A 53 megawatt platform whose financing capacity rises and falls with the share price it is meant to support

01Snapshot

Key figures, each tagged by provenance.

02Business model

Sphere 3D operates Bitcoin mining and data-center infrastructure. Block rewards run at 3.125 BTC following the April 2024 halving, plus transaction fees, and halve again roughly every four years.

The company was a virtualization and data-management software business before shifting to pure-play mining. Since the June 1, 2026 Cathedra merger it describes itself as a digital infrastructure company spanning 53 MW across five United States data centers in Iowa, Kentucky and Tennessee, targeting AI and high-performance computing alongside mining. Management has proposed renaming the company DarkHorse Technologies Inc., with the reserved ticker DRK, at a shareholder vote on August 24, 2026.

The merger issued 2,405,300 common and 1,387,117 preferred shares with an aggregate fair value of $7.266 million, struck at $1.90 per common share as of May 29, 2026. Former Cathedra security holders held approximately 33% of the voting rights against roughly 67% for pre-existing shareholders. Cathedra’s information circular states a different figure on a different basis: approximately 49% of the issued and outstanding shares on a partially diluted basis, assuming full conversion of the Series H and Series I preferred, exercise of certain options, and vesting of replacement and existing restricted units. Both hold on their own footing. The distance between 33% and 49% is the measure of what the preferred stock and the unvested equity carry, and it is the reason a voting figure and an economic figure should not be used interchangeably here. Holders of subordinate voting shares received 0.123014 of a common share each and holders of multiple voting shares 12.3014 common shares each, economically equivalent for both classes. Certain key Cathedra shareholders were subject to a 7% post-closing ownership cap, with consideration above the cap taken in the new series of non-voting preferred, which is where the Series I stock originates. The securities were issued unregistered in reliance on the exemption at Securities Act Section 3(a)(10).

The board is Tim Hanley as Chair with Marcus Dent, Kurt Kalbfleisch, Nicholas Gates and Joel Block, three of whom are independent. Mr Dent had served as a Cathedra director since 2021. Joel Block is chief executive, Kurt Kalbfleisch chief financial officer, Tiah Reppas chief accounting officer, and Thomas Masiero Head of Strategy. Cathedra’s shares were delisted from the TSX Venture Exchange and the OTCQB at the close on June 2, 2026, and Cathedra stated it would apply to cease to be a reporting issuer in Canada.

The 53 MW reconciles for the first time. Cathedra’s circular records 45 MW across data centers in Tennessee and Kentucky, having completed a new 15 MW Kentucky data center in late October 2025 that raised its capacity by 50%, implying 30 MW before it, with a proprietary fleet producing approximately 400 PH/s. Cathedra’s 45 MW plus the 8 MW Iowa site accounts for the 53 MW, and 400 PH/s plus 0.84 EH/s accounts for the 1.2 EH/s. Neither reconciliation appears in any SEC filing reviewed.

The operating logic is that electrical power capacity is the asset, and that it can be directed to whichever use pays best: proprietary mining, third-party co-mining, or eventually AI compute. The Bitdeer agreements of June 25, 2026 are the first executed instance.

Sources: Form 10-Q for Q1 2026, Overview; Form 424B5, July 31, 2026, Recent Developments; company releases of 1 and 25 June and June 30, 2026; definitive proxy statement, July 13, 2026.

03Revenue streams

All reported revenue is Bitcoin mining revenue, recognized at the spot price at contract inception. Three streams exist or are in prospect.

Proprietary mining. Bitcoin mined and periodically sold under a hybrid treasury policy: hold where possible, sell to fund working capital. Q1 2026 production was 25.3 BTC, down 17.0% year on year, with 26.2 BTC held at quarter-end against 37.3 at the start, the difference being sales exceeding production.

Co-mining. The Bitdeer agreements of June 25, 2026 put 30 MW to work across three Tennessee and Kentucky sites. Bitdeer supplies SEALMINER hardware and the parties share net mining proceeds on one-year renewable terms. Financial terms are undisclosed and have not appeared in a periodic filing.

Owned Kentucky operations. North Campbell Land Company is a wholly owned subsidiary operating roughly 15 MW inside the Hopkinsville Electric System Holland Substation, participating in a flexible-load program permitting rapid curtailment when TVA demand spikes.

AI and HPC. Not a revenue source. The company is evaluating positioning its TVA-region power for AI and HPC deployment. No compute contract has been signed.

Sources: Form 10-Q for Q1 2026, Overview, Revenue Recognition and Note 4; company release of June 25, 2026; Hoptown Chronicle and Kentucky New Era, July 27 to 30, 2026.

