Fermi Inc. is a nineteen-month-old, pre-revenue AI power and data center developer that went public in the fall of last year, has since spent more than a billion and a half in site spend on a Texas Panhandle campus called Project Matador, and now carries a going-concern warning on its own balance sheet. The equity is a binary bet on whether the company can convert its first anchor lease with TensorWave into financed construction, and whether the governance turbulence of the first eight months is a cost already paid or the beginning of a longer conflict. The stock has repriced at least once already, and the market has moved from a narrative premium to a distress discount in under a year.
The near-term evidence is mixed in a specific way. The TensorWave lease signed in early August is a real, binding, multi-billion-dollar contract that converts Fermi from a land-and-permit story into a construction-and-financing story, but the same quarterly report discloses a going-concern doubt, a large first-half loss, a convertible note issued after quarter-end, and four separate lawsuits including one from the former CEO. The contract is the single most important fact, the capital stack is the second, and everything else is noise with option value attached to it.
The market cap sits in the low single-digit billions against a physical asset base that is itself in the low single-digit billions, and against a contracted revenue stream that is an order of magnitude larger. The price is not paying for revenue, or for EBITDA, or even for the first turbine; it is paying for the probability that a fifteen-year lease becomes a funded, delivered, and occupied facility. That probability is the entire argument that follows.
Fermi Inc. is a Texas corporation formed as Fermi LLC in January of last year and converted to corporate form in the fall of last year, immediately after its registration statement went effective and before the shares began trading. The company raised roughly three-quarters of a billion in net proceeds in its initial public offering, including a full exercise of the underwriters' over-allotment option, and it is dual-listed in New York and London. It had no revenue as of the most recent quarter-end, and its only operating asset is a single campus in the Texas Panhandle.
The entire business case sits on Project Matador, a multi-phase development on a site of more than five thousand acres held or to be held under a long-term ground lease with the Texas Tech University System in Carson County, Texas, with adjacent acreage under contract or option bringing the expanded campus to roughly eight and a half thousand acres. The design end-state is more than eleven gigawatts of predominantly private power generation, expandable toward seventeen gigawatts with the optioned land and additional state air permits, supporting up to fifteen million square feet of AI-ready hyperscale compute over a multi-decade timeline.
The strategy is to develop and lease private-grid powered data center space, not to sell power into the wholesale market or operate a utility. Fermi integrates on-site natural gas-fired generation, supplemental grid-supplied power from a local utility, battery energy storage, solar, and a longer-term nuclear baseload program, all behind the meter, and leases the resulting capacity to hyperscale and neocloud tenants under modified-net leases. The company states that it does not expect to generate operating revenues until it commences delivery under definitive tenant lease agreements, and the single signed agreement as of this writing is the TensorWave lease.
The reporting context matters for reading the numbers. The most recent quarterly report covers the second quarter of this year, and the first full fiscal year report covers the prior calendar year. The company also disclosed a deferral of its previously announced REIT election: it intended to elect REIT status for the short taxable year ending at the close of the prior calendar year, but deferred the election, is taxable as a C corporation for the current and next tax years, and may never make the election. The deferral signals that the company is not yet ready to commit to the REIT framework while it is still pre-revenue, and the question of whether an election ever happens is one of the named thesis variables.
