the notes, and the value of the contracted lease revenue has to clear that interest burden before it creates any net benefit for the parent.
The governance changes are smaller but they point at the same place. On July 13, 2026 the board appointed Steven Bandrowczak, the former Xerox chief executive officer, as a director effective that month, with a seat on the audit committee. The consequence is that the audit committee, the body that signs off on the accounting for a company running a net long digital asset position and a large construction program, now has a public company operator with a technology background in place. On August 5, 2026 the chief accounting officer, Robert Rico, announced his resignation effective at the end of the third quarter, to pursue an outside opportunity, with no disagreement reported. He continues in the role until a permanent replacement is named. The combination is a normal transition, but it lands in the middle of a period of complex accounting, and the interim period is a window in which the quality of the numbers matters more than usual. The timing is also a signal about how the company plans to staff the next stage, because an experienced audit committee member arrives just as the construction and the debt take on more of the reporting burden.
The dominant downside is that the equity is paying for a data center that has not yet earned a meaningful fraction of the revenue it promises, while simultaneously carrying a net long position in an asset that is falling. The firm transacts significantly and holds net long positions predominantly in bitcoin and ether. Bitcoin fell 14 percent in the second quarter, and ether fell 25 percent over the same window, with the first half declines even steeper. The mechanism of the loss is the mark on that position, and a shareholder who enters at a multiple of book is exposed to a second leg of the drawdown. If bitcoin stays depressed, the brokerage earns thin spread revenue against a shrinking asset base, the treasury takes further unrealized losses, and the premium over book that the data center story has built up compresses on both sides at once. The asymmetry of that downside is that a further fall in bitcoin hurts the treasury mark and the brokerage margin at the same time, while the data center has not yet added enough contracted income to offset it.
The second risk is concentration and counterparty. The data center growth story is a lease to a single named tenant, CoreWeave, and most of the contracted campus points at that one relationship. A tenant that is itself a high growth, high leverage hyperscaler is a strong anchor today, but the credit and delivery risk is concentrated, and the remaining 830 megawatts of approved power has no contracted tenant. The parent has also taken on a completion guarantee for the build, which means the ring fencing of the 9.875 percent notes is a credit ring fence, not an absolute one, and a construction overrun or a delayed delivery could pull the parent into the subsidiary's cost base. The mechanism of the concentration is that a single tenant's ability to pay and to take delivery sets the ceiling on the data center cash flow, so the segment cannot be valued as if the load were spread across many independent counterparties the way the brokerage revenue is.
The strongest counterargument to the bear case is that the market is already pricing a weak crypto outcome and that the data center is a genuinely different asset with contracted cash flows that should be valued separately. The bull reading is that 3.507 billion of subsidiary debt proves a real credit market is funding the Helios build. A multi year lease to a named hyperscaler, with over 1.5 gigawatts of ERCOT approved power behind it, is a defensible position. It is also the case that the 2.5 billion of cash and stablecoins keeps the brokerage alive long enough for the data center to mature. The tension between these two readings is the whole stock, and it is why the valuation below values the two segments separately before combining them. A further element of the bull case is that the brokerage itself is optionality, because if the crypto cycle turns, the spread and fee business that has been depressed by the drawdown reopens, and the same balance sheet that is currently a drag becomes a source of cash flow that can fund further expansion without additional external capital.
The framework is a sum of the parts, because the two segments answer to two different economics. The Digital Assets segment is a spread and asset management business that earns fee income, so it is valued on an earnings multiple applied to a normalized run rate, not on the current loss. The Data Centers segment is a contracted power and real estate business, so it is valued on the present value of the leased megawatts. The Treasury and Corporate sleeve is a portfolio of digital assets and investments, so it is valued roughly at its mark. The market capitalization near 14 billion sits against a book of 2.72 billion of equity. That multiple of book near 5 is the anchor the whole valuation has to explain. The reason a sum of the parts is the right tool here is that the two segments have different reference points, and valuing the combined entity on a single multiple either over pays for the brokerage or under pays for the data center.