04Industry & market backdrop

Halving pressure. Block rewards halve roughly every four years, so miners must grow efficiency and scale to hold position. The next halving falls around April 2028.

Rising network difficulty. As more capable hardware is deployed, aggregate network hashrate grows and an existing operator must keep adding hashrate to hold its share of rewards. Difficulty adjusts roughly every two weeks.

Consolidation and the AI pivot. The dominant theme across 2025 and 2026: lower hashprice and rising competition pressured mining economics, and smaller operators have sought scale while repositioning as AI and data-center landlords.

Rate and liquidity conditions. The Federal Reserve held at 3.50% to 3.75% on July 29, 2026, a fifth consecutive hold, but on a 9 to 3 vote with three dissents favoring an increase. Bitcoin traded near $63,600 to $64,400, roughly 49% below its October 2025 peak, with more than $465M of spot-ETF outflows on 23 and 24 July.

Sources: Bitcoin protocol schedule; Federal Reserve FOMC calendar and the July 29, 2026 decision; market data as at the August 3, 2026 close; Form 10-Q for Q1 2026, Overview.

05Competitive position

Named peers on comparable metrics, not a general market description.

Sphere 3D is among the smallest listed miners. Larger United States peers include MARA Holdings, Riot Platforms, CleanSpark, Core Scientific, TeraWulf, Cipher Mining and Bitdeer, most of which operate hundreds of megawatts to gigawatts against this company’s 53 MW. At roughly 1.2 EH/s the company holds a fraction of a percent of global network hashrate. Its competitive levers are power cost and the optionality of its sites.

The Mawson precedent

Big Digital Energy is the closest comparable, and not only strategically. It was Mawson Infrastructure until April 2026, when, following an Endeavor Investor Group campaign, it appointed five new directors under a cooperation agreement, formed a Strategic Transactions Committee and renamed itself, moving to the BGDE ticker on 30 April. Endeavor participant Phil Stanley became chief executive. The same group filed a Schedule 13D on Sphere 3D on July 31, 2026.

The detail that matters for this file is what happened to the activist’s own position: it fell from about 44.9% to 30.0% between January and May 2026 as that company issued shares. An activist diluted by the financing pattern this company is now accelerating is a specific precedent, not a general analogy.

Sources: peer filings and disclosures; Mawson Form 8-K and press release, April 23, 2026; Schedule 13D, July 31, 2026.

06Financial performance

FY2024 net loss is derived: FY2025 losses were reported as 126.8% larger than FY2024, implying roughly $9.5M. Estimate

Revenue has fallen in each successive year. Bitcoin production fell 61% in 2025, from 286.3 to 111.6 BTC, driven by the April 2024 halving, a shift to lower-cost hosting and equipment upgrades. The Q1 2026 decline reflected temporary capacity reductions during a fleet refresh that was substantially complete by quarter-end.

Management published a forecast against the first of these years, and the outcome is the only measure this file carries of what its guidance is worth. The company’s January 2023 investor presentation, still the only deck on its investor relations presentations page when that page was read on August 20, 2026 set out a 2023 revenue forecast, premised on all 1.66 EH/s of miners at 29.5 J/TH being installed and hashing. Its downside case, at Bitcoin $15,000, was $30.4M; the base case at $20,000 was $39.5M and the upside case at $30,000 was $48.6M. Estimate FY2023 revenue was $21.9M, as the table above records, 28% below the company’s own downside case. The fleet behind the forecast never arrived either: the same deck projected 1,666 PH/s by April 2023, and the company was at 0.84 EH/s three years later, before the Cathedra merger lifted it to roughly 1.2. This is a record of the issuer’s forecasting, not a current claim about the business, and it is recorded because no other guidance track record appears in this file.

Revenue of $1.9M compares with $2.8M a year earlier. The Q1 2026 loss narrowed to $4.1M from $8.8M, or $1.18 per share, helped by a 23% reduction in general and administrative expense and the absence of a prior-year investment loss. Against that, cash stood at $3.1M on total assets of $21.7M, with $3.7M of operating cash burn in the quarter.

The second quarter turned the loss from a run-rate into an event. Revenue rose to $2.45M from $1.9M in the first quarter, the first period carrying any Cathedra contribution, and against $3.02M a year earlier. Everything below the revenue line went the other way. The company wrote down $7.04M of property and equipment and $0.59M of intangible assets, together three times the revenue of the period and neither of them present a year earlier, and general and administrative expense more than doubled to $4.83M from $2.13M as the merged company absorbed the costs of combining. Total operating costs were $16.27M against $5.65M. The net loss was $13.83M, or $2.68 per share, against $4.1M in the first quarter.

