Hyperscale Data has positioned itself as a domestic AI infrastructure operator, and the single largest claim on the stock is a decade long colocation contract that has not yet produced a dollar of revenue. The contract anchors a 617,000 square foot campus in Michigan, and management values the deal in excess of 1.2 billion over the maximum term. The stock trades near its 52 week low, and the fully diluted share count has more than doubled in a single year. The company reports a net loss for the first half of 2026. The capital structure is the central risk in the story, and the dilution is the most consistent headwind on the stock.
The most important recent development is the June 2026 master services agreement with an unnamed California neocloud provider. The mechanism is a shift of the Michigan campus from Bitcoin mining to high margin AI colocation, and the customer paid 5 million in non refundable upfront fees against a retrofit that costs 100 million or more. But the contract value is locked to the customer exercising its extension and expansion options, not to any delivered revenue, and the company cannot yet fund the retrofit from its own cash flow.
The central risk is the capital structure. Shares outstanding grew from 323 million at the start of 2026. By the end of June, the count reached 581 million. The company still carries 95 million in current notes payable. The 2027 guidance calls for revenue of 300 million at the low end. The high end is 350 million. It depends almost entirely on the Ault Capital Group portfolio that management expects to divest.
The catalyst to watch is Phase 1 of the Michigan retrofit reaching ready for service status in late September. That milestone is the first point at which colocation revenue begins to land, and it is the first test of whether the anchor customer stays. The Series F exchange offer that carries the ACG divestiture defines the second half of 2026.
The peer set for Hyperscale Data is not a single category. The data center business sits in the independent colocation and AI infrastructure space, where the closest analogues are the neocloud and hyperscale colocation operators that serve AI training and inference workloads. The Bitcoin mining operations align the company with the digital asset treasury names that own mining fleets and hold a strategic reserve of coins. The Ault Capital Group portfolio, which includes crane rental, defense electronics, hotels, and private credit, has no close public peer set at all, and it is the reason the consolidated revenue line looks like a diversified holding company rather than an AI infrastructure operator.
The strategic center of gravity has shifted from a diversified holding company to an AI infrastructure operator, and the shift is explicit in the company name change from Ault Alliance to Hyperscale Data. The June 2026 master services agreement is the load bearing event that makes the shift credible. It pairs a 10 year initial term with two 5 year extension options. It also gives the customer a right of first offer on an additional 32 megawatts of capacity if the initial deployment is met. The total contract value over the maximum term is in excess of 1.2 billion. It rises above 3 billion if the customer exercises the additional 32 megawatts. The customer is identified only as a California based neocloud provider, and the company has not disclosed the identity, which is a deliberate choice that protects the competitive position of the deal.
The corporate structure is the second strategic feature. The company is a Delaware corporation headquartered in Las Vegas, Nevada, and it operates through two main subsidiaries. Sentinum, Inc. owns and operates the Michigan data center campus and the Bitcoin mining operations. Ault Capital Group, Inc. is a hybrid private equity firm that acquires, finances, and holds controlling interests in businesses across financial services, digital assets, defense, industrial services, hospitality, and real estate. The company expects the divestiture of ACG to occur in 2027, and upon completion the company becomes a pure play data center operator and digital asset holder. The divestiture is carried out through a voluntary exchange offer in which holders of the Series F exchangeable preferred stock surrender their shares in return for ACG common and preferred stock, and the offer is structured so that only those who agree to surrender receive ACG shares.
The competitive position in each of the three business lines is distinct. In AI colocation, the company competes against the hyperscale operators and the neocloud providers that are building out their own data centers, and its advantage is the existing power capacity and the anchor customer contract that locks in 20 megawatts of demand. In Bitcoin mining, the company competes on energy cost and has approximately 28 megawatts of mining capacity at the Michigan campus, a figure that declines as the AI retrofit absorbs power. In the ACG portfolio, the company does not compete directly in any of the operating businesses, and the portfolio companies are operated by their own management teams, with ACG acting as a holding company that allocates capital and oversees governance. The principal structural risk is that the consolidated financials blend all three of these lines, and the market cannot easily isolate the value of the data center business from the value of the portfolio companies and the mining operations.