The product is not a software platform or a single building; it is a bundled, site-specific energy and compute infrastructure package. The tangible advantage, to the extent one exists, is a combination of long-term site control in a region with cheap gas and available water, secured long-lead-time turbine equipment, a first-mover position in a specific AI infrastructure market, and the contracted revenue of the TensorWave lease. None of these is a durable moat in the classic sense, and the honest read is that the moat is a position, not a franchise. The TensorWave lease is the anchor commercial asset. It is a fifteen-year modified-net build-to-suit lease for a facility of more than two hundred megawatts on roughly two and a half hundred acres of the campus, with about three hundred thousand square feet across four data halls and one network data hall, and an expected contracted revenue of roughly six and a half billion over the initial term. The lease includes two five-year extension options and an expansion option for two additional buildings that would triple the facility power to about six and a half hundred megawatts, with the expanded buildings required to be energized by the end of 2028. The tenant is a subsidiary of TensorWave Inc., which provides an unconditional parent guaranty, and Fermi is negotiating an additional guaranty of the tenant's rent obligations with a global AI infrastructure firm. Fermi provides a completion guaranty for its own construction, and target delivery spans from late 2027 into the first quarter of 2028, with rent credits for delay and a tenant termination right if a phase misses its outside date. The power strategy is layered. Phase One is six smaller gas turbines capable of nearly three hundred megawatts, with site excavation complete and a balance-of-plant contractor engaged. Phase Two is three larger turbines rated at more than seven hundred megawatts in simple-cycle mode, and the large units arrived at the Port of Houston in July, bringing the company's landed power assets to about one and a half gigawatts. The company's own generation program is targeted at roughly four and a half gigawatts when combined with the Hillcore alliance described next. A front-end engineering design study for the nuclear component, completed in July by a major EPC contractor, supports commissioning one reactor unit in each of the mid-2030s, with the first unit now expected about a year later than previously disclosed, and the nuclear regulator has formally acknowledged the company's draft environmental impact statement under a pilot review program.
The Hillcore framework agreement, signed in early August, is structurally important because it shifts the capital burden. Under a build-own-operate-transfer arrangement on roughly four hundred acres, Hillcore Energy Capital intends to finance, construct, own, and operate a gas-fired, solar, and battery storage facility of up to about two and a half gigawatts, with Fermi as anchor offtaker under long-term take-or-pay power purchase agreements. Fermi commits no capital and issues no debt for the Hillcore plant, and its guarantee of payment obligations is capped at a limited number of months of fixed capacity charges. The framework is subject to definitive documentation within ninety days and a forty-five-day diligence period, so it is not yet a binding commitment, and the company's generation buildout would roughly double if it closes.
The company had zero revenue for the first half of this year and for the comparable stub period of last year, and its total first-half loss was more than two hundred million, or about thirty-four cents per share, with the second-quarter loss in the mid-twenties of millions, or about four cents per share. The loss is dominated by non-cash items, and the more useful cash metric is the first-half operating cash burn, which was in the mid-fifties of millions, alongside investing outflows for property, plant, and equipment in the low six-hundreds of millions.
The loss composition is worth dissecting because it is not operating burn in the traditional sense. The first-half general and administrative expense of roughly a hundred and ninety-three million included more than a hundred and eighteen million of net share-based compensation expense, nearly forty million of outside professional services, and a five million contract cancellation fee. The second-quarter figure was smaller in part because executive departures triggered the reversal of previously recognized stock-based expense, producing a net credit of more than fifteen million, and the first-half loss also absorbed a roughly twenty-five million non-cash charge from the extinguishment of a term loan with an equipment finance company in the first quarter. The income statement, in other words, overstates the cash cost of the business, and the balance sheet is where the real story lives.
