SLNH
SLNH · Soluna Holdings, Inc. · Equity Research Report · as of August 13, 2026
Key figures, each tagged by provenance.
| Metric | Value | Basis |
|---|---|---|
| Share price | $1.35 | Market Aug 12, 2026 close, up 6.30% on the day |
| Market capitalization | $330.2M | Estimate Aug 12, 2026 close on 244,590,575 shares, common only |
| Revenue (TTM) | $42.1M | Filed FY2025 $29.7m plus H1 2026 $24.45m less H1 2025 $12.09m, from the filings |
| Revenue (Q2 2026) | $15.06M | Filed 10-Q, Aug 13, 2026. Up 60.3% sequentially, 144.6% year over year, or 73.1% excluding a $4.4m presentation change |
| HPC revenue (Q2 2026) | $0.0M | Filed 10-Q, Aug 13, 2026. Still zero |
| Net loss (Q2 2026) | ($22.62M) | Filed 10-Q, Aug 13, 2026. $(20.71M) attributable to Soluna |
| Net loss (FY2025) | ($56.99M) | Filed 10-K, filed late Mar 2026 |
| Cash, unrestricted | $113.36M | Filed 10-Q, balance sheet at Jun 30, 2026 |
| Net cash | $80.2M | Filed $113.36m unrestricted cash less $33.12m carrying debt at Jun 30, 2026. This row carried an estimate of ~$5–10M reconstructed from deal terms and it was wrong by about $105 million, see section 07 |
| Shares outstanding | 244,590,575 | Filed 10-Q cover page, at Aug 10, 2026 |
| Series A liquidation preference | $123.0M | Filed 4,920,045 shares at $25.00, unchanged, 10-Q at Jun 30, 2026 |
| Series A dividends in arrears | $35.1M | Filed 10-Q, at Jun 30, 2026 |
| Operating capacity | ~192 MW | Filed 10-Q at Jun 30, 2026, across three active locations. Supersedes the 171 MW sum of five site phases at the Jul 14 update |
| Capacity under construction | 14 MW | Filed 10-Q at Jun 30, 2026, at Kati 1 |
| Stated development pipeline | ~6,300 MW | Filed 10-Q at Jun 30, 2026, of which over 1,600 MW in advanced development. Was 4,300+ MW at the FY2025 release |
| Short interest | ~14.1% | Market 34.41m shares at the Jul 31, 2026 close, reported then as 21.8% of the 157,747,354 shares outstanding and 33.8% of float, 1.85 days to cover. Restated here against the 244,590,575 now outstanding; no newer reading is carried |
| Beta | 5.13 | Market 52-week range $0.41 to $5.14 |
The shape matters more than any single figure. All of the operating capacity serves Bitcoin customers. The AI and HPC capacity the equity story rests on, Kati 2 at 350 MW and Dorothy 3 at 300 MW, had a definitive development joint venture at Kati 2 signed June 3, 2026 with Metrobloks, one signed tenant letter of intent, and zero signed leases as of the Q2 report of August 13, 2026.
The company's own presentation sizes both of those campuses differently, and this report did not carry either figure. Its investor presentation whose pipeline page is stamped “as of August 1, 2026”, the deck carrying no cover date prints, at page 25, “150-300 MW TARGET CAPACITY” for Dorothy 3. The 300 MW above is the top of that range read as a point figure, and the bottom of it is half the top, on a line this report calls the capacity the equity story rests on. Estimate At page 23 the same deck heads Kati 2 “300MW+ Tier III AI Campus Plan” and describes approximately 100 MW of critical IT load targeted for service in 2027 with “at least 200 MW more” in 2028. Neither the 350 MW here nor the companion calendar's 100 plus 250 matches that. Estimate The two are not reconciled here. The figures above come from the filings and these come from the issuer's current presentation; on Kati 2 the deck's own wording is open-ended, so “at least 200 MW more” does not exclude 350. What a reader should take is that the campus figures this valuation leans on are targets with a range beneath them, not fixed capacity.
| Stage | Capacity | Share of pipeline | Basis |
|---|---|---|---|
| Operating, earning revenue | ~192 MW | 3.0% | Filed 10-Q at Jun 30, 2026 |
| Under construction | 14 MW | 0.2% | Filed 10-Q at Jun 30, 2026, at Kati 1 |
| Advanced development | 1,600+ MW | 25% | Filed 10-Q at Jun 30, 2026 |
| Stated total pipeline | ~6,300 MW | 100% | Filed 10-Q at Jun 30, 2026 |
The operating figure, reconciled. Several figures circulate and each is correct at its own date. The five operating sites carry a nameplate of 171 MW: Dorothy 1A 25, Dorothy 1B 25, Dorothy 2 48, Sophie 25 and Kati 1A 48. The company stated total operational capacity of 123 MW once Dorothy 2 was fully energised, which is the same four sites before Kati 1A, and Kati 1A adds the remaining 48 MW. The 147 MW under management cited in the Q1 release is that quantity at March 31, 2026, when Kati 1A was roughly half energised: 123 plus 24. Until the Q2 filing this document used 171 MW as the operating nameplate at the July update and cited 147 MW where the reporting date was 31 March. A third figure of roughly 159 MW appearing in an investor relations summary is undated and is corroborated by neither the filings nor the monthly updates; it was not used in the body and the masthead deck carried it anyway, which is corrected in this revision. The Q2 Form 10-Q supersedes all of this. It states approximately 192 MW of capacity across three active locations, Murray in Kentucky and Silverton and Willacy County in Texas, at June 30, 2026, with 14 MW under construction at Kati 1. That is a different basis from the 171 MW sum of five named phases, and the filing does not decompose it, so the change of basis is recorded here rather than a site list being inferred. Filed monthly updates of Dec 9, 2025 and Jul 14, 2026; Q1 2026 results release, May 18, 2026
What the company sells, to whom, and how it is paid.
Wind and solar plants periodically produce power nobody can take: the grid is congested, demand is low, or the transmission line is full. Rather than curtail and earn nothing, the generator would rather sell that energy cheaply to a load standing next to the plant.
Soluna is that load. It builds modular data centers behind the meter, co-located with renewable generation, so the electricity never touches the public grid. The data center buys power no grid-connected competitor can match on price, and the generator improves its capture rate. Soluna's control software, marketed as MaestroOS, decides when to run and when to stand down.
The model requires interruptible computing, work that can be paused for hours without breaking. Bitcoin mining is the archetype: when power gets scarce or expensive, the machines switch off and only that period's revenue is lost.
The central tension. AI training and inference are not interruptible on the same terms. A tenant renting 100 MW expects roughly 99.9% availability, not operation contingent on wind. The characteristic that makes the Bitcoin business cheap, flexible curtailable behind-the-meter power, is the one AI tenants are least willing to accept.
The company's answer is vertical integration and hybrid power: own the generation, add interconnection and firming, and deliver contracted uptime. That is what the April 2026 purchase of the Briscoe Wind Farm is for, and why Soluna announced a collaboration with Siemens on GPU power volatility. Whether the answer is sufficient will be settled by whether a creditworthy tenant signs a lease, not by argument.
Soluna does not own outright most of what it operates. Individual projects are financed through joint ventures with capital partners, principally Spring Lane Capital, and are consolidated as variable interest entities. The balance sheet shows the consequence: of $210.4 million of total equity at June 30, 2026, $54.5 million belonged to non-controlling interests, leaving $155.9 million attributable to Soluna Holdings. That share fell from 58% to 26% across the June quarter, because the company bought out the Dorothy 1A and Dorothy 1B minorities rather than because anything was revalued. Filed 10-Q, Aug 13, 2026
So that capacity is under management, not capacity wholly owned by the common. Revenue consolidates in full while a meaningful share of project-level economics accrues to partners. Through 2026 the company has been buying those partners out, consolidating the Dorothy 1 campus in April and May, which raises the shareholders' share of the economics and consumes cash to do it.
Consolidated revenue and consolidated book value both overstate what reaches SLNH shareholders while non-controlling interests remain at 26% of total equity, down from 58% before the Dorothy 1 buy-ins. No adjusted per-share economics are presented in the filings, so the size of that gap is an inference from the equity split rather than a disclosed figure.
Each stream, with its share of the total and its durability.
| Revenue line | Q1 2026 | Q1 2025 | Change | Share |
|---|---|---|---|---|
| Data hosting | 6,688 | 2,402 | +178% | 71.2% |
| Cryptocurrency mining, proprietary | 2,169 | 2,999 | −28% | 23.1% |
| Demand response services | 537 | 507 | +6% | 5.7% |
| High-performance computing | 0 | 28 | n/a | 0.0% |
| Total revenue | 9,394 | 5,936 | +58.3% | 100% |
Figures in $ thousands as filed. Filed 10-Q for the quarter ended Mar 31, 2026, filed May 15, 2026.
Data hosting, 71% of revenue. Soluna provides space, power and cooling; the customer owns the machines. Soluna is paid for power and service whether or not the customer's mining is profitable. Contracted, recurring, and not a direct bet on the Bitcoin price. It grew 178% year over year as Dorothy 2 and Kati 1A came online, and customers prepay or post deposits, which is why customer deposits sit as a liability.
Proprietary mining, 23%. Direct Bitcoin exposure, down 28%. Soluna runs its own machines at Dorothy 1B. The line moves with the Bitcoin price and network difficulty and is the lower-quality earnings stream. Management is deliberately shifting away from it.
Demand response, 6%. Small but structurally interesting. ERCOT pays Soluna for its willingness to shut down on command. High-margin revenue that exists precisely because the load is flexible.
HPC and AI: zero. Not small, zero. A prior attempt at an AI cloud business through a Hewlett Packard Enterprise GPU agreement failed in March 2025 and left a $19.3 million contract termination liability on the balance sheet, on which no payments have been made. The company has attempted this once already, at a different scale, and it went badly. Filed 10-Q, Note 6, Mar 31, 2026
Customer concentration. Three customers account for the substantial majority of revenue, with project-level concentration disclosed for Dorothy 1A, Sophie, Dorothy 2 and Kati 1. Blockware and Sazmining are the named counterparties; Blockware has expanded four times, surpassing 17 MW as of April 2026. Losing one hosting customer would be a material revenue event, and hosting customers in this industry are themselves leveraged to Bitcoin. Filed 10-Q concentration note; press releases Apr 21 and 28, 2026
Over roughly eighteen months, listed Bitcoin miners discovered their most valuable asset was never the machines but energised land with grid interconnection. AI demand has made large blocks of deliverable power the binding constraint across the sector.