The liquidity answer is thinner than the quarter before it. Cash stood at $2.85M at June 30, 2026 against $3.71M at December 31, 2025, and the Bitcoin position fell to 20.5 coins worth about $1.2M from 26.2 coins at March 31. The company mined 29.0 Bitcoin in the quarter against 30.9 a year earlier. Operating activities used $9.02M of cash across the six months, of which roughly $5.3M falls in the second quarter once the first quarter’s $3.7M is taken out, so cash and Bitcoin together are under one quarter of runway at the period’s own rate. The merger placed $3.28M of goodwill on a balance sheet that carried none at the year end, and total assets were near flat at $24.87M against $25.12M.

The share count is not restruck here. This report prices off the August 3, 2026 close and uses 8,619,150 shares as at July 29, 2026. The Q2 cover states 8,704,816 as of August 11, 2026, which is eight days after the pricing date where the count in use is five days before it, so the newer figure is not plainly the better basis for this price. Every capitalisation, enterprise value, per-share and ownership figure below therefore stands on the July 29 count, and the newer one is recorded rather than applied.

Operationally the quarter read better than the revenue line. The fleet averaged 17.1 J/TH at March 31, 2026 with capacity of 0.84 EH/s, the refresh substantially complete.

The acquired entity, Cathedra, reported $21.2M of revenue and a $9.7M net loss for 2025, also with a going-concern warning. The combined outcome turns on integration and financing.

The combined basis, retrieved

The figures above are pre-merger standalone. A combined basis does exist and has since 3 June: unaudited pro forma condensed combined financial information was filed as Exhibit 99.4 to the closing Form 8-K, prepared under Article 11 of Regulation S-X, with Cathedra’s audited 2025 and 2024 statements at Exhibit 99.2 and its Q1 2026 unaudited statements at Exhibit 99.3. This report previously stated that no combined balance sheet would exist before the Q2 Form 10-Q. That was wrong, and the correction is recorded in the log.

Against standalone FY2025 revenue of $11.2M and a $21.5M loss, the combined basis carries $25.9M of revenue and a $30.8M loss. Revenue more than doubles; the loss grows by roughly half.

The liability line is the one that matters. Cathedra brought roughly $6.7M of liabilities against $2.6M of current assets, and transaction accounting adds a further $2.2M, taking combined liabilities to $10.6M against Sphere’s standalone $1.7M, a little over six times. The composition is payables, accruals, customer and contract liabilities and lease obligations rather than borrowings, which is consistent with management’s description of a debt-free balance sheet, but they are claims all the same and section 8 reprices the enterprise value on them.

Sources: Form 10-K for FY2025, March 27, 2026; Form 10-Q for Q1 2026, May 15, 2026, Overview, Note 4 Bitcoin and Liquidity; Cathedra FY2025 results.

07Capital structure & dilution

Shares, preferreds, ATM capacity, shelf, warrants, convertibles.

For a company that must keep selling equity to operate, the share count is the central financial fact rather than a footnote. It rose 154% in about seven months.

The shelf ceiling, and why it moves

With public float below $75 million the company may sell no more than one-third of public float in any rolling twelve months off its shelf. The 31 July prospectus supplement gives the arithmetic: float of $46,107,198 measured at $5.65, the 3 June price, a one-third cap of $15.37M, less $5,018,576 already sold, leaving $10,350,490. The facility was sized at $10,300,000, the maximum the rule allowed.

Float is re-measured at prevailing prices, so capacity moves with the stock.

The headline $10.3M facility may therefore prove far less available than it appears when float is next measured. For this issuer the share price is not only a valuation output but an input to whether financing exists at all.

The four dilution channels

The at-the-market facility. The 31 July Form 8-K discloses $5,131,036 raised across 2,172,789 shares through 30 July at roughly $2.36 average, equal to about 25% of the current count. Shares issued after Q1 went out at roughly $1.54. The prior $8.0M offering with A.G.P., against which the Q1 10-Q recorded roughly $1.7M drawn and about $6.3M nominally remaining, was terminated and replaced by an Amended and Restated Sales Agreement adding Maxim Group as a second agent at a $10,300,000 ceiling. At $2.46 the full facility implies about 4.19 million further shares, roughly 49% of the current count. Either party may terminate on two days’ notice.