The balance sheet at the end of the second quarter shows total assets in the low one-point-sevens of billions, of which the bulk is property, plant, and equipment, net, up sharply from year-end, with cash and cash equivalents down to the low sixties of millions from four hundred and eight million at the start of the year, and a new restricted-cash line in the high twenties of millions. Total liabilities are in the mid-sevens of hundreds of millions, of which debt is in the low fives of hundreds of millions, and stockholders' equity is in the low one-point-zeros of billions with an accumulated deficit in the mid-threes of hundreds of millions. The share count is in the low six-hundreds of millions, and the company's debt is secured by the financed equipment, whose value exceeds the related obligations. The most important financial fact in the filing is the going-concern disclosure. Management concluded that the company's cash and restricted cash at the end of the second quarter, measured against forecasted disbursements under the current operating plan, are not sufficient to satisfy obligations within one year of the filing date, raising substantial doubt about the ability to continue as a going concern. The mitigating factors cited are the post-quarter-end convertible note issuance, undrawn committed borrowing capacity, the ability to monetize equipment whose value exceeds the related debt, and the intent to sequence capital expenditures with definitive tenant agreements and the corresponding project-level financing. One equipment facility includes a mandatory prepayment requirement if an approved customer agreement has not been received by the end of the year, which is a live covenant trigger that the TensorWave lease is designed to satisfy. The convertible note offering in July is the pivotal capital event. The company issued more than four hundred million of five percent convertible senior notes due in 2031 in a private offering to institutional buyers, including an additional tranche on full exercise of the initial purchasers' option. Net proceeds were in the high four-hundreds of millions after discounts and estimated expenses, and in connection with the offering the company entered into capped-call transactions at a cost in the mid-thirties of millions, which effectively raise the initial conversion price, in the low single digits per share, to a cap in the mid-teens. The notes carry no scheduled amortization and no financial maintenance covenants, and the capped-call structure means that existing shareholders face minimal dilution unless the stock rises well above the current level, at the cost of a meaningful slice of the net proceeds. The stated use is general corporate purposes, which in practice means bridging the construction period until project-level financing can be secured against the TensorWave lease.
The execution risk is concentrated in a narrow window. The closing of the TensorWave lease by the end of September, subject to an extension provision, the securing of project-level financing sufficient to fund construction of the facility, and the delivery of the phases on the target schedule that runs from late 2027 into the first quarter of 2028, form a narrow execution window. The lease effectiveness is subject to conditions including execution of work letters and guaranties, finalized operations schedules, board approvals, and the landlord obtaining project-level financing, and if the conditions are not met by the closing date or its extension, either party may terminate the lease. This is the single most consequential date in the company's near-term calendar, and the Keystone prepayment covenant described earlier makes it a solvency event as well as a commercial one.
The capital requirement for the TensorWave facility is not yet fully quantified in the public filings, but the company has already invested more than a billion and a half in site buildout as of the second quarter, and the first-half capital outflows were largely for campus infrastructure and turbine acquisition rather than for the building itself. The Hillcore framework, if consummated, would reduce the direct capital for more than two gigawatts of generation, but the data center building, the electrical infrastructure, and the site improvements remain the company's capital responsibility. The company has stated that it is pacing future outlays with commercial commitments, which is the right framing and also the highest-risk one, because the first commercial commitment is the one that requires the most capital.
The nuclear timeline, with the first unit now targeted for the mid-2030s rather than a year earlier, is a long-duration execution risk that does not affect the near-term thesis but does affect the long-term narrative. The regulatory milestone of the pilot environmental review is a genuine de-risking event, and the completed front-end study provides initial cost and schedule estimates, but a first unit that is still several years out remains subject to the full licensing process. The nuclear program is a narrative asset and a real option, not a near-term cash flow, and the one-year slip on the first unit should be read as a schedule, not a commitment. The management transition is a live execution risk, and it consumed a large share of the company's first year as a public company. The board removed its chief executive officer from the role in mid-April and terminated the executive for cause at the end of April, its chief financial officer resigned in the same period, an interim co-presidency and an interim finance chief carried the company through the summer, and the board appointed a permanent chief executive officer in early August. The new chief executive has served on the board since the fall and has more than forty years of infrastructure experience across a major engineering firm and a nuclear developer, and the appointment of a new general counsel who is also the founder and chief executive of a related-party counterparty in the Firebird litigation is a governance complication that the company discloses and that shareholders should monitor.
The REIT question is a forward-looking variable with option value. The company deferred its election and is a C corporation for the current and next tax years, and if it eventually elects REIT status the capital structure would need to be restructured to comply with the distribution and asset-composition requirements, shifting the investor base toward income-oriented buyers. If it never elects, the equity remains a development bet with no dividend expectation. The deferral is not a negative signal in isolation; it is a signal that the company is not yet in a position to commit to that framework, and the timing of any election is one of the named thesis variables alongside the lease closing, the Hillcore definitive documentation, the first delivery phase, and the resolution of the active litigation.