The re-rating logic is simple. Bitcoin miners historically traded at roughly 6 to 12 times EV/EBITDA; established data-center operators trade closer to 20 to 25 times. Convert a mining site into a leased AI campus with an investment-grade tenant and the multiple re-rates. That arbitrage is the largest force acting on this group.
The market moved past announcements to signed, long-dated leases with creditworthy counterparties:
| Operator | Signed AI/HPC agreement | Reported contracted value |
|---|---|---|
| TeraWulf | Anthropic, roughly 401 MW, 20 years | ~$19bn |
| Cipher | AWS 300 MW; Google-backstopped Fluidstack | ~$9.3bn backlog |
| Applied Digital | CoreWeave, roughly 400 MW, 15 years | ~$11bn |
| Hut 8 | Fluidstack, 245 MW, 15 years | ~$7bn |
Peer figures are indicative of scale and contracting status as publicly reported between December 2025 and July 2026. Market Not verified against each peer's own filings.
By March 2026 the group had reportedly signed more than $70 billion of AI computing contracts in aggregate. Compass Point argued in July 2026 that the market still under-credits these signed leases, while noting that hyperscaler backstops create severe customer concentration.
Soluna has no signed AI lease. Its most advanced AI project, Kati 2, reached a signed letter of intent with one potential tenant as of the July 14, 2026 update. A letter of intent is ordinarily non-binding: it establishes intent to negotiate, not an obligation to lease.
Soluna is competing for tenants against peers who have already closed, at ten to a hundred times its scale, with completed or near-complete facilities and investment-grade credit support. The sector tailwind is real. Whether it lifts a company at Soluna's stage of the funnel is a separate question.
Through July 2026 the market stopped paying for the re-rating thesis. Over roughly a month Nebius fell about 36%, IREN about 37% and CoreWeave about 28%, with Applied Digital, Core Scientific and TeraWulf also down sharply, a 25% to 38% derating across the group.
Two observations matter more than the percentages. Applied Digital reported a 61% revenue beat and sold off. IREN announced a five-year, $3.4 billion NVIDIA cloud contract plus up to $2.1 billion of NVIDIA investment, and fell 37% over the month regardless. Reported drivers included competition fears after Meta signalled plans to rent out excess AI capacity, valuation concerns, and unease about the debt funding the buildout.
The bull case at Kati 2 assumes that signing a lease produces a re-rating. On July's evidence, signing may be necessary without being sufficient: a company that has not yet signed is being asked to clear a bar that is not currently paying out for those who have cleared it. This is an inference from peer price action, not a claim about Soluna's own prospects, and it would be falsified by a peer re-rating on a new contract.
Project financing is tightening. On July 29, 2026 Bloomberg reported that CoreWeave was seeking to raise the interest rate on a $2.6 billion loan tied to capacity for Anthropic, and credit-default-swap costs for technology borrowers have been rising. Long rates moved with it: the 30-year Treasury reached about 5.25% in late July, its highest since 2007, with the 10-year above 4.7%. Long rates are the discount rate against which a fifteen or twenty year lease is underwritten. If the best-capitalized operators are repricing debt upward, a first-time sponsor of Soluna's size faces a harder market than the sector narrative implies. Market as reported Jul 29 to 31, 2026
Siting risk is no longer theoretical. On July 14, 2026 New York's governor imposed a moratorium on new large data center development statewide, citing utility bills and resource use. Soluna operates in Texas and Kentucky with no direct New York exposure, but the precedent is that a jurisdiction can close to new capacity quickly.
Speed to power is the currency, and incumbency compounds. Tenants award capacity to developers who can prove a delivery date. Every quarter a peer energises a campus, the pool of uncontracted demand available to a first-time AI developer narrows.
ERCOT summer peak mechanics reduce uptime. Texas sets certain transmission charges by a large customer's load during four coincident summer peak intervals, so flexible loads curtail heavily from June through September. The July update notes increased curtailment across Dorothy 1A and 1B, Dorothy 2 and Kati 1A. This is expected, and seasonally negative for the June and September quarters.
Given the curtailment disclosures across all Texas sites in the June and July updates, and Briscoe repair work still ongoing and delayed by high winds, Q2 2026 is unlikely to show the clean sequential revenue progression the company has highlighted for four straight quarters. This is an inference from disclosed operating conditions, not company guidance, and the August 13, 2026 report settles it.
Named peers on comparable metrics, not a general market description.
| Company | Signed AI/HPC leases | Indicative scale |
|---|---|---|
| IREN | Microsoft AI cloud partnership | ~$16.7bn market cap, 4.5 GW pipeline |
| TeraWulf | Anthropic, ~401 MW, 20 years | ~$19bn contracted revenue |
| Applied Digital | CoreWeave, ~400 MW, 15 years | ~$11bn contracted revenue |
| Cipher | AWS 300 MW; Fluidstack | ~$9.3bn backlog |
| Hut 8 | Fluidstack, 245 MW, 15 years | ~$7bn contracted revenue |
| Soluna | No tenant. A definitive joint venture with Metrobloks to develop Kati 2, signed Jun 3, 2026. | ~$330.2M market cap, ~192 MW operating |
Peer figures indicate scale and contracting status only. Market as reported Dec 2025 to Jul 2026, not verified against each peer's filings.
| Measure ($m unless stated) | Q2 2026 | Q1 2026 | Q2 2025 | FY 2025 |
|---|---|---|---|---|
| Total revenue | 15.06 | 9.39 | 6.16 | 29.7 |
| Total cost of revenue | 14.29 | 7.48 | 4.97 | n/a |
| Gross profit | 0.77 | 1.92 | 1.19 | 6.46 |
| Gross margin | 5.1% | 20.4% | 19.3% | 21.7% |
| Total general and administrative | 17.64 | 18.54 | 7.80 | n/a |
| of which stock compensation | 9.48 | 10.22 | n/a | n/a |
| Operating loss | (16.94) | (16.63) | (6.62) | (33.68) |
| Net loss | (22.62) | (17.90) | (7.78) | (56.99) |
| Net loss attributable to Soluna | (20.71) | (17.47) | (7.38) | (53.41) |
| Adjusted EBITDA, company defined | (1.56) | (2.09) | n/a | (13.2) |
| Loss per share | (0.18) | (0.24) | (0.93) | n/a |
| Weighted average shares | 130,975,761 | 84,101,320 | 11,146,141 | n/a |
| Operating cash flow | (5.19) | (6.37) | n/a | n/a |
Filed 10-Q filed Aug 13, 2026 for Q2 2026 and Q2 2025; 10-Q filed May 15, 2026 for Q1 2026; FY2025 10-K and results release, filed late Mar 2026. Quarter figures for Q2 are the filed three-month column; Q1 2026 reconciles as the filed half less that column. TTM revenue is computed from the filings rather than taken from a data provider.
The headline loss is mostly non-cash, but not costlessly so. Net loss more than doubled to $17.9 million, driven by $10.2 million of stock-based compensation against $1.8 million a year earlier. Strip out stock compensation and depreciation and the company's own adjusted EBITDA was a $2.1 million loss.
The nuance is that non-cash is not the same as costless. In Q1 2026 stock compensation exceeded total revenue. June 2026 equity grants totalled at least 12,631,694 restricted shares, roughly 8% of shares outstanding, across five insiders: Michael Toporek 7,310,534 taking him to 17,132,816; John Belizaire 3,018,802; William Hazelip 726,401 taking him to 1,702,859; Mary Jennifer O'Reilly 726,401 taking her to 1,703,675; and David Marusak 849,556, with director awards vesting on separation. Shareholders pay in dilution rather than cash, and the share count confirms it. Filed Forms 4, Jun 1 to 3, 2026
The dilution has been extraordinary and it accelerated in the June quarter. Weighted average shares went from 11.1 million in Q2 2025 to 131.0 million in Q2 2026. Shares outstanding rose from 10.6 million at end-2024 to 102.6 million at end-2025 to 157,747,354 by May 12, 2026, to 225,821,479 at June 30 and 244,590,575 by August 10, 2026. This is why loss per share improved from $(0.93) to $(0.18) while the absolute loss grew: the denominator grew faster than the numerator. It is also why a market capitalization that rose 73% since the previous revision reflects a 12% move in the price and a 55% move in the count.
Historical per-share figures are not comparable. Soluna executed a 1-for-25 reverse stock split on October 16, 2023. Any price chart predating that is not comparable to today's prices, and the share count has since been rebuilt many times over through at-the-market sales, a standby equity purchase agreement with Yorkville, warrant exercises and restricted stock.
Gross margin is thin and definition-dependent. On the filed statement, 20.4%. Excluding depreciation from cost of revenue, roughly 44%. At least one major data provider displays a trailing gross margin near 66%, which does not reconcile to the filing on any basis. The filed figure governs.
Cash flow is the plainest read: $5.2 million of operating outflow in the June quarter and $11.6 million in the half, against $13.7 million of capital expenditure and deposits and $51.4 million for the Briscoe acquisition net of cash acquired, funded by $113.5 million of at-the-market proceeds, $24.5 million of note proceeds and $18.9 million from the standby facility. The half-year figures are stated directly here rather than a trailing reconstruction, because the acquisitions make a twelve-month roll-up of this business a comparison between two different companies. The business is funded by capital markets and project partners, not by itself.