The Michigan data center campus is the flagship asset and the source of the company's only durable moat. It spans 617,000 square feet across 83 acres in Dowagiac, Michigan. The campus has approximately 30 megawatts of power energized to date. The expansion program targets long term potential in excess of 300 megawatts, subject to utility agreements, regulatory approvals, and financing. The moat is the combination of energized power capacity, the existing physical plant, and the anchor customer contract that locks in 20 megawatts of demand for a decade. A new entrant cannot replicate the power interconnection and the retrofit economics, and the 300 megawatt end state represents a 10x expansion of the current footprint.
The retrofit for the anchor customer is the most significant capital commitment in the company's history. It covers approximately 60,000 square feet of the campus. The estimated cost is 100 million at the low end. The high end is 120 million for the initial 20 megawatts. The retrofit converts space currently used for Bitcoin mining into AI colocation, and the power reallocation is progressive as each phase is commissioned. Phase one delivers 10 megawatts with a ready for service date targeted at late September. Phase two delivers the remaining 10 megawatts by the end of the year. The cost is not funded from operating cash flow, and the company has been drawing on debt and equity to cover the buildout.
The Bitcoin mining operations are the legacy asset that is being converted. The company mined 212 Bitcoin during fiscal 2025. That generated approximately 22.6 million in revenue with approximately 1.83 exahashes per second of capacity. The Bitcoin treasury holdings totaled 959 Bitcoin. The holdings were valued at approximately 60.8 million as of early August 2026. The mining business is a cash generator, but it is also the source of the power that the AI retrofit absorbs, and the two lines compete for the same megawatts. As the colocation phases come online, the mining capacity at the Michigan campus declines, and the company has stated that it continues to operate mining at its Montana facility during the transition.
The Ault Capital Group portfolio is the second product line, and it has no technology moat that the market can price with confidence. The portfolio includes crane rental through Circle 8 Crane Services, defense electronics through Gresham Worldwide, hotel operations through Ault Global Real Estate Equities, and commercial power electronics through TurnOnGreen. Each of these businesses is operated by its own management team, and the portfolio companies carry non recourse debt that is not guaranteed by the parent. The principal value of the portfolio to a GPUS shareholder is the revenue and EBITDA that it contributes to the consolidated results, and the principal risk is that the portfolio is being divested in 2027, which means the consolidated financials that the market is currently pricing look materially different after the exchange offer completes.
The first half of 2026 produced total revenue of 78.9 million. That was up from 50.9 million a year earlier. The revenue mix is dominated by the ACG portfolio. Defense solutions revenue jumped to 23.8 million from 3.3 million a year earlier. The jump was driven by the reconsolidation of Gresham Worldwide after it emerged from bankruptcy in late last year. Crane operations revenue was 22.1 million, down from 25.4 million. Crypto asset mining revenue was 10.0 million, roughly flat. The data center and AI colocation business contributed no revenue in the first half, because the retrofit was not yet complete. The consolidated revenue line is therefore a proxy for the portfolio, not for the AI infrastructure story that the stock is trading on.
The net loss for the first half was 49.1 million, compared to 22.1 million a year earlier. The loss widened for three distinct and clearly identifiable reasons. General and administrative expenses nearly doubled to 38.4 million from 19.1 million, reflecting the cost of the public company infrastructure and the acquisition integration. Research and development spending jumped to 9.1 million from 241,000, driven by the robotics and AI initiatives. A 16.1 million gain on the extinguishment of a settlement obligation provided a one time offset that flattered the result. The operating cash flow for the period was negative 9.9 million, and the company burned cash through operating activities even as it generated revenue.
The balance sheet is the most concerning line item. Cash and cash equivalents stood at 36.8 million at the end of June. Restricted cash added another 28.3 million, bringing the combined total to a substantial figure. Current notes payable were 95.2 million, and the guarantee liability added nearly 40 million. Total stockholders equity fell over the year. The accumulated deficit grew to 790.5 million. The equity base is thin relative to the debt, and the current notes are the most immediate liquidity constraint. The company raised 50.2 million in new common stock in the first half. It also drew 60.1 million in new notes payable. The net effect was to increase the share count by 258 million shares while the cash balance grew only modestly.
The cash flow statement tells the same story in more detail. Operating cash flow was negative 9.9 million. Investing cash flow was negative 45.1 million. The largest drivers were 17.3 million of property and equipment purchases and 12.7 million of non marketable equity investments. Financing cash flow was positive 70.8 million, driven by the equity raise and the note issuance. The net increase in cash was 15.9 million for the period. The company is funding its capital expenditure and investment program entirely through external financing, and the operating business is not yet generating enough cash to cover its own costs. The 16.1 million gain on the extinguishment of the settlement obligation is a one time item that does not recur, and the underlying cash burn is worse than the headline net loss suggests.