Restricted stock. RSUs and restricted stock awards rose from 653,815 at 31 March to 1,686,834 at 29 July. Within that, 472,222 RSUs were granted in Q1 2026 at a $1.38 grant-date fair value, taking RSUs outstanding from 179,165 to 616,779, with a further 86,871 vested awards granted after quarter-end.

Series I Preferred. Non-voting, carrying 8% payment-in-kind dividends for three years with staged conversion rights, subject to a 7% ownership cap and a Nasdaq exchange cap. Conversion would add 1,387,117 shares.

Warrants. Three tranches expire deeply out of the money in August and September 2026. What remains afterwards is 873,643 warrants at $9.40 running to January 15, 2031, immaterial at $2.46 but a ceiling on any recovery above that $9.40 strike.

Executive equity awards

Two entitlements recorded in Cathedra’s circular appear in neither company’s SEC filings as reviewed. Joel Block held 1,447,584 Cathedra restricted share units which, alone among Cathedra’s units, did not vest on closing; they were exchanged for replacement units on the same economic and vesting terms, about 178,073 units at the 0.123014 ratio. Exhibit 99.4 confirms that figure exactly: it records that approximately 2.9 million Cathedra restricted units were outstanding at May 29, 2026 and that replacement units over 178,073 common shares were issued to a single holder, with the remainder accelerating. Their fair value was about $0.3M, recognized as post-combination expense over 2.4 years. Being issued at closing, they sit inside the 1,686,834 restricted stock figure at 29 July. Separately, under an employment agreement dated September 25, 2025 as amended, Mr Block is entitled to a bonus of US$1,600,000 payable on consummation of the arrangement and achievement of certain performance milestones, and not payable until those milestones complete. Against $3.1M of cash at 31 March that is a contingent claim on liquidity, and it is not a share-count item.

The chief executive received the larger of the two disclosed awards on appointment. The closing release of June 1, 2026 states that Joel Block is entitled to a one-time inducement equity award of 500,000 restricted stock units, settled in common shares, vesting bi-annually in four equal installments over two years with the first tranche at the six-month anniversary of the grant date, subject to continued employment and to Compensation Committee and board approval, granted under Nasdaq Rule 5635(c)(4). On a 1 June commencement the first tranche falls due around December 2026.

Taken with the chief financial officer’s grant below, disclosed executive restricted stock totals 750,000 units, roughly 8.7% of the 8,619,150 shares outstanding.

Insider grant with change-of-control acceleration

A Form 4 filed June 10, 2026 records that the chief financial officer received 250,000 restricted stock units on 8 June, vesting in four installments from December 2026 to June 2028 but accelerating in full on a Vesting Event, defined as a Change in Control or a dissolution, liquidation or wind-up. The grant predates the activist Schedule 13D by seven weeks. The acceleration clause bears on any campaign ending in a change of control, and on the wind-up branch of a going-concern outcome.

The transaction cost the combined company roughly $3.9M in one-off charges. Exhibit 99.4 records acquisition-related transaction costs of about $3.4M, comprising $2.4M of professional, legal and accounting fees and $1.0M of strategic advisory fees, plus a transaction bonus of $0.49M and $0.735M of accelerated restricted stock. Set against $3.1M of cash at 31 March, these are the charges that accompanied a merger presented as balance-sheet strengthening.

Fully diluted

The warrants are far out of the money and unlikely to convert. The restricted stock and Series I are not.

A fourth consolidation is already authorised

The reverse-split risk in this file has been framed as a consequence of the August vote. It is not. At the special meeting of May 15, 2026 shareholders passed a special resolution, requiring two-thirds of votes cast and carrying 1,231,229 to 220,731, or 84.57%, amending the current Ontario articles to permit a consolidation of the common shares on a one-for-up-to-five basis, at an exact ratio and a date to be determined by the Board in its sole discretion. The definitive merger proxy states no expiry for that authority. It remains unexercised, and it sits on top of the 1-for-10 consolidation effected on February 9, 2026.

Two consequences. A fourth reverse split, after those of 2017, 2018 and February 2026, requires no further shareholder vote and no continuance: it can be effected at any time by board resolution fixing the ratio. And the authority was granted at the meeting that approved the merger, so it has been in force since before closing while this file dated the risk to a vote three months later.

Governance change pending

The 24 August vote would move the company from Ontario to British Columbia. The new articles at Appendix A, Article 9.1.1.3 would permit directors to consolidate issued shares by directors’ resolution or ordinary resolution, which lowers the procedural bar further rather than creating it, and Article 9.3 would permit a name change by directors’ resolution. The Board may also abandon either proposal in its sole discretion after shareholder approval, and registered holders following the Ontario Business Corporations Act section 185 procedure may dissent and demand fair value in cash against a $2.85M balance at June 30, 2026.