Internal controls: material weakness, and still open at June 30, 2026. In the FY2025 Form 10-K the CEO and CFO concluded disclosure controls and procedures were not effective, and the Q2 2026 Form 10-Q reaches the same conclusion at June 30, 2026 on the same weakness. Management states it cannot be considered remediated until the remedial controls have operated long enough to be tested. The identified errors concerned classification of current versus long-term debt, a lease classification and valuation, and an overstatement in deposits on equipment and current liabilities. KPMG was appointed for FY2026, replacing UHY LLP, which audited FY2025 and FY2024. Some of the identified errors touch exactly the balance-sheet lines this analysis depends on. Filed FY2025 10-K, Item 9A
Board structure limits shareholder recourse. The board is divided into three classes serving staggered three-year terms, so only one class stands for election at any annual meeting and replacing a majority takes at least two successive years. Directors can be removed only for cause, and only on the affirmative vote of 75% or more of shares entitled to vote. Special meetings can be called only by the chairman, chief executive or president, or on written request of a majority of the board or of voting power. Filed Form S-3, Apr 2026
Shares, preferreds, ATM capacity, shelf, warrants, convertibles.
Comparing SLNH to peers using any standard data provider means using a balance sheet that no longer exists.
| Item ($m) | Jun 30, 2026 | Note |
|---|---|---|
| Cash, unrestricted | 113.36 | Was 68.57 at Mar 31; the increase is at-the-market issuance, not operations |
| Restricted cash | 17.93 | $10.0m current, $7.9m non-current; covenant and deposit purposes |
| Total assets | 293.51 | Includes $137.8m property, plant and equipment, up $63.0m on the Briscoe purchase |
| Total debt, carrying | 33.12 | $37.92m principal less $4.80m unamortised costs. All of it classified current on the covenant matters below |
| HPE contract termination liability | 19.35 | Unchanged. Unpaid since Mar 2025 |
| Total liabilities | 81.82 | n/a |
| Equity, Soluna Holdings | 155.93 | Includes Series A preferred. Additional paid-in capital rose $140.6m in the half |
| Equity, non-controlling interests | 54.45 | Outside project partners, now 26% of total equity, down from 58% on the Dorothy 1 buy-ins |
| Accumulated deficit | (405.89) | Cumulative historical losses |
Filed 10-Q, condensed consolidated balance sheet at Jun 30, 2026, filed Aug 13, 2026.
What the 8-K/A disclosed about Briscoe. The amended filing of June 17, 2026 puts total cash consideration at approximately $55.9 million against the $53.0 million headline, and discloses the asset's 2025 revenue of $8,950,927 against a net loss of $14,773,838. That loss sits against management's projection of $6 million to $11 million of year-one adjusted EBITDA. On the same pro-forma basis, combined 2025 revenue would have been $35,081,000 with a net loss attributable to Soluna of $54,129,000. Filed 8-K/A, Jun 17, 2026
Working from the disclosed transaction terms, the estimated cash impact across the June quarter is:
The acquisition total, reconciled. The three acquisitions sum two ways depending on which Briscoe figure is used, and the difference is the whole of the gap. On the restated basis, $55.9m plus $16.5m plus $8.765m is $81.2 million. On the headline purchase price, $53.0m plus the same two is $78.3 million, which is where a rounded $79 million came from. This document states $81 million throughout, on the restated total cash consideration in the amended filing, because that is what left the balance sheet. Filed 8-K/A, Jun 17, 2026; 8-K, Apr 2, 2026; press release, Apr 16, 2026; 8-K and Exhibit 99.1, May 20, 2026
That totalled roughly $59 million to $64 million of cash out against $68.6 million of unrestricted cash at the start of the quarter, and this document inferred something in the region of $5 million to $10 million, $8 million at the midpoint, with the +$40.6 million net cash position at 31 March plausibly becoming net debt. The Q2 Form 10-Q shows that inference was wrong by about $105 million. Unrestricted cash was $113.4 million at June 30, 2026 and carrying debt $33.1 million, so net cash of $80.2 million. The outflow arithmetic above was not the error; the error was treating an opening balance less known outflows as a closing balance, in a company whose cash flow statement shows $113.5 million of at-the-market issuance over the same six months.
That inference would have described a company nearly out of money. It was not, and the reason is the second half of the picture, which this document did identify.
Those were calculations from disclosed deal terms, not company figures, made before a post-acquisition balance sheet existed. The Q2 10-Q settled it, and against the calculation. They are kept here, marked wrong, because the reasoning that produced them is the reasoning a reader should be able to check.
The acquisitions were funded by selling stock. The share count makes this unambiguous.
| Date | Shares outstanding | Added | Basis |
|---|---|---|---|
| March 31, 2026 | 111,717,040 | n/a | Filed 10-Q balance sheet |
| April 6, 2026 | 113,181,690 | +1.5m | Filed prospectus, Apr 6, 2026 |
| April 24, 2026 | 141,347,055 | +28.2m | Filed Form 424B3, ~May 12, 2026 |
| May 12, 2026 | 157,747,354 | +16.4m | Filed 10-Q cover page |
Roughly 46 million shares, about 41% of the starting count, were issued in six weeks, and 28.2 million of those came in the eighteen days between 6 and 24 April alone. The June quarter then added more than the six weeks did: 68.1 million more shares to June 30 and 18.8 million after it, on roughly $23.6 million raised under the at-the-market program following quarter end. That window brackets the Briscoe and Dorothy 1A closings precisely. It is the funding source for the acquisitions, and it is why the company is not in the distress the cash arithmetic alone would suggest. For scale on price, the same 6 April prospectus reported a last sale price of $0.9259, so a large part of that issuance was done below a dollar.
Liquidity and dilution are not two questions, and Q2 2026 settled it in the direction this paragraph anticipated. The cash balance was comfortable, $113.4 million, and the share count moved 55% to produce it. The company had the cash because it sold the shares, and of the two figures the share count is the one that cannot be reversed. Read the two together or neither means anything.
The 9.0% Series A Cumulative Perpetual Preferred is the quietest large item in the structure. At June 30, 2026 there were 4,920,045 shares outstanding with a $25.00 liquidation preference, $123.0 million, and approximately $35.1 million of cumulative dividends in arrears, the share count unchanged across the quarter. Dividends have not been paid since roughly late 2022, and cumulative means the unpaid amounts accrue rather than lapse, at about $11.1 million a year.
Combined, roughly $158 million of claims rank ahead of the common against a $330.2 million common market capitalization, about 48% of the equity beneath it. That ratio improved because the common grew, not because the claim shrank: the claim grew by $2.7 million of arrears over the quarter and will keep doing so. The preferred also carries a protective provision requiring two-thirds holder approval before issuing anything senior to it, and became callable from August 23, 2026, a right the company is in no position to exercise.
What the preferred market price implies. Insider filings show a director selling Series A at $10.75 and $11.00 per share on 17, June 18 and 22, 2026, and the chief executive selling small amounts at $7.25 in April and $10.40 in late May under a Rule 10b5-1 plan. At roughly $11 the preferred trades near 44% of its $25 face. The market for the senior security is not pricing full recovery of face plus arrears. Filed Forms 4, Apr to Jun 2026
The Generate credit agreement carries a minimum trailing debt service coverage ratio of 1.60:1.00 and a minimum forward contracted ratio of 1.20:1.00, secured by first-priority liens on substantially all project assets, with a mandatory cash sweep applying a percentage of free cash flow to principal. The 10-Q states compliance as of that filing and separately notes a limited waiver obtained in connection with diligence requests. Coverage covenants tied to project cash flow are the mechanism by which a weak operating quarter becomes a financing problem. Filed 10-Q, Note 8
The machinery is fully loaded: an at-the-market program with H.C. Wainwright dated April 29, 2025; two standby equity purchase agreements with YA II PN dated August 12, 2024 and March 24, 2026; and a shelf registration filed March 9, 2026 registering up to $1.0 billion of securities, of which $500 million is carved out for an at-the-market sales agreement prospectus. Registered capacity is roughly three times the entire market capitalization. It was five times before the June quarter, and it fell because the capitalization rose, not because any registration lapsed.
| Instrument | Capacity | Basis |
|---|---|---|
| Shelf registration, total | $1.0bn | Filed Form S-3, Mar 9, 2026, amended Mar 30, 2026 |
| of which ATM sales agreement prospectus | $500m | Filed Form 424B5, Apr 1, 2026 |
| Resale registrations, April 2026 | 31.66m sh | Filed 26,512,815 issuable to Yorkville under the 2024 standby agreement, of which 3,000,000 issued by 1 Apr; 2,459,400 warrant and Harmattan merger shares; 2,700,000 Generate warrant shares. These are resale registrations and produce no cash for the company |
| Warrants outstanding | 27,549,840 sh | Filed Form 424B3 |
| Generate warrant package | 2,700,000 sh | Filed a pre-funded warrant plus common warrants, issued with the $12.5m Tranche C loan |
| Company ATM illustration | 760,178,588 sh | Filed shares issuable were the ATM fully sold at $0.77 |
Schedule 13G filings show Robert L. Bugbee moving from 9,168,000 shares, 9.3%, to 13,336,362 shares, 8.5%. He bought roughly 4.2 million shares while his percentage fell, which is dilution outrunning accumulation by a single holder. Insiders hold roughly 35% of shares outstanding. Beyond that single holder the register has four more reportable positions and one exit, and the filings divide on the rule they are made under rather than on the schedule. BlackRock, Inc. reported 8,321,125 shares, 5.3%, on July 30, 2026 for a June 30, 2026 event date, with sole voting power over 8,242,809 and sole dispositive power over all of them. It is the only statement here filed under Exchange Act Rule 13d-1(b), the institutional route, and the only institution on the register that is not also a financing counterparty. Filed Schedule 13G, Jul 30, 2026
Three further holders each report exactly 9.99%, and all three are counterparties to this company's own financing rather than outside investors: Generate Capital, PBC and its funds at 3,345,829 shares on September 18, 2025, the lender under the credit agreement described above; YA II PN, Ltd. and the Yorkville entities at 1,059,641 shares on February 6, 2025, the standby equity purchase and promissory note counterparty; and Intracoastal Capital LLC with Mitchell Kopin and Daniel Asher at 2,997,785 shares on July 18, 2025. Intracoastal then amended to nil on November 7, 2025, a complete exit. Alpha Capital Anstalt reports 1,045,318 shares, 1.531%, on November 12, 2025. Filed Schedules 13G and 13G/A, Feb 2025 to Jul 2026
Three unrelated holders reporting the identical figure of 9.99% is the signature of a contractual beneficial ownership blocker rather than a coincidence of position sizes. On that reading the 9.99% is a ceiling on what each may hold at any moment through conversion or exercise, not a measured block of common stock, and it is not comparable with the 5.3% BlackRock reports or with the 8.5% held outright by the individual holder above. The filings state the percentage and do not state the blocker, so the ceiling is inferred from the figure and from the instruments these counterparties hold.