The primary downside scenario is a failure to close the TensorWave lease by the end of September, or its extension, which would remove the company's first contracted revenue stream and likely trip the equipment-finance prepayment covenant and aggravate the going-concern doubt. In that scenario the company would hold more than a billion and a half of physical assets with no contracted revenue, more than half a billion of debt, and a going-concern warning, and the equity would be a pure optionality bet on finding a replacement tenant and restructuring the capital stack. The market cap would likely compress toward a net asset value of the physical assets minus the debt, which is a floor but not a reliable one given the specialized nature of the assets and the thin market for them.
The second downside scenario is a successful lease closing followed by a construction delay or cost overrun that pushes the first delivery phase beyond the outside date, triggering the tenant's termination right and the rent-credit escalations. In that scenario the company would have the capital committed to construction but no revenue to service it, and the going-concern issue would resurface with more debt on the balance sheet. The completion guaranty the company has provided means it is on the hook for construction performance, and a delay would also damage its credibility with the Hillcore framework and with any future tenant negotiations, compounding the financing risk.
The third downside scenario is a governance escalation. The former CEO's litigation, the proxy contest that preceded it, and the bylaws amendment imposing a supermajority requirement for shareholder amendments to specified provisions are all symptoms of a contested control dynamic. If the former CEO's team gains board representation or obtains an injunction that disrupts operations, both the management transition and the construction timeline would be impaired. The company discloses that these actions have caused and are expected to continue to cause substantial costs, and to divert management's attention, and the five active legal matters collectively represent a material distraction and a potential cash drain. The litigation stack is worth itemizing because each matter has a different risk profile. A securities class action filed in early this year alleges materially false and misleading statements in the offering documents and other public statements relating to tenant demand and funding arrangements, covering a window in the final quarter of last year, and names the company, certain directors and officers, and certain underwriters, with the company unable to estimate a loss range. A brokerage-commission dispute over the sale of a turbine contract, now proceeding after the court denied a motion to dismiss in early August, seeks a compensatory amount in the high single-digit millions plus exemplary damages. A New Jersey equipment vendor's breach-of-contract claim seeks more than eight million, and a petition from two consultants alleging the wrongful reclamation of units in the predecessor entity seeks relief alleged to exceed four hundred million, which the company has deemed not probable as of the quarter-end.
The counterargument to the bear case is that the TensorWave lease, with a parent guaranty and a pending global infrastructure guaranty, is a fundamentally different asset from the uncontracted pipeline the company had when the stock was at its high. The market priced the equity at that level on the promise of the campus, and the contract converts a portion of that promise into a binding obligation. The convertible note, with its capped call, is shareholder-friendly capital that extends the runway without immediate dilution, and the landed turbines are a real, physical asset that can be monetized if the construction plan fails. The bear case is that the company cannot execute, and the bull case is that the contract, the capital, and the assets are sufficient to get through the construction phase, and the equity is currently priced for a high probability of failure. The hardest risk to model is the interaction between the going-concern disclosure and the market's perception of solvency. A going-concern warning in a pre-revenue development company is not rare, but it is a signal that management and the auditors believe current resources are insufficient for the next year without the mitigating actions, and those mitigants include a plan to sequence capital with tenant agreements, which is a plan and not a guarantee. The risk is that the market treats the warning as a red flag regardless of the mitigants, and the stock trades at a distressed multiple even when the fundamentals are improving, which is the dynamic that has carried the price to the low end of its range.
The valuation framework for a pre-revenue development company is not a traditional multiple exercise, because there is no revenue, no EBITDA, and no internally consistent trailing or forward multiple. The appropriate framework is a probability-weighted valuation of the assets and the contract, with three named scenarios. The current market cap, at the recent share price on the outstanding share count, sits in the low single-digit billions, below the book value of the physical assets plus the post-quarter-end cash and notes, and well below the contracted revenue of the lease. The market is pricing the equity as a high-probability failure scenario, and the question is whether the probability of the base case is underpriced. In the bear case the TensorWave lease does not close, or it closes but the construction fails to deliver on schedule. The physical assets are impaired in a distress scenario, the existing debt plus the new convertible notes rank senior to the equity, and the specialized nature of the turbines, the site, and the half-built infrastructure means the recovery rate in a liquidation is uncertain. A reasonable bear-case value for the equity is a small fraction of the current market cap, and this is the scenario the market is currently pricing with a high probability, with the stock near the low end of its fifty-two-week range as the clearest expression of that view.