Roughly a tenth of the register votes by contract

Cathedra’s circular records that the Key Holders, defined as Joel Block, Thomas Masiero and Jialin Qu, delivered voting agreements as a condition of closing. For 24 months following the effective date, each agrees to vote all common shares held at any meeting of shareholders in accordance with the recommendations of the board, subject to exceptions where a Key Holder is materially and disproportionately adversely impacted relative to other holders. On a June 1, 2026 effective date the undertaking runs to about June 1, 2028.

Two consequences. It covers the 24 August vote on the continuance and the rebrand, where Masiero and Qu together hold roughly 11.5% of shares outstanding and the chief executive is the third party to the agreement. And it is the governance counterweight to an activist holding 5.80%, with the qualifications that it is contractual rather than voluntary, that it expires, and that at this particular vote the activist’s own weight is far below 5.80% because most of the position was bought after the record date.

The circular also explains why both Key Holder positions sit just below the 5% threshold rather than far above it. Their Cathedra holdings were large: the Togetsu Trust held about 36.15% of Cathedra and the Thy Kingdom Trust about 34.63%, both trustees being Cathedra directors. At the 0.123014 ratio those would have converted to well over a million common shares each. The 7% ownership cap capped the common portion and routed the excess into Series I preferred, so the common holdings recorded below understate the economic interest of both holders.

Shareholder register

The board answered the register with a shareholder rights plan. Adopted August 7, 2026, with the Rights Agreement dated August 10, 2026 and TSX Trust Company as rights agent, it attaches one Right to each Voting Share outstanding at the Close of Business on August 20, 2026. A Take-over Bid is defined at 20% or more of the outstanding Voting Shares, which is the threshold to read this table against: no holder above is close to it today, and the plan raises the cost of any accumulation toward it without a formal bid. Filed on Form 8-K of August 10, 2026 under Items 1.01 and 3.03, with a Form 8-A12B registering the Rights the same day. Filed

Masiero and Qu both received their shares in exchange for Cathedra stock, so the merger created a concentrated former-Cathedra bloc. One correction matters to how the table reads: Thomas Masiero is Head of Strategy of the combined company, so his 5.96% is an insider holding rather than an outside one, and the largest position on the register belongs to an officer. He has also appeared publicly as the representative of North Campbell Land Company in the Hopkinsville zoning proceeding described in section 9. Adding insiders as a group, roughly a quarter of the company is held by identified parties, which reduces effective float and helps explain why modest buying produces outsized price moves.

One row on this table is not a filed position, and it should be read differently from the other three. The three holdings above 5% are Schedule 13D and 13G filings and can be checked against the documents. The Alpine Fox line cannot: no SEC filing discloses a position in this issuer by Alpine Fox or by Mike Alfred. A 199,000-share holding is 2.31% and sits below the 5% threshold that would compel a Schedule 13, so no such filing is owed. Alpine Fox Capital LLC does file Form 13F-HR quarterly, and its most recent, filed May 14, 2026 for the March quarter, does not name this issuer. The June-quarter 13F, due August 14, 2026, is the first document that could corroborate the position, and it is the item carried on the companion calendar as the institutional 13F disclosure date. The holder’s own statement could not be retrieved on August 2, 2026. A self-reported holding is evidence that a claim was made, not that it is accurate, so the row is retained with that provenance rather than removed, and it carries no weight in the subtotal above or anywhere else in this report.

The percentages above are restated. The filers computed theirs against earlier counts, 7,639,893 at 19 June for the two 13G filings and 7,641,767 at 8 July for the 13D, so each stake is smaller than filed: 6.72% to 5.96%, 6.54% to 5.80% and 6.24% to 5.53%. Two consequences follow. Dilution is measurable within weeks rather than hypothetical, and all three holders now sit close to the 5% threshold from above, so continued issuance could carry them below it without any of them selling a share.

What the activist can vote on 24 August is a fraction of what it owns. The Schedule 13D reports no sole voting or dispositive power at all. Endeavor Blockchain holds the 500,000 shares directly and Joshua Kilgore indirectly as its managing member, and both report the same 500,000 as shared on every row, so the group’s aggregate is 500,000 and not twice that. Item 6 discloses no contract, arrangement or understanding of any kind, so the block is uncommitted. But the meeting votes the register as it stood at the 8 July record date, and the filing’s dated purchases run 22 to 30 July. All five of its purchase dates fall after the record date, so none of the 400,000 shares bought on them can be voted.