One Schedule 13D has been filed and it is closed out. Wayne W. Boos reported 2,010,000 shares on August 1, 2023, rose to 2,910,000 by September 5, 2023 and fell to 145,000 by February 14, 2024, amending twice more to April 26, 2024. Item 4 states the shares are held for investment purposes and disclaims plans under every one of subsections (a) through (j), so the activist schedule carries no activist purpose here. No Schedule 13D has been filed since. Filed Schedule 13D of Aug 1, 2023 and its amendments to Apr 26, 2024 Filed Schedule 13G, Feb and May 2026
Selling has concentrated in the Series A preferred rather than the common. Director William Phelan sold 18,000 preferred shares on 17, 18 and 22 June at $10.75 to $11.00, leaving 106,233. Chief People Officer Mary Jennifer O'Reilly sold her entire preferred holding of 6,800 shares on 10 and 11 June at $10.80 and $11.10, ending at zero. Chief Accounting Officer Jessica Thomas sold 2,500 preferred at $10.90 and 4,838 common at $1.70. The chief executive sold 1,000-share lots in April and May under a Rule 10b5-1 plan adopted December 16, 2025. The pattern is consistent: officers exiting the preferred, not the common. On the buying side, chief financial officer Michael Picchi purchased 100,000 common shares at $1.63 on May 22, 2026, an open-market purchase rather than a grant, and the most substantive insider signal of the period. Insiders hold roughly 35% of shares outstanding. Filed Forms 4 for W. Phelan, J. Belizaire, J. Thomas and M. O'Reilly, Apr to Jun 2026
The chief executive's June 1, 2026 restricted stock grant of 3,018,802 shares is compensation, not a purchase, and separately the Compensation Committee raised his base salary to $600,000 retroactive to January 1, 2026 with a target bonus of 100% of salary. Filed 8-K, Jul 24, 2026, Item 5.02(e)
Multiples shown with their basis and pricing date. No target price.
There is no usable earnings multiple: the company has never earned money and EBITDA is negative. That leaves revenue multiples, asset-based measures and capacity-based measures, each with a caveat.
| Metric | Value | Basis |
|---|---|---|
| Share price | $1.35 | Market Aug 12, 2026 close, above the 50-day average of $1.34 and below the 200-day of $1.43, both struck at the 12 Aug close |
| Market capitalization | $330.2m | Estimate 244.59m shares at Aug 10, 2026 against the 12 Aug close, common only |
| Enterprise value | $250.0m | Filed $330.2m plus $33.12m carrying debt less $113.36m unrestricted cash, all at Jun 30, 2026. No longer reconstructed |
| Enterprise value, as previously reconstructed | ~$238m | Estimate superseded. It added ~$53m of debt and deducted ~$8m of estimated cash; the filed figures are $33.1m and $113.4m |
| Price / sales, TTM $42.1m | 7.9× | Estimate the multiple rose although revenue did, because the share count rose faster |
| EV / sales, TTM $42.1m | 5.9× | Estimate down from ~7.2× reconstructed, because the cash was there |
| Market cap less net cash | $250.0m | Estimate the same quantity as enterprise value here, shown because $80.2m of net cash is a quarter of the market capitalization |
| EV per operating MW | ~$1.3m | Estimate $250.0m divided by ~192 MW operating |
| Price / book, reported total equity | 1.6× | Estimate total equity of $210.4m still includes $54.5m of non-controlling interests |
| Price / book, attributable equity | ~2.1× | Estimate $330.2m divided by $155.9m attributable equity |
Price at publication is the close on the last trading day before the action, which is what the firm could see when it struck the figure. Where the action carries a date on which no session traded, the last session that did still supplies the close: it always resolves, which is the reason for the convention. One action here is dated a Saturday.
Two ratings exist on the stock. Both are the published views of the firms named and neither is adopted here.
| Firm | Analyst | Date | Rating | Target | Price at publication | Basis and disclosed conflict |
|---|---|---|---|---|---|---|
| Zacks Research | n/a | Jul 21, 2026 | Hold, upgraded | n/a | $1.36 | Market The close on July 20, 2026, the last session before the action. A quantitative ratings service rather than broker research: model-derived, no target, no named analyst. Not carried before this revision, and it is the most recent action on the stock. |
| H.C. Wainwright | Mike Colonnese | Jul 20, 2026 | Buy | $4.00 | $1.13 | Market The close on July 17, 2026, the last session before the action; the shares rose 20% on the day the note was published. The conflict here is filed, not merely reported. The Q1 2026 Form 10-Q records that the at-the-market offering agreement through which the company sells newly issued shares is with H.C. Wainwright & Co., LLC, dated April 29, 2025. Filed The only Buy on the stock is published by the firm earning commissions on its share issuance, and a higher price directly increases the proceeds available under that facility. At least one wire described the action as coverage assumed rather than initiated, which suggests an analyst handover, and some services still show a superseded $5.00 attributed to a different Wainwright analyst. |
| Weiss Ratings | n/a | Jun 12, 2026 | Sell, D−, upgraded from E+ | n/a | $1.42 | Market The close on June 11, 2026, the last session before the action. A quantitative ratings service rather than broker research: model-derived, no target, no named analyst. |
| Wall Street Zen | n/a | Apr 25, 2026 | Hold, upgraded from Sell | n/a | $1.20 | Market The close on April 24, 2026, the last session before the action: the grade carries April 25, 2026, a Saturday, on which no session traded, so the preceding Friday supplies the close. A quantitative ratings service rather than broker research. |
A published aggregate over three contributors stood at Hold with a $4.00 target when it was read on August 3, 2026. The close on July 31, 2026 was $1.22, the last session before the aggregate was published, and August 3 closed at $1.22 as well, so the figure is the same on either reading. Its composition has changed since this file first read it: one Buy, one Hold and one Sell, where it was one Buy and one Sell before the Hold of July 21 was added. It is that service's own published figure and is not a center struck in this document. Market
Wainwright's own published estimates project losses throughout: $(0.11) for Q2 2026, $(0.45) for FY2026 and $(0.27) for FY2027. Its $4.00 target implies a market capitalization near $631 million on 157.75 million shares, roughly 19 times trailing revenue and about 228% above the 31 July close. The target and the loss estimates are consistent only on the assumption that the pipeline converts, so the target rests on conversion rather than on forecast earnings. That is an observation about the published work, not an adoption of it. Market as reported Jul 20 to 23, 2026
The conflict attaching to that coverage is recorded in section 14.
At roughly 6 to 8 times sales the market is not valuing the Bitcoin hosting business. A stable contracted hosting operation of this size with thin margins and no growth would command a low single-digit revenue multiple. The premium is paid for the AI option: Kati 2, Dorothy 3, and the 6.3 GW behind them.
That option has a strike price, and the strike price is capital. Building 100 MW of AI capacity costs well over a billion dollars. Soluna's equity market value is $330 million and filed unrestricted cash at June 30, 2026 is $113 million. That is a materially stronger starting position than this document previously described, and it is still perhaps a tenth of one 100 MW campus. Any conversion of the pipeline will be funded by project debt, third-party project equity and further issuance of SLNH shares. The AI upside is real; existing shareholders will own a materially smaller fraction of it than the pipeline headline implies.
Ranked by how much each would change the investment case.
Everything else is secondary. Kati 2 is planned as a 350 MW critical-IT campus in Willacy County, Texas, 100 MW in Phase I and 250 MW in Phase II, through a joint venture formed June 3, 2026 with Metrobloks: Soluna's subsidiary holds all Class A interests and manages the venture, contributing the Phase I property, a purchase agreement for the Phase II property and roughly $3.5 million of operating expenses, with total commitments reported around $21 million to $22.5 million. Phase I is 100 MW; Phase II adds 250 MW. The distribution waterfall returns capital plus a 14% internal rate of return plus $100,000 per Gross PPA MW, then splits 50/50 with Metrobloks. Filed 8-K, ~Jun 9, 2026
As of July 14, 2026: general contractor selected and onboarding, detailed design about 50% complete, long-lead electrical equipment letters of intent signed, and formal commercial negotiations with one potential tenant under a signed letter of intent. A binding lease with a named creditworthy tenant would re-rate the equity. Its absence is why the stock trades where it does.
A planned 300 MW AI and HPC campus near Silverton, Texas, on 300 adjacent acres secured under a definitive purchase agreement, next to Dorothy 1, Dorothy 2 and Briscoe. This is the clearest expression of the vertical-integration thesis: owned generation plus owned land plus adjacent operating experience. It is also early, with environmental due diligence, survey work, fiber studies and schematic design as of July 2026 and a community engagement consultant still being hired. Realistically a 2028 or later revenue contributor.
Management projected $6 million to $11 million of year-one adjusted EBITDA on $20 million to $24.4 million of annualised revenue. Against a $53.0 million headline price that implies roughly 5 to 9 times EBITDA. The 8-K/A of June 17, 2026 complicates the picture: total cash consideration was approximately $55.9 million, and the asset's 2025 accounts show revenue of $8,950,927 against a net loss of $14,773,838.
The $20 million to $24.4 million annualised revenue projection should be read carefully. Briscoe's primary purpose is to power Soluna's own co-located data centers. Power sold internally is an intercompany transaction that eliminates on consolidation, appearing as lower cost of revenue rather than added revenue. Only third-party power sales and renewable energy credit sales flow through the consolidated top line, so the asset may improve consolidated margin without adding the projected revenue to reported totals. The July update also disclosed repair work still ongoing and delayed roughly two weeks by high winds, requiring an additional crew and crane.
The nearest-term, highest-certainty driver. Kati 1 is an 83 MW site: K1A Galaxy at 48 MW is operating, and K1B at 35 MW is in phased construction with Phase 2 modular data centers of 9 MW substantially complete and Phase 3 of 14 MW described as ahead of schedule. This is Bitcoin hosting, so it extends the current business rather than transforming it, but it is contracted, incremental and visible.