In the base case the TensorWave lease closes by the end of September, project-level financing is secured, and the first delivery phase is achieved in late 2027 or the first quarter of 2028. The company begins recognizing revenue in 2028, the going-concern doubt is alleviated, and the equity re-rates toward a multiple of the contracted revenue stream, which averages to roughly four hundred million per year over the initial term. The base case implies a market cap meaningfully above the current level, because the current price does not fully reflect the probability of successful execution, and the capped-call structure on the convertible notes reduces the dilution overhang that would otherwise cap the re-rating.
In the bull case the lease closes, the Hillcore framework is consummated, the first delivery phase is on schedule, and the tenant exercises the expansion option to triple the facility power by the end of 2028. The contracted revenue more than doubles, the generation capacity is sufficient to support the expanded footprint, and the equity re-rates toward a multiple of the expanded stream, with the nuclear program and a favorable resolution of the REIT question adding option value. The probability of the full bull case is low, and it is not meaningfully priced in, but the tail is real, and it is the asymmetry between the low-probability bull and the high-probability bear that defines the risk-return of the equity.
The convertible note structure is a valuation input that cuts both ways. The initial conversion price is well above the recent share price, and the capped-call strike is further above still, so the note holders hold substantial equity upside and the company has paid a meaningful amount to cap that upside for existing shareholders. The coupon is a reasonable cost of capital for a pre-revenue company carrying a going-concern warning, and the absence of amortization and maintenance covenants preserves flexibility. The valuation conclusion is that the equity is a binary option on execution, the current price reflects a high probability of the bear case, and the base case is underpriced if the lease closes and the construction proceeds on schedule, but the downside is severe and the going-concern warning is a live variable that the market may not forgive.
Fermi Inc. is a high-risk, high-reward equity that has moved from a pure narrative at its offering to a contract-and-construction story in under a year, with the share price moving from a high in the mid-thirties to a low in the mid-fours and back to the mid-fives in the same period. The thesis of this report is that the TensorWave lease is the load-bearing fact, a multi-billion-dollar, fifteen-year, binding contract with a parent guaranty and a pending global infrastructure guaranty, and that the convertible note issued after quarter-end is the second load-bearing fact, because it extends the runway and removes the immediate solvency risk that the going-concern warning highlights. The named thesis variables are the lease closing and project financing, the Hillcore definitive documentation, the first delivery phase, the REIT election decision, and the resolution of the active litigation.
The counterargument, stated plainly, is that the company is nineteen months old, has no revenue, carries a going-concern warning, lost its chief executive and chief financial officer in the same quarter, is the subject of five lawsuits including one from its former chief executive, and is asking the market to trust that a new chief executive can deliver a multi-hundred-megawatt data center on a schedule the market has already doubted once. That counterargument is legitimate and the report does not dismiss it. The bear case is that the company cannot execute and the physical assets are a partial floor but not a reliable one, and the bull case is that the contract, the capital, and the assets are sufficient and the equity is priced for a high probability of failure.
The judgment is that the probability of the base case is underpriced at the current market cap. The market has made a large bear move, and the current price near the low end of the fifty-two-week range reflects a high probability of the bear case, but the contract, the capital, and the assets make the base case more probable than the price implies, and the risk is real and the downside is severe. The equity is appropriate for investors who can hold through a binary outcome and who are not sensitive to the governance turbulence, and it is not appropriate for investors who require a revenue base, a dividend, or a clean balance sheet. The final assessment is that the stock is a high-probability option on execution, and the current price does not fully reflect the chance that the option finishes in the money.