The residual settles the rest. Item 5 covers every transaction in the sixty days to 31 July and accounts for 400,000 of the 500,000, so the other 100,000 shares were acquired on or before June 1, 2026 and were held at the record date. The cover page corroborates it: 5% of 7,641,767 is 382,088, and the holding reaches 390,000, or 5.10%, on 29 July only if 100,000 were already in hand. Without them the crossing falls on 30 July rather than the 29 July the filing reports as its event date. The activist therefore votes about 100,000 shares, or 1.31% of the 7,641,767 entitled to vote, against an economic position of 5.80%. Both proposals need two-thirds of votes cast, and the contractually committed bloc of roughly 11.5% was held at the record date in full. Estimate 500,000 owned less the 400,000 the filing itemises for the sixty-day window, over the 7,641,767 record-date count.

The largest line in the company’s own ownership table is not on the register above, and it is not a share position. The definitive proxy carries Armistice Capital, LLC at 9.99%, ahead of every filed holder, through 873,643 warrants at $9.40 exercisable under a 9.99% beneficial-ownership limitation. Armistice holds no common shares, so it votes nothing on 24 August and does not belong in a table of the share register, and the percentage is the limitation’s ceiling rather than a computed stake. The warrants are the same tranche carried under capital structure above, running to January 15, 2031 and far out of the money against $2.46.

Sources: Form 424B5, July 31, 2026, cover page, Offering and Dilution; Form 8-K, July 31, 2026, Items 1.01 and 8.01; Schedule 13D, July 31, 2026; Schedules 13G, July 8 and 20, 2026; Form 4, June 10, 2026; Form 10-Q for Q1 2026, Notes 9, 10 and 11; definitive proxy statement, July 13, 2026, Appendix A, Rights of Dissent, Voting Shares and Principal Holders Thereof, and Security Ownership of Certain Beneficial Owners and Management.

08Valuation

Multiples shown with their basis and pricing date. No target price.

Earnings multiples do not apply: the company is loss-making, so price to earnings is undefined. Miners are ordinarily assessed on infrastructure scale, through enterprise value per megawatt of power and per exahash of capacity, and on net tangible book value.

The stock trades below tangible book

The prospectus supplement puts net tangible book value at $4.98 per share at 31 March, falling to a pro forma $3.11 after the merger and issuance, and $2.82 if the facility is fully drawn. At $2.46 the shares change hands at roughly 87% of pro forma as adjusted net tangible book value. New money entering through the facility receives an immediate increase of $0.47 per share in net tangible book value while existing holders take a $0.29 decrease, so the transfer runs toward the incoming buyer.

The same multiple on a combined basis

Every balance-sheet input above is pre-merger standalone, which pairs a post-merger market capitalization with a company that no longer exists in that form. The pro forma combined figures give the like-for-like reading, and the difference is not cosmetic.

On the combined basis the enterprise value is 48% higher and the per-megawatt multiple close to half as much again. The reason is entirely the liability line: Cathedra brought $6.7M of obligations, and an enterprise value is market capitalization less cash plus what is owed. The standalone reading understated it because it credited the merger’s megawatts while ignoring the merger’s liabilities.

This report treats ~$498K per MW as the better figure for comparison against peers, and retains the standalone column so a reader can see which basis any given number came from. Both are struck at the 31 July close against a 31 March balance sheet.

Two readings of the infrastructure multiple

At roughly $498K per megawatt on the combined basis, the capacity is still priced below what data-center developers ordinarily spend to build, which supports the case if that capacity is genuinely deliverable to AI or HPC tenants. Against that, the market has reasons for the discount: no power purchase agreements, a going-concern qualification, and a runway of under one quarter on the second quarter’s own burn rate.

The multiple has also re-rated quickly. At the $1.63 close of 21 July the implied figure was roughly $180K per megawatt on the same share count; at the 3 August close of $2.46 it is about $339K, an 88% expansion in nine trading sessions with no change in the underlying 53 MW. What changed was the buyer base rather than the asset.

Third-party coverage

One firm, one target, and it is fourteen months old. It was struck before the February 2026 consolidation and before the June 1, 2026 merger that created the current company, so it is a claim about a different capital structure and a different business. Against the $2.46 close it implies more than twelve times the current price. It is reported here as market data, attributed and dated, and is not adopted: this document issues no rating and derives no target of its own. No center is computed, there being one figure.