The bear case values the brokerage at a depressed multiple for a spread business in a falling market and gives the data center little credit. It applies a 4x multiple to a normalized Digital Assets run rate of about 300 million of annualized fee and spread income, worth 1.2 billion. It values the data center at a conservative 1.5 billion for a pre scale asset with one tenant, and it marks the treasury at 1.5 billion against net asset value. The sum, roughly 4.2 billion, implies the shares are well above the sum of the parts in a bear scenario, and the gap is the risk the market is underpricing. The logic of the bear case is that a spread business in a falling market should earn a low multiple, and that a data center with one tenant and no delivered megawatts of scale cannot carry the difference between book value and the market price on its own.
The base case values the brokerage at a 6x multiple on a normalized run rate of 400 million, worth 2.4 billion. It gives the data center a 3.5 billion valuation and marks the treasury at 1.8 billion. The base case is the scenario where nothing goes wrong but nothing exceptional happens either. The sum, roughly 7.7 billion, is below the current market capitalization near 14 billion, which means the base case says the shares are expensive even before adjusting for the bitcoin position the treasury carries.
The bull case assumes the data center matures on schedule into a scaled, contracted power business and the brokerage recovers with the crypto cycle. It values the brokerage at a 7x multiple on a 500 million normalized run rate, worth 3.5 billion. It values the data center at 7 billion for a multi gigawatt campus with contracted and available megawatts, and it marks the treasury at 2.5 billion as bitcoin recovers. The sum, roughly 13 billion, is close to the current price, which means the bull case is what the market is already paying for, and it requires the 2027 delivery to happen on time and bitcoin to stop falling. The logic of the bull case is that once the contracted load is energized, the data center segment has a revenue stream that is locked for the lease term and is not subject to the quarter to quarter swings that have defined the brokerage, and that gives the equity a floor that the current loss does not show.
The verdict is that the current price is paying for the bull case in full while the income statement is still printing the bear case. The base case sum of the parts, roughly 7.7 billion, sits well below the current market capitalization near 14 billion. The only scenario that justifies the price is the bull case, which requires a 2027 delivery on schedule and a bitcoin that stops falling. That is a lot to require of a company whose latest quarter lost 85.3 million and whose first half lost 301.6 million, and whose growth segment contributed a fraction of the revenue. The gap between the base case and the market price is the entire debate, and it is a gap that can only close through a delivered megawatt or a recovered bitcoin, not through a statement in a filing.
The single variable that decides the outcome is the monetization of the Helios campus. If the Phase II and Phase III load delivers on schedule and the Data Centers segment moves from a rounding error to a real contracted cash flow, the 7.7 billion base case becomes the floor and the data center re rates toward the 7 billion bull value. If the delivery slips, or if CoreWeave's own leverage becomes a credit concern, or if the 9.875 percent subsidiary debt proves expensive against the actual yield, the base case becomes the ceiling and the premium over book has no support. The parent's completion guarantee means a bad construction outcome does not stay ring fenced, and that is the tail risk that a clean sum of the parts understates. The watch item to follow is therefore the segment reporting, because the first quarter in which the data center leasing revenue moves from a rounding error to a line item is the point at which the thesis either validates or stalls, and every quarter before that is a quarter in which the market is paying for a promise.
The brokerage is a real asset with a real moat in counterparty access, and the 2.5 billion of cash and stablecoins is a real buffer, but neither of those is what the price is buying at a multiple of book near 5. The shareholder is buying the data center with a brokerage attached, and the brokerage is currently a cost center marked against a falling bitcoin. The honest read is that the stock is a levered bet on a single delivery date and a single asset price, and that the market has already paid up for the win. The governance transition, with a former Xerox chief at the head of the audit committee and a chief accounting officer departing, is a reminder that the next couple of quarters of numbers have to be clean, because the multiple is priced for perfection on a story that has not yet produced the cash flow.
The deeper judgment is about what the two segments are allowed to contribute to the price. The brokerage cannot support a multiple of book near 5 on its current loss, and it has to be the data center that does that work. But the data center cannot do that work until the contracted megawatts are delivered and the lease revenue actually clears the heavy interest on the subsidiary debt. That leaves the equity in an uncomfortable middle, where the price reflects the full bull case while the income statement reflects the bear case, and the resolution of that gap comes only when the next set of contracted megawatts is energized and the first meaningful data center earnings appears in the segment. The framing that ties the whole argument together is that the market has decided in advance that the data center is the real business, and the brokerage is the funding vehicle that has to survive the crypto drawdown long enough for that decision to pay off.