The guidance for the coming year is the single most important forward document in the company's history, and it is structured as a three platform forecast. The data center, AI infrastructure, and robotics platform is guided to 40 million at the low end and 50 million at the high end in revenue. The lending, financial services, and digital assets platform is guided to 100 million at the low end and 150 million at the high end. The portfolio companies platform is guided to 150 million at the low end and 200 million at the high end. Consolidated adjusted EBITDA is guided to 60 million at the low end and 80 million at the high end. The data center platform, which is the asset that the market is pricing, contributes less than a fifth of the guided revenue.
The execution risk is concentrated in the Michigan retrofit. Phase one is targeted for a ready for service date in late September, and Phase two is targeted for the end of the year. The retrofit cost is 100 million at the low end. The high end is 120 million, and the company funds it from external capital while the operating business is still cash negative. The anchor customer has paid 5 million in upfront fees. It has also provided a 5.6 million security deposit, which is reduced by one third on each of the first, second, and third anniversaries of the Phase two target delivery date. If the company misses the Phase one deadline, the customer has termination rights, and the contract value that management is using to value the stock is at risk.
The ACG divestiture is the second major execution risk, and it is scheduled for 2027. The divestiture is carried out through a voluntary exchange offer in which holders of the Series F exchangeable preferred stock surrender their shares in return for ACG common and preferred stock. Only those who agree to surrender receive ACG shares, and the offer is structured to leave the remaining shareholders with a pure play data center company and a digital asset treasury. The risk is that the exchange ratio and the timing of the offer are not yet fixed, and the market is asked to price a company whose capital structure is about to change. The 10 percent special dividend of Class B stock announced in early September is a signal that the company is preparing for the divestiture, and the Class B shares carry 10x the voting power of Class A.
The guidance is explicitly preliminary, and management has listed a set of assumptions that it expects to hold. These include the timely deployment of capacity at the Michigan campus, the availability of sufficient debt and equity financing, the continued growth of the lending and digital asset operations, and the absence of material adverse changes in economic or regulatory conditions. The guidance is not a commitment, and management has stated that it updates the guidance periodically. The principal risk is that the guidance is too optimistic for the portfolio companies, which are being divested, and that the data center revenue that the market is pricing does not arrive until the following year at the earliest, after the share count has expanded further to fund the buildout.
The first risk is the capital structure. The company has 95 million in current notes payable. It has 36.8 million of unrestricted cash, and the gap is bridged by the 28.3 million in restricted cash that is not available for general corporate purposes. The company has been funding its operations and its capital expenditure through continuous equity issuance, and the share count has grown by 258 million shares in the first half of 2026 alone. The 10 percent special dividend of Class B stock announced in September adds another 20 million shares to the mix, and the Class B shares carry 10x the voting power of Class A. The dilution is the most consistent headwind on the stock, and it is not a risk that the company can mitigate except by generating operating cash flow.
The second risk is the anchor customer. The master services agreement is the single largest asset in the company's balance sheet, and it is locked to one unnamed customer. The customer has paid 5 million in upfront fees. It has also provided a 5.6 million security deposit. The contract value of 1.2 billion over the maximum term depends on the customer exercising its two 5 year extension options. If the customer terminates at the end of the initial 10 year term, or if it declines to exercise the extensions, the revenue that the market is pricing evaporates. The customer is a California based neocloud provider, and the company has not disclosed its identity, which means the market cannot independently assess the credit quality of the counterparty.
The third risk is the ACG divestiture. The portfolio companies are the source of the majority of the consolidated revenue, and they are being divested in 2027. After the divestiture, the company is a data center operator with a Bitcoin treasury and a robotics business, and the revenue base is materially smaller. The guidance for the coming year of 300 million at the low end assumes the portfolio companies are still consolidated. The high end is 350 million. If the divestiture occurs earlier than expected, or if the portfolio companies underperform, the guidance is revised down. The robotics business, which includes the Omnipresent Robotics subsidiary and the AGIBOT partnership, has not yet generated meaningful revenue, and it is excluded from the current year earnings guidance.