Two things about the coverage are findings in their own right. Its currency is not established: the issuer’s investor relations site did not respond when it was requested on August 3, 2026, so the company’s own statement of who follows it could not be read, and a firm that ends coverage commonly leaves its last target in aggregator data indefinitely. And the services disagree with each other, one showing $3.00, one $30.00 and one $30.60, with one of them stating a current price of $0.89 that is not the traded price. A figure taken from a page that is wrong about the price is not made reliable by being about a target instead.

Sources: market data at the August 3, 2026 close; Form 424B5, July 31, 2026, Dilution; Form 10-Q for Q1 2026 balance sheet and cash-flow statement. Multiples and runway calculated from those inputs. Third-party target and rating from aggregator coverage pages read on August 3, 2026.

09Growth drivers

The Hopkinsville expansion. The most concrete plan disclosed. Through North Campbell Land Company the company has proposed transitioning part of its existing 15 MW beyond mining, developing a new 50 MW data center, and funding construction of a new 65 MW substation, of which about 15 MW would remain available to other Hopkinsville Electric System customers. That last provision offers the utility and the community a direct benefit. No financing has been disclosed and the company held $3.1M of cash at Q1.

The Bitdeer co-mining agreements. 30 MW under contract since June 2026, adding hashrate without the company funding hardware.

Activist engagement. Endeavor Blockchain’s stated intent to engage the Board on value creation is, on the Mawson precedent, a possible route to asset sales, recapitalisation or a faster AI pivot. Whether it reads as a driver or a disruption depends on execution.

AI and HPC optionality. Management cites a modular build timeline under a year against three to five years for conventional campuses. Not contracted.

The proposed rebrand. If approved on 24 August, it formalises the AI and HPC positioning.

Fleet efficiency, already delivered. At 17.1 J/TH with the refresh substantially complete at March 2026, this lowers cost per Bitcoin today. It is not pending upside unless new hardware is deployed.

Bitcoin price. The largest external lever on revenue.

Sources: Hoptown Chronicle and Kentucky New Era, July 27 to 30, 2026; company releases of 25 and 30 June and July 10, 2026; Form 10-Q for Q1 2026, Overview and Note 12; Schedule 13D, July 31, 2026.

10Risks

Severity-ranked, most severe first.

Sources: Form 10-Q for Q1 2026, Going Concern, Note 2 Concentration Risk, Note 12 Commitments and Overview; Form 424B5, July 31, 2026; definitive proxy statement, July 13, 2026; Federal Reserve decision of July 29, 2026; local reporting, July 23 to 30, 2026.

11Bull / base / bear

Each case with its preconditions: what must be true, not what might be.

These are conditional scenarios, not forecasts, and no probability is assigned to any of them.

12Research summary

Sphere 3D is a micro-cap Bitcoin miner mid-transformation into an AI and HPC infrastructure story. The June 1, 2026 all-stock merger with Cathedra roughly doubled its footprint to 53 MW across five United States data centers at about 1.2 EH/s. Since then it has signed 30 MW of co-mining with Bitdeer, proposed renaming itself DarkHorse Technologies with a 24 August vote, and moved to position its TVA-region power for AI and high-performance compute.

The filed record is unambiguous. Revenue has fallen across 2023 to 2025, from $21.9M to $16.6M to $11.2M, and stood at $1.9M in Q1 2026 and $2.45M in Q2 2026, the first quarter carrying any Cathedra contribution. The going-concern warning is live and was reaffirmed at the second quarter, cash was $2.85M at June 30 against roughly $5.3M of burn in that quarter, and the at-the-market facility has drawn $5,131,036 across 2,172,789 shares, with a fresh $10,300,000 ceiling and a second sales agent added on 31 July. Set against that: losses that narrowed by more than half year on year, a contracted Bitdeer relationship, owned Kentucky infrastructure with a proposed 50 MW expansion, and a register that has attracted three holders above 5% in three weeks.

The ten days to 31 July changed the shape of the file. The shares rose 48% from $1.63 to $2.42 while an activist bought 400,000 shares across five sessions at $1.54 to $2.45, lifting a holding of 500,000 shares that cost roughly $988,384 in all, and the company issued into the same window. The Endeavor Blockchain filing is the most consequential new fact, not because 5.80% is control but because the same group ran a campaign at Mawson Infrastructure that ended in five new directors, a Strategic Transactions Committee, one of its participants as chief executive, and a rename to Big Digital Energy. This company had already proposed its own rename before they arrived. Whether that convergence produces cooperation or conflict is unresolved.

The methodological point worth carrying forward is that infrastructure metrics rather than earnings multiples govern here, and that at this size the register moves the price more than the operations do. On 24 July a $380,000 position added roughly $1.15M of market capitalization, about three times what was bought.