Soluna was added to the Russell 3000 and Russell 2000 Value indices in the June 2026 reconstitution, and holds its first quarterly earnings conference call on August 13, 2026. That a company of this profile has never held an earnings call is itself notable; starting one is a governance improvement. Institutional ownership is around 8%, so there is room for it to rise. Ryan Carver, previously of Microsoft, was appointed Chief Development Officer in July 2026 to lead the AI and HPC platform, which is relevant credibility for tenant negotiations.
Severity-ranked, most severe first.
Shares outstanding went from 10.6 million to 244,590,575 in roughly nineteen months, including 46 million shares in the six weeks to May 12, 2026 and 87 million more in the thirteen weeks to August 10. An ATM program, two Yorkville standby equity facilities and a $1.0 billion shelf including a $500 million ATM prospectus all remain available. Operating outflow was $11.6 million in the half against $65.1 million of investing outflow, and filed unrestricted cash at June 30, 2026 is $113.4 million, raised rather than earned. Any Kati 2 development capital raised at the parent level dilutes further.
Zero HPC revenue, zero signed AI leases, one non-binding letter of intent, and one failed prior attempt leaving a $19.3 million unpaid liability. Competitors with completed facilities, investment-grade backstops and billion-dollar balance sheets are bidding for the same tenants. If Kati 2 does not secure a binding lease, the company reverts to a sub-scale Bitcoin hosting business priced at 6 to 7 times sales.
This risk read "unverified" until August 13, 2026. It is now verified, and against the estimate rather than for it. Unrestricted cash was $113.4 million at June 30, 2026, not the roughly $8 million section 07 reconstructed, with $17.9 million more restricted. The balance was not earned. Operating cash flow was an outflow of $11.6 million in the half against $113.5 million of at-the-market proceeds, so the cash position is a function of the share price and the appetite of buyers, and it is consumed by pre-development spending that produces no revenue. Covenants include a 1.60:1.00 trailing debt service coverage ratio and a cash sweep on free cash flow, and neither ratio was tested at June 30: Amendment No. 3 of August 7, 2026 removed the requirement, conditional on prepaying the Tranche A-1 and A-3 loans, done August 10. The entire debt balance is classified current in consequence. The severity stays Severe on the funding mechanism, not on the size of the balance.
On August 3, 2026 the Governor of Texas directed the Public Utility Commission of Texas and ERCOT to conduct a comprehensive verification and audit of all data centers advancing through ERCOT's interconnection process, and stated that any project failing to comply will be denied connection to the grid. ERCOT suspended the Batch Zero Large Load classification notifications due by August 7, 2026 in market notice M-A080326-01 and filed its request for good cause exceptions on August 10, 2026, ahead of the Commission's open meeting of August 20, 2026. It is pausing the Batch Zero study process itself, not only the notifications, and states that the verification "will take several months". No completion deadline has been named.
The exposure is the pipeline that carries the valuation premium, and the Q2 filing enlarged it. The pipeline was restated to approximately 6.3 GW, with capacity redesignated for AI at Rosa, Hedy, Ellen and Fei, and the Form 10-Q filed August 13, 2026 places Fei and a newly named Project Gladys in ERCOT interconnection planning in terms. Dorothy 3 at 300 MW sits behind them. It also reaches the same Commission tracked in the calendar on the large-load interconnection standard and the co-location rule. Both statuses were checked against the Commission's docket on August 13, 2026 and neither was what this file assumed: the co-location rule was adopted on March 26, 2026 as 16 TAC 25.205, five months before this document first described it as possibly pending, and the large-load standard is not adopted, its Project 59142 running to August 13 with no order. The audit's scope moved against this company on August 10: ERCOT's filing states that data centers are interpreted to include virtual currency mining facilities and other computational loads, and that community impact information will be sought from all such facilities of 25 MW or more that have not yet energised. What remains unestablished is whether any specific site here was submitted into Batch Zero.
None of this appears in any SEC filing by this issuer, and that was re-tested rather than carried. The Form 10-Q filed the same day as this revision names ERCOT eighteen times, and names the Public Utility Commission, the Governor, the directive and market notice M-A080326-01 not once. The severity here is moderate rather than higher because the June quarter removed the liquidity edge: a schedule delay against $113.4 million of unrestricted cash is a slower build, not a financing emergency.
$123.0 million of liquidation preference plus roughly $32.4 million of accruing arrears rank ahead of the common. Arrears grow at 9% of face annually while unpaid. Any resolution, whether cash payment, exchange into common or renegotiation, is likely to be expensive for common shareholders. The preferred trading near 44% of face suggests that market does not expect full recovery.
Roughly 23% of revenue is direct Bitcoin mining. Hosting revenue is contractually insulated in the short term but not economically insulated: if Bitcoin falls far enough, hosting customers stop paying, stop expanding, or fail. FY2025 revenue fell 21.8% year over year largely on weaker Bitcoin-linked economics. The current backdrop is unhelpful, with Bitcoin near a 21-month low around $58,000 to $64,000 after record spot-ETF outflows of roughly $4.5 billion in June 2026.
Three customers account for the substantial majority of revenue. Blockware alone exceeds 17 MW across four expansions. The loss or failure of a single counterparty would be material, and hosting customers in this industry are themselves leveraged to Bitcoin.
Management concluded disclosure controls were not effective as of December 31, 2025, with identified errors in debt classification, lease classification and valuation, and overstated equipment deposits. The auditor changed to KPMG for FY2026. A new auditor plus an unremediated material weakness plus a complex web of consolidated variable interest entities warrants extra scrutiny of reported figures, particularly since some identified errors concern exactly the balance-sheet lines this analysis depends on.
Four of five sites are in Texas. Summer coincident-peak curtailment reduces uptime by design. Interconnection queues gate Projects Annie, Fei, Ellen, Hedy, Rosa and Grace. Texas power market rules, transmission charge design and any data-center-specific regulation flow directly through to margin.
Multiple consolidated variable interest entities, Class A and Class B membership tiers, priority return waterfalls, a related-party relationship with Harmattan Energy whose President and director Matthew Lipman also sits on Soluna's board, and a $54.5 million non-controlling interest balance, down from $65.8 million on the Dorothy 1 buy-ins. Consolidated revenue overstates what reaches SLNH shareholders, and reported book value does the same.
Short interest was 34.4 million shares at the July 31, 2026 close, reported then as 21.8% of shares outstanding and 33.8% of float, at 1.85 days to cover. Against the 244,590,575 shares on the August 10 cover page the same figure is about 14%, and no more recent reading is carried here. Beta is 5.13 and average daily volume roughly 15.5 million shares. This produces violent moves in both directions on modest news, as the $0.41 to $5.14 twelve-month range demonstrates and as the 6.15% single-day decline on 31 July illustrates in miniature. The stock spent much of 2025 below $1.00 and regained Nasdaq minimum bid compliance only on May 1, 2026; at the 12 August close of $1.35 the buffer is roughly 35%.
Each case with its preconditions: what must be true, not what might be.
The letter of intent lapses. Better-capitalized competitors absorb available demand and Kati 2 cannot be financed on acceptable terms, a 350 MW campus being too large for a $330.2 million company to fund without transferring most of the economics. Simultaneously the operating business weakens: summer curtailment compresses revenue, Bitcoin softens, a hosting customer defaults or fails to renew. Cash of $113.4 million at June 30, 2026 is drawn down by pre-development spending and operating burn toward the level where the 1.60:1.00 coverage covenant and the cash sweep bind. That path is longer than this document previously described but it is the same path, and the June 30 test was already not taken: it was waived on August 7 and conditioned on a prepayment. Financing then arrives at whatever price is available, which at $1.35 into a count that has already grown 55% this year is punishing. The $19.3 million HPE liability and $158 million of preferred claims remain, senior and unaddressed, the latter growing by about $2.8 million a quarter. In the most adverse version, sustained trading below $1.00 puts the listing back in question.
Requires: no AI lease, a weak Bitcoin tape, and a financing need at a depressed price.
Kati 1B finishes and adds Bitcoin hosting revenue. Briscoe improves consolidated margin. Kati 2 advances through design and negotiation but a binding lease takes longer than the market hopes, a normal outcome for a first-time AI developer competing against completed facilities. Dorothy 3 remains in development. Revenue grows off a small base, adjusted EBITDA moves toward breakeven, and GAAP losses continue, consistent with Wainwright's published estimates of $(0.45) for FY2026 and $(0.27) for FY2027. The company returns to capital markets through the ATM, the Yorkville facilities or a project-level raise. Share count keeps climbing. The stock trades on Kati 2 headlines and Bitcoin beta, and the Series A arrears accrue unresolved in the background.
Requires: continued hosting execution while AI conversion slips, funded by further issuance.
The letter of intent converts into a signed long-dated lease with a named hyperscaler or well-funded neocloud. The engaged investment bank closes project financing at the joint-venture level. Briscoe's owned generation proves to be the differentiator management claims, letting Soluna offer contracted uptime at a cost structure grid-connected developers cannot match. Dorothy 3 follows on the same template with land and power already secured. The sector comparison becomes relevant rather than aspirational, and Soluna re-rates from a Bitcoin hoster at 5.8 times sales toward a contracted infrastructure developer. What must be conceded: this case assumes signing produces a re-rating, and July 2026 supplied evidence against that assumption. If the market now wants delivered cash flow rather than backlog, the bull case needs an energised revenue-producing campus, which is years rather than quarters away.
Requires: a binding Kati 2 lease with a creditworthy tenant, plus project financing that does not gut the parent's equity share.
These paths are not independent draws. The same conditions cluster: a tight financing market makes lease negotiations harder, weak Bitcoin reduces hosting demand and depresses the share price used to raise equity and pressures coverage covenants. Modeling the catalysts as separate probabilities overstates the number of ways this works out. Nearly everything routes through one variable, whether a creditworthy tenant signs.