The fourth risk is the Bitcoin treasury. The company holds 959 Bitcoin, valued at approximately 60.8 million. The crypto assets on the balance sheet fell from 46.2 million at the start of the year to 4.4 million by the end of June. The company recorded 7.1 million in impairment of restricted crypto assets in the first half. The Bitcoin price is the swing factor for the treasury, and a decline in the price of Bitcoin directly reduces the equity base. The company has pledged a portion of the treasury as collateral for borrowings, and a sharp decline in the price of Bitcoin could trigger a margin call or a forced sale.
The valuation framework for Hyperscale Data requires a sum of the parts, because the consolidated financials blend three businesses with very different economics. The data center business is an AI infrastructure operator with a 10 year anchor contract and a retrofit under way. The ACG portfolio is a diversified holding company with crane rental, defense electronics, hotels, and private credit. The Bitcoin treasury is a digital asset position with 959 coins. The robotics business is a pre revenue venture. The market cap is approximately 27 million. The company has 581 million Class A shares outstanding, which means the equity is priced at less than 5 cents per share on a fully diluted basis.
The bear case assumes the retrofit slips, the anchor customer does not exercise the extensions, and the ACG divestiture removes the majority of the consolidated revenue. In this scenario, the data center business generates 40 million in revenue in 2027 with a margin profile that is not yet visible, and the equity base is diluted further to fund the buildout. The bear case value for the equity is the Bitcoin treasury at 60 million, the Michigan campus at a depreciated cost basis, and the ACG portfolio at a net asset value that is difficult to estimate, less the 95 million in current debt. The bear case equity value is in the range of 20 million to 40 million, which is below the current market cap, and the bear case implies the stock is fairly valued at the current price.
The base case assumes the retrofit is completed on schedule, the anchor customer exercises the first extension, and the ACG divestiture occurs in 2027 as planned. In this scenario, the data center business generates 50 million at the low end in revenue in the coming year. The high end is 80 million. The ACG portfolio is exchanged and the remaining shareholders hold a pure play data center company. The base case equity value is the present value of the data center cash flows, the Bitcoin treasury, and the net asset value of the remaining assets, less the debt. The base case implies an equity value of 80 million at the low end. The high end is 120 million, which is several times the current market cap.
The bull case assumes the retrofit is completed ahead of schedule, the anchor customer exercises both the extensions and the additional 32 megawatts, and the Bitcoin price rises. In this scenario, the data center business generates 100 million at the low end in revenue in the coming year. The high end is 150 million. The total contract value of 3 billion is locked in, and the Bitcoin treasury is worth more than 100 million. The bull case implies an equity value of 200 million at the low end. The high end is 400 million, which is many times the current market cap. The bull case requires the anchor customer to exercise all of its options, the Bitcoin price to rise, and the company to fund the expansion without further dilution, and the probability of the full bull case is low.
Hyperscale Data is a binary story, and the binary is the Michigan retrofit. The company has 959 Bitcoin, a data center campus of 617,000 square feet, and a decade long anchor contract. The contract is worth 1.2 billion over the maximum term, and the company also has a portfolio of operating companies that is being divested. The stock is trading near its 52 week low, and the share count has more than doubled in a year. The question is whether the retrofit is completed on schedule and the anchor customer stays, and the answer to that question determines whether the stock is a deeply undervalued asset or a dilution machine that is pricing in a future that never arrives.
The counterargument to the bear case is that the 1.2 billion contract is real, and that the upfront fee and the security deposit are cash that has already been received. The counterargument to the bull case is that the contract value depends on the customer exercising options that are not yet exercised, and that the company is funding the retrofit with dilutive equity at a price that is near the 52 week low. The net effect of these two counterarguments is that the stock is priced at the intersection of the execution risk and the dilution risk, and the current price reflects a market that is skeptical of both.
The judgment is that the stock is a speculative position with a defined catalyst and a defined risk. The catalyst is the Phase 1 ready for service date in late September, and the risk is the dilution and the divestiture. An investor who believes the retrofit is on schedule and the anchor customer is a quality counterparty has a case for the stock being undervalued. An investor who does not believe either of those things has a case for the stock being overvalued. The 2027 guidance is a portfolio forecast, and the data center revenue that the market is pricing is not yet visible in the financials. The stock is a bet on the retrofit, not on the portfolio, and the portfolio is being sold.