Soluna Holdings is a genuine if sub-scale renewable-powered data center operator that has grown revenue for five consecutive quarters to $15.1 million in Q2 2026 and runs approximately 192 MW across three active locations. Behind-the-meter siting and, since April 2026, ownership of the Briscoe Wind Farm give it a real power cost advantage. Among micro-cap AI-infrastructure stories, having revenue, customers and generating assets places it in a better category than most.
The valuation is not being set by that business. At roughly 8 times trailing sales the market is paying for the AI pipeline: Kati 2 at 350 MW, Dorothy 3 at 300 MW, and a stated 6.3 GW behind them. As of the Q2 report that pipeline had produced a definitive joint venture with Metrobloks to develop Kati 2, signed June 3, 2026, and still no signed tenant lease and no HPC revenue, in a sector where competitors have closed multi-billion-dollar agreements with Anthropic, AWS, CoreWeave and Fluidstack, and where through July 2026 even those signed contracts stopped supporting share prices as the peer group derated 25% to 38% in a month. Converting even Phase I of Kati 2 requires capital exceeding the entire market capitalization by a wide margin.
Three things are easily missed. First, the dilution: 10.6 million shares to 244,590,575 in about nineteen months, with the ATM, two Yorkville facilities and a $1.0 billion shelf still live, 46 million shares issued in the six weeks that funded the acquisitions and 87 million more in the thirteen weeks after. Second, the $158 million of Series A claims, $123.0 million of preference plus $35.1 million of accruing arrears, ranking ahead of a $330.2 million common stock, with that preferred trading near 44% of face. Third, that essentially every publicly displayed valuation metric is built on a 31 March balance sheet that roughly $81 million of April and May acquisitions rendered obsolete.
The FY2025 material weakness, the concurrent auditor change to KPMG, and the still-unpaid $19.3 million liability from the previous failed AI compute attempt all argue for verifying figures at source rather than relying on summaries, including this one.
The question this reduces to. Will a creditworthy tenant sign a binding lease at Kati 2, and on terms that leave today's shareholders with a meaningful share of the economics after the capital required to build it?
Everything else is second order. The Bitcoin hosting business roughly funds the lights. Briscoe improves the cost structure. The 6.3 GW pipeline is worth approximately nothing until a tenant with a credit rating signs a document, and it grew by 2 GW in a quarter without one. If that happens, the current valuation will look small. If it does not, this is a sub-scale Bitcoin hoster with a preferred overhang priced at nearly six times sales.
Primary documents are free at sec.gov/edgar under Soluna Holdings, Central Index Key 0000064463. The company posts monthly operating metrics at solunacomputing.com/investors.
Requested categories with nothing to report, stated explicitly rather than omitted.
| Category | Status | Basis |
|---|---|---|
| Analyst coverage | Covered above | One broker and three quantitative ratings services are reported in section 08 as market data, each with its date and the close on the last session before it: H.C. Wainwright Buy at $4.00 dated Jul 20, 2026, and grades from Zacks Research of Jul 21, 2026, Weiss Ratings of Jun 12, 2026 and Wall Street Zen of Apr 25, 2026, alongside a published consensus of one Buy, one Hold and one Sell. Coverage is thin but it is not absent, and only one contributor is broker research. |
| Price target derived by this document | None | Checked against this document's own fifteen sections and found empty. No target is derived anywhere. Third-party targets are reproduced as market data, attributed and dated, and are not adopted. |
| Rating issued by this document | None | Checked against this document's own fifteen sections and found empty. No buy, sell or hold recommendation is issued. |
| Common dividend | Not established | Never examined. The document records $2.1m of Series B preferred dividends paid on conversion and roughly $32.4m of Series A arrears, but no source was checked for a common dividend. Also named in what was not checked. |
| Share buybacks | Not established | Never examined. The document establishes that the company is a structural net issuer with $1.0bn of registered capacity, but no source was checked for a repurchase program. Also named in what was not checked. |
| Securities class action | Not established | Never examined. No litigation of any kind is addressed in this document and no docket or legal proceedings disclosure was checked. Also named in what was not checked. |
| Credit rating | Not established | Never examined. No rating agency was checked, so the absence of a rating in this document is a gap rather than a finding. Also named in what was not checked. |
| Going concern opinion, FY2025 | Not established | The opinion itself was not read. UHY LLP audited FY2025 and FY2024, and the EXPERTS paragraph incorporated into the April 2026 Form S-3 references that report without going-concern language, which is suggestive but is not the opinion. Also named in what was not checked. |
Pricing basis. All price-dependent figures are struck at the August 12, 2026 close of $1.35. The exchange reports that session closed and stamps it at 16:00 Eastern, and volume of 9,370,569 shares corroborates it at about 0.90 times the mean of the nine preceding sessions. The 13 August session had also closed when this revision was written and the exchange had not yet published it, so 12 August is the latest completed session this document can read. Share counts are the Form 10-Q cover figure at August 10, 2026.
Tags record where a claim came from and nothing else. They are not color-coded, because color is reserved for meaning and identity and provenance is neither. A Filed fact can be terrible news.
| Tag | What it asserts |
|---|---|
| Filed | Stated in an SEC filing or company release, cited by form and date. This also covers a dated official publication by a named non-SEC issuer, such as an exchange notice, a central bank statement or a schedule published by its organizer, with the issuer named in the source line. What the tag asserts is unchanged: a specific body published this, on this date, and the reader can go and look. It never covers a third party's characterisation of someone else's facts, which is Market. |
| Estimate | Derived or inferred here. The arithmetic is shown. |
| Open | Expected but unconfirmed. Nothing filed either way. |
| Market | Price, volume, float, short interest, and published targets and ratings, stamped with the close or publication date. A third party's target or rating is market data, attributed to the firm that issued it and never adopted here. |
Primary sources. The Form 10-Q for the quarter ended June 30, 2026, filed August 13, 2026, and the results release of the same date; the Form 10-Q for the quarter ended March 31, 2026, filed May 15, 2026; the FY2025 Form 10-K; 8-K current reports of 20 January, 31 March, 2 April, 15 April, 17 April, 18 May, 20 May, 9 June, 17 June (8-K/A), 14 July, 24 July and July 30, 2026; Forms 3 and 4 filed April to June 2026; Schedule 13G filings of February and May 2026; the prior-year DEF 14A; the Series A prospectus supplement; the Form S-3 filed March 9, 2026 and amended March 30, 2026, the further Form S-3 of April 29, 2026, and Forms 424B5 of April 1, 2026 and 424B3 of April and May 2026. Company press releases and monthly business updates dated 10 March, 30 March, 2 April, 9 April, 10 April, 16 April, 21 April, 28 April, 1 May, 7 May, 18 May, 20 May, 9 June, 23 June, 14 July, 16 July and July 30, 2026.
Secondary sources. S&P Global Market Intelligence for market data at the July 31, 2026 close, and the exchange's own daily record for the closes the price-dependent figures and the coverage table are struck at. Sector context from publicly reported peer lease disclosures and commentary between December 2025 and July 2026, including the July 2026 peer derating, the reported CoreWeave loan repricing, the New York data center moratorium and US Treasury yield levels. Sell-side actions as reported July 20 to 23, 2026.
Per the Q1 2026 10-Q, Soluna's at-the-market offering agreement, the facility through which it sells newly issued shares into the open market, is with H.C. Wainwright & Co., LLC, dated April 29, 2025. On July 20, 2026 H.C. Wainwright published the only Buy rating on the stock, with a $4.00 target.
Nothing about this is improper, and research and banking coexist under disclosure rules at every firm. Wainwright's own estimates candidly project losses through 2027. But a bullish note from the firm earning commissions on the company's share issuance is not independent corroboration, and a higher share price directly increases proceeds available under the ATM. Read that firm's own disclosure section before assigning weight.
Section 07 was thinned in this revision. It previously carried several figures that this document had never examined and that reached it from the companion catalyst calendar: the placement agent warrants carried in mezzanine equity, the weighted average exercise price of the outstanding warrants, the component share counts inside the Generate warrant package, the Green Cloud conversion shares and price, the authorised common share count, the share warrant attached to the Dorothy 1A note, and the commitment fee shares settled under the standby equity facility. Section 09 lost the stated cost of the Kati 2 Phase II property on the same basis, and the pre-funded warrant share count inside the Generate package went the same way, leaving that row at its 2,700,000 total. A document is not entitled to a fact it never examined, in the same way a null row cannot claim an absence nobody checked; both assert evidence the file does not have. The figures have been removed rather than re-sourced, and the gaps are named here rather than backfilled from the companion.
What section 07 still covers on this document's own evidence: the March 31, 2026 balance sheet in full; the April and May acquisitions with their consideration, funding and dates; the reconstructed post-acquisition liquidity position and its arithmetic; the share count progression from December 31, 2024 to May 12, 2026 with the April issuance window; the Series A preferred, its liquidation preference, its arrears and what the preferred market price implies; the Generate covenants, the cash sweep and the security package; the standing issuance facilities by counterparty and date, the $1.0 billion shelf and the $500 million at-the-market prospectus, the April resale registrations and the total warrants outstanding; the Schedule 13G holder; and insider activity in both directions.
What it does not cover: the exercise prices and component structure of the warrant book beyond its total; the terms of the Green Cloud conversion; the authorised share count and therefore the headroom remaining before a further authorisation would be needed; and the specific consideration for the Kati 2 Phase II property. Each is a real disclosure in the filings and each is carried in the companion calendar on that document's own evidence. Restoring any of them here requires this document to examine the underlying filing, which is new research.
One limitation remains open. Four that stood at the previous revision have been closed and the reconciliations are recorded above and in the document log.
The FY2025 audit opinion has now been read, and it settles what the indirect evidence could only point at. The opinion is unqualified: the statements “present fairly, in all material respects” the financial position at December 31, 2025 and 2024 in conformity with United States generally accepted accounting principles. It carries no going-concern explanatory paragraph and no emphasis of matter, and the opinion paragraph runs straight into the basis for opinion with nothing between them. The earlier reading from the FY2024 comparison and the incorporated experts paragraph was right, and the document now stands in place of the inference. Going concern appears instead as a critical audit matter, the second of two, the first being the fair value of roughly 30.6 million warrants. The auditors record that the conditions “raised substantial doubt”, that management “developed plans to alleviate the substantial doubt” including debt and equity financing and, if necessary, cost reduction, and that they identified the evaluation of those plans as the matter requiring especially challenging judgment. That is a different disclosure from a qualified opinion: doubt raised by the conditions and addressed by management’s plans, not doubt left standing. The report is dated March 27, 2026 at Albany, New York, and the firm has served as auditor since 2021. Filed 10-K for FY2025, filed Mar 27, 2026; 10-K for FY2024, filed Mar 31, 2025
Facts and assumptions. Figures attributed to filings, releases and market data providers are facts as reported, with dates. Every derived or forward-looking estimate is carried in a violet assumption block and is not company guidance. Where sources conflict, both figures are shown and the discrepancy is recorded above.
Newest first. The original build entry is never removed or rewritten.
Dorothy 3 was carried at 300 MW, and the company states a range of 150 to 300 MW. The 300 was the top of the issuer's own target range read as a point figure, on a line this report describes as the capacity the equity story rests on. The range is now stated. Kati 2 is carried here at 350 MW and the same deck heads it a 300 MW plan, approximately 100 MW for service in 2027 and at least 200 MW more in 2028. Both are recorded as disagreements rather than resolved, because the figures already here came from the filings and the deck's Kati 2 wording is open-ended.
The deck carries no date on its cover, and is used only because it dates itself inside: its pipeline page is stamped as of August 1, 2026. That stamp is cited rather than a cover date the document does not have. It also predates this report's August 13 build, so nothing here is newer information; it was available and was not read. The as-of does not move and nothing is repriced.
The post-acquisition cash estimate was wrong by about $105 million, and it was load-bearing. This document reconstructed unrestricted cash after the three acquisitions at roughly $8 million, from disclosed deal terms, and expected the +$40.6 million net cash position at 31 March to have become net debt. The Form 10-Q for Q2 2026 puts unrestricted cash at $113.4 million and carrying debt at $33.1 million, so $80.2 million of net cash. The method is named rather than dropped: it netted disclosed acquisition outflows against a disclosed opening balance and did not model the at-the-market issuance that funded them, which the cash flow statement puts at $113.5 million for the six months. Every sentence that reasoned from the $8 million figure has been rewritten, in section 07, section 08, the risk register and the bear case, rather than left standing beside the new number.
The share count was the other half of the same correction, and it moved further than the price. 157,747,354 at May 12, 2026 to 244,590,575 at August 10, 2026, a 55% increase, against a 12% move in the price. The market capitalization therefore rises from $190.9M to $330.2M while the enterprise value rises only from a reconstructed ~$238m to a filed $250.0m, because the cash the reconstruction had spent was in fact raised and held.
The two Commission rules the entry above recorded as owed were fetched from the Commission's docket the same day, and neither status was what this file assumed. The co-location and net-metering rule was adopted March 26, 2026 as 16 TAC 25.205 under Project 58479, five months before this document described it as possibly pending. The large-load interconnection standard is not adopted; its docket is Project 59142, a number this file had never carried, which is why the status had never been checkable. The adopted co-location text puts its mandatory obligation on the generation resource rather than on the load and leaves behind-the-meter curtailment permissive, while the order's reasoning states that generation capacity is made available "first and foremost through the full curtailment of the large load customer", so the exposure is not withdrawn. Whether the rule binds the Briscoe and Dorothy arrangement turns on ERCOT registration facts this revision could not settle. The detail sits in the companion calendar.
Q2 2026 results, reported August 13, 2026 and filed the same day. Revenue $15.06 million, a fifth consecutive sequential increase, up 60.3% on Q1 and 144.6% year over year, or 73.1% excluding a $4.4 million presentation change that grosses up pass-through electricity in both revenue and cost of revenue. Gross profit fell to $766 thousand from $1.92 million on $1.5 million of Briscoe maintenance and the Kati 1 ramp, so gross margin went 20.4% to 5.1%. Net loss $(22.62) million against $(17.90) million, including a $4.2 million loss on debt extinguishment. Adjusted EBITDA improved to $(1.56) million from $(2.09) million. Briscoe's first full quarter under ownership produced $366 thousand of wind energy generation revenue against $2.25 million of cost before depreciation: the wind farm reduced gross profit rather than lifting it, which is the opposite of what this document flagged to look for.
Two answers the market wanted were negative. The material weakness was not remediated: management concluded disclosure controls and procedures were not effective at June 30, 2026 on the same balance sheet classification and presentation weakness. Covenant compliance was not demonstrated but waived: Amendment No. 3 of August 7, 2026 provided that neither the Debt Service Coverage Ratio nor the Forward Contracted DSCR was required to be tested at the June 30, 2026 measurement date, conditional on prepayment in full of the Tranche A-1 and A-3 loans, which occurred August 10. The entire outstanding principal is classified as current in consequence, and Briscoe is first tested on September 30, 2026.
The operating and pipeline figures both moved, and the pipeline moved more. Operating capacity is stated at approximately 192 MW across three active locations at June 30, 2026, superseding the 171 MW sum of five site phases this document used from the July 14 update; the filing does not decompose the 192 MW, so the reconciliation note records the change of basis rather than pretending to a site list. The development pipeline is restated at approximately 6.3 GW from 4,300+ MW, with over 1.6 GW in advanced development, after redesignations for AI at Rosa 187 to 242 MW, Hedy 120 to 198 MW, Ellen 100 to 145 MW and Fei 120 to 240 MW. Kati 2 is no longer a letter of intent: a definitive joint venture with Metrobloks was signed June 3, 2026, Phase I at 100 MW with Phase II adding 250 MW, and the competitive position table is corrected, since it still read "None. One letter of intent at Kati 2." That was already wrong before this quarter.
A Texas regulatory review this document did not know about. On August 3, 2026 the Governor of Texas directed the Commission and ERCOT to audit every data center advancing through ERCOT's interconnection process, with non-compliant projects to be denied grid connection; ERCOT suspended Batch Zero classification the same day in market notice M-A080326-01, pending a good cause exception at the Commission's open meeting of August 20, 2026. Nothing about it is on EDGAR, and that was measured rather than assumed: the Form 10-Q filed the same day names ERCOT eighteen times and the Commission, the Governor, the directive and the notice not once. Carried as a new risk here and as a dated catalyst in the calendar. What the 10-Q does supply is the exposure in the company's own terms, placing Fei and a newly named Project Gladys in ERCOT interconnection planning.
Repriced to the 12 August close. $1.21 → $1.35, up 6.30% on the session against the one before, on volume of 9,370,569 shares, about 0.90 times the mean of the nine preceding sessions. The 13 August session had closed when this was written and the exchange had not published it, so 12 August is the latest completed session this document can read. Capitalization $190.9M → $330.2M on 244,590,575 shares. The buffer above the $1.00 minimum bid goes 21% to 35%. The claims ranking ahead of the common are $158.1 million, $123.0 million of preference plus $35.1 million of arrears, which is 48% of the common market capitalization where it was 81%: the ratio improved because the common grew, not because the claim shrank. The 52-week range of $0.41 to $5.14 was re-measured over the window ending 12 August rather than carried forward, and both ends survive it.
The price shown beside a third-party view was struck on the date the action carries rather than on the session before it. The price beside a published rating or target is what the firm could see when it struck the figure, so it is the close on the last trading day before the action; a close struck on the action date itself is set after the action is out and sometimes moves because of it. Zacks Research, July 21, 2026: $1.37 → $1.36. H.C. Wainwright, July 20, 2026: $1.36 → $1.13. Weiss Ratings, June 12, 2026: $1.48 → $1.42.
The Saturday row is why the convention resolves and the other one does not. Wall Street Zen's grade carries April 25, 2026, on which no session traded, so the old basis had no answer at all and the cell read that no close could be struck. The last session that traded before it is Friday 24 April, and it closed at $1.20. That figure was already in the row, described as the preceding close and shown as evidence that none could be struck; it is now the price at publication, which is what it always was.
Two figures did not move, and the reason is worth stating. The published consensus dated August 3, 2026 was struck against $1.22, and the last session before it, 31 July, also closed at $1.22, so the figure survives the correction and only its basis changes. The H.C. Wainwright row moved furthest: the shares rose 20% on 20 July, the day that Buy was published, so the same-day close carried the move the note was published into and the preceding close was $1.13.
Repriced to the 4 August close. The session ended at 16:00 Eastern, stamped by the exchange, on volume of 11,971,770 shares, about 1.10 times the mean of the nine preceding sessions. $1.22 → $1.21, capitalization $192.5M → $190.9M, price to attributable book ~4.1× → ~4.0×, and the buffer above the $1.00 minimum bid requirement 22% → 21%. The share of the equity ranking behind the $155M of prior claims is unchanged at about 81%.
The file described its own stamp, and the description had gone stale. The pricing basis read that figures were struck at the 31 July close and that the stamp was 2 August, while the masthead carried 3 August and the coverage table was struck on 3 August data. Three snapshot and valuation rows also named the 31 July close against figures that had moved. The basis now names one close, the 4 August one, and says what establishes that the session had ended.
Filing sweep re-run to August 4, 2026 against Central Index Key 0000064463. The index paginates: 1,547 filings across the current view and one overflow file it references, and the whole record was read rather than the recent page of it. Four filings since the previous sweep: a Form S-8 of 28 July, a current report and a Schedule 13G of 30 July, and a Form 4 of 4 August. Three notifications of late annual filing stand across the whole index and no NT 10-Q, NT 20-F or NT 40-F appears anywhere on it.
Third-party coverage extended, not rebuilt. The four entries already carried keep their firm, rating and target exactly. Added are the analyst, the date as its own column, the close on that date, and the disclosed conflict, which on this issuer is filed rather than reported: the Q1 2026 Form 10-Q names H.C. Wainwright as the counterparty to the at-the-market offering agreement of April 29, 2025, and that firm publishes the only Buy on the stock. A Zacks Research upgrade to Hold of July 21, 2026 is added, which the earlier reading had not carried and which is the most recent action; the published consensus is accordingly one Buy, one Hold and one Sell where it was one Buy and one Sell.
One date resolves to no price. Wall Street Zen carries an action dated April 25, 2026, which was a Saturday, so no close can be struck against it and the cell says so rather than borrowing the preceding session. Three of the four contributors are quantitative services rather than broker research and are now marked as such, so the count of covering firms is not read as four analysts.
The ownership discussion carried one holder and described the rest of the register as "institutions roughly 8%". Five further reportable positions are on file and were not recorded, including BlackRock at 8,321,125 shares, 5.3%, filed on July 30, 2026, the same day as the Form 8-K this document treated as the most recent filing. Three more holders each report exactly 9.99% and each is a counterparty to this company's own financing. A Schedule 13D also exists, filed by Wayne W. Boos in August 2023 and run down to 145,000 shares by February 2024. Section 7 now sets out the register with the rule each statement is filed under, because a position reported under the institutional route and a position capped by a contractual blocker are not the same kind of holding and were being aggregated into one percentage.
Six figures were tagged as published market data and are computed in this document. The masthead market capitalization, both body statements of it, the screened enterprise value, and the price-to-sales, EV-to-sales and price-to-book multiples all read Market. Each is derived here from a price and a filed input, so each now reads Estimate. The tag asserted that a data provider had published the figure; the arithmetic behind each was shown alongside it throughout, so the figures were checkable, but the label said where they came from and it said the wrong thing.
The statement that no late-filing notification had ever been made was false. This report recorded that no NT filing of any kind exists and treated it as informative for an issuer that had changed accounting basis and filer status. Three exist: a Form NT 10-K of December 30, 1996 and Forms NT 10-K/A of March 27, 2000 and March 27, 2001. They are old, and none bears on the current accounts, but the claim was a claim about the whole record. It was made against the 1,000 filings the current index view returns; the complete index for this issuer holds 1,546, and the three sit in the older portion that view does not reach.
The FY2025 audit opinion has been read and the question it was left open on is closed. The opinion is unqualified and carries no going-concern explanatory paragraph and no emphasis of matter. Going concern appears instead as the second of two critical audit matters, the first being the fair value of roughly 30.6 million warrants, with the conditions recorded as having raised substantial doubt and management plans recorded as alleviating it. The report is dated March 27, 2026 at Albany, New York. The earlier inference from the FY2024 comparison and the incorporated experts paragraph pointed the right way and is now replaced by the text itself.
Repriced to the August 3, 2026 close, at an unchanged price. The 3 August session has closed and the basis moves July 31, 2026 → August 3, 2026. The close is $1.22 on both dates, so market capitalization stays $192.5M and no figure derived from the price moves; what changes is the date the figures are struck on. The day traded 12,550,120 shares against a recent average of 17,575,327, and the exchange reported the session closed and the figure settled. The filing record was swept again on August 3, 2026 against the complete index of 1,546 filings, reached by issuer central index key and including the overflow page the current view does not return. Nothing has been filed after July 30, 2026, but two filings were made on that date rather than one: the Form 8-K already recorded, and the BlackRock Schedule 13G described in the correction above.
The acquisition total is stated once, as $81 million, on the restated total cash consideration in the amended filing of June 17, 2026. The figure of $79 million carried in the research summary and in the enterprise value note came from the $53.0 million headline purchase price and is withdrawn: $55.9m + $16.5m + $8.765m is $81.2 million, against $78.3 million on the headline basis.
Operating capacity is stated as 171 MW of nameplate at the July 14, 2026 update, being Dorothy 1A 25, Dorothy 1B 25, Dorothy 2 48, Sophie 25 and Kati 1A 48, with 147 MW cited where the reporting date is March 31, 2026 and Kati 1A was roughly half energised. The figure of roughly 159 MW used throughout the previous revision is withdrawn: it comes from an undated investor relations summary and is corroborated by neither the filings nor the monthly updates. Enterprise value per operating megawatt moves from $1.5m to $1.4m as a consequence, and operating capacity from 3.7% to 4.0% of the stated pipeline.
Net change since the previous entry. No repricing: July 31, 2026 was a Friday and the as-of date is the following Sunday, so the close the figures come from has not moved and the pricing basis is unchanged at $1.22.
Three unreconciled figure pairs closed. The acquisition total and the operating capacity are recorded in the correction above. The third was the trailing net loss: a data provider's $(78.97) million against $(63.3) million computed from the filings. The difference is preferred stock. Adding the Series A cumulative dividend of $11.070 million a year and the $2.1 million of Series B dividends paid on conversion accounts for $13.2 million of the $15.6 million gap, and the $2.48 million residual is consistent with a deemed dividend on that conversion. The provider reports net loss attributable to common shareholders; this document carries the filed attributable figure and now says what the provider's number includes.
Filing sweep. Swept 8-K, 6-K, 10-Q, 10-K, 20-F, 40-F, S-1, S-3, 424, 13D, 13G, the full NT series, Forms 3 and 4 and DEF 14A, most recent first, reaching August 2, 2026. Nothing has been filed since the 8-K of July 30, 2026. Three late-filing notifications do exist, a Form NT 10-K of December 30, 1996 and Forms NT 10-K/A of March 27, 2000 and March 27, 2001, none of them recent. They sit in the older portion of the filing index, which the complete record reaches only when the index is read in full: 1,546 filings, not the 1,000 the current view returns. The FY2025 annual report was located and dated March 27, 2026, which narrows the outstanding audit opinion question to the opinion text itself.
The previous entry follows unchanged.
This revision supersedes the 26 July build. Net changes from that version:
Liquidity, corrected twice. A third acquisition was added: Soluna paid $8,765,490 in cash for the remaining 49% of Project Dorothy 1B on May 19, 2026, referenced qualitatively in the original but absent from the cash arithmetic. Briscoe was restated: the 8-K/A of June 17, 2026 puts total cash consideration at approximately $55.9 million against the $53.0 million headline, and discloses the asset's 2025 revenue of $8,950,927 against a net loss of $14,773,838. Estimated post-acquisition unrestricted cash falls from roughly $16–19 million to roughly $5–10 million, $8 million taken as the midpoint. Section 6 now also carries the other half of that picture: roughly 46 million shares were issued between 31 March and May 12, 2026, about 41% of the starting count, which is how the acquisitions were paid for, so liquidity and dilution are one question, and the 13 August cash balance and share count must be read together.
Dilution capacity and compensation. Registered issuance capacity corrected from the $87.65 million previously cited to a $1.0 billion shelf including a $500 million ATM prospectus, separated into primary issuance versus resale overhang: roughly 31.7 million shares registered for resale across three April registrations, 27,549,840 issuable on warrants, and the company's own illustration of 760,178,588 shares were the ATM fully sold at $0.77. June 2026 equity grants totalled at least 12,631,694 restricted shares, about 8% of shares outstanding, not the 3,018,802 to the CEO alone previously reported; director awards vest on separation. CEO base salary was raised to $600,000 retroactive to January 1, 2026 with a 100% target bonus. Kati 2 economics were quantified: capital returned plus a 14% IRR plus $100,000 per Gross PPA MW, then 50/50 with Metrobloks.
Market data, sector and macro. All price-dependent figures reset from the 24 July to the July 31, 2026 close of $1.22 (−6.15% on the day): market capitalization $192.5 million, price/sales 5.8×, adjusted enterprise value approximately $238 million, EV/sales approximately 7.2×, price/book on attributable equity approximately 4.1×. Section 11 adds a market-capitalization discrepancy resolving to the Series A preferred trading near $11.24, or 45% of face, independently corroborating the June insider sale prices. Section 3 gains the July 2026 derating across the AI infrastructure peer group of 25–38%, in which Applied Digital beat revenue by 61% and IREN announced a $3.4 billion NVIDIA contract yet both sold off; tightening project-finance conditions including the reported CoreWeave loan repricing and a 30-year Treasury yield near 5.25%, its highest since 2007; and the New York data center moratorium of 14 July as siting precedent. The bull case in Section 10 is qualified accordingly: signing a lease may be necessary without being sufficient. Bitcoin's 21-month low near $58,000 was added to the risk section.
Ownership and nulls. Schedule 13G holder Robert L. Bugbee moved from 9,168,000 shares (9.3%) to 13,336,362 shares (8.5%), buying while being diluted. A systematic review across 8-K, 6-K, 10-Q, 10-K, 20-F, 40-F, S-1, S-3, 424, 13D, 13G, Forms 3 and 4 and the full NT series confirmed no NT, 13D, 6-K, 20-F or 40-F filings exist; the last three cannot, as Soluna is a domestic filer. Nothing has been filed since July 30, 2026. That review reconciled two figures used here: the three April resale registrations sum to 31.66 million shares (26.5m SEPA, 2.7m Generate warrants, 2,459,400 YA and HEL), and the June grant cluster sums to 12,631,694 (Toporek 7,310,534, Belizaire 3,018,802, Hazelip 726,401, O'Reilly 726,401, Marusak 849,556). It also added four details: Briscoe's nameplate is approximately 149.85 MW rather than a round 150; the board is classified into three staggered classes with directors removable only for cause on a 75% vote; Soluna director Matthew Lipman is also President and a director of related party Harmattan Energy; and UHY LLP audited FY2025 and FY2024, with the incorporated EXPERTS paragraph carrying no going-concern language.
Thirteen sections built from the Form 10-Q for the quarter ended March 31, 2026, the FY2025 Form 10-K, 8-K current reports and Forms 4 filed March to July 2026, and company press releases and monthly business updates through July 14, 2026, with market data at the July 24, 2026 close of $1.26 (market capitalization $198.8 million, price/sales 6.0×). That version established the analytical spine carried forward here: that the screened enterprise value of roughly $158 million rested on a 31 March balance sheet made obsolete by the April acquisitions, and reconstructed it at roughly $235 million; that approximately $155 million of Series A preferred claims, $123.0 million of liquidation preference plus roughly $32.4 million of arrears, rank ahead of the common and appear in no equity multiple; that HPC revenue was zero against a valuation priced on the AI pipeline; that the sole covering broker, H.C. Wainwright, is also the company's at-the-market sales agent; and that the FY2025 material weakness warranted verifying figures at source. It estimated post-acquisition unrestricted cash at roughly $16–19 million and flagged four discrepancies between third-party data and the filings.