The question that resolves the file

Can the company fund itself to the point where its 53 MW is contracted to a paying tenant at better-than-mining economics, without diluting existing holders past the point where it matters? The rebrand, the domicile change and the ticker are preamble. The Bitdeer arrangement is a partial, Bitcoin-priced answer. The Hopkinsville plan is the most concrete attempt at a full one, and it is unfunded.

The 31 July filings frame the tension precisely. On one day the company enlarged its share-selling facility to $10.3M and added a second agent, and an activist disclosed a 5.80% stake with intent to engage the Board on value creation. One party is solving for survival through issuance; the other has just been diluted from 44.9% to 30.0% doing exactly that elsewhere, and has already seen its own newly filed 6.54% restate to 5.80% on the corrected share count.

13Null categories

Requested categories with nothing to report. Status reads Covered above, None where a named source was checked and found empty, or Not established where the category was not examined. A category nobody looked at is a gap, not an absence.

14Methodology & sources

Pricing basis. Equity market figures are struck at the August 3, 2026 close: $2.46, with a $1.08 to $12.60 fifty-two week range that is unchanged across the full 251-session window. Beta of 3.48 and earnings per share of $5.12 negative are provider statistics as at the 31 July close and are not restruck here, so they carry their own date. The masthead carries that close date beside the as-of date.

Provenance tags

Every material claim carries its provenance. Two categories are used in this file for the first time, and two extensions are stated here rather than assumed.

Primary sources

Form 424B5 prospectus supplement, July 31, 2026, read in full. Form 8-K, July 31, 2026, Items 1.01 and 8.01. Schedule 13D, Endeavor Blockchain LLC and Joshua Kilgore, July 31, 2026. Schedules 13G, Thomas Masiero July 8, 2026 and Jialin Qu July 20, 2026. Definitive proxy statement, July 13, 2026, read in full. Form 4, June 10, 2026. Form 10-Q for the quarter ended March 31, 2026, filed May 15, 2026, read in full. Form 10-K for FY2025, filed March 27, 2026. Merger Form 8-K filings, March 11, 2026. Company releases of 6 February, 5 March, 1, 8, 10 and 25 June, 30 June, 10 July and July 29 to 30, 2026. Mawson Infrastructure Form 8-K and release, April 23, 2026, for the comparable in section 5.

Derivations

Market capitalization is $2.46 multiplied by 8,619,150 shares. Enterprise value is market capitalization less cash and Bitcoin of $4.93M plus approximate liabilities of $1.71M, the latter taken as total assets less equity from the Q1 balance sheet. Per-megawatt and per-exahash figures divide that enterprise value by 53 MW and 1.2 EH/s. Implied runway divides cash and Bitcoin, $4.04M at June 30, 2026, by the second quarter’s operating cash use of roughly $5.3M, being the $9.02M of the six months less the $3.7M of the first, which is under one quarter. Shelf headroom applies the one-third-of-float test to float re-measured at the stated prices, less the $5,018,576 already sold. Ownership percentages divide filed share counts by 8,619,150. The FY2024 net loss is inferred from the reported 126.8% year-on-year increase.

What was not checked

A full filing sweep was run on August 11, 2026 against the complete EDGAR submissions index for this issuer, Central Index Key 0001591956, covering 807 filings from November 18, 2013 to August 10, 2026, with no supplementary index holding older material back. Exhibit lists were opened rather than form types counted. The three filings made after the previous sweep were examined: the Form 8-K and Form 8-A12B of August 10, 2026 carrying the shareholder rights plan, and the DEFA14A of August 7, 2026. The pricing basis is unchanged at the August 3, 2026 close.

Categories not examined. Investor days and industry conferences were not searched. Analyst coverage and third-party price targets were sought on August 3, 2026 and are carried in section 08; what was not established there is the currency of the one target found and the conflict its firm discloses, neither of which an aggregator page settles.

Filing families swept, with results. All eighteen form types were checked against the full index. No NT 10-K, NT 10-Q, NT 20-F or NT 40-F has ever been filed. That is now a sourced negative rather than an assumption: the entire index was searched for late-filing notifications and returns none across the company’s filing life. The foreign private issuer families are confirmed dormant rather than presumed inapplicable: 150 Form 6-K filings ending December 21, 2022, three Form 20-F filings ending March 31, 2022, and two Form 40-F filings ending March 30, 2016, which is consistent with the transition to domestic-filer status and with the historical Canadian material that formerly reached the SEC by that route. Form 3 and Form 4 were enumerated: 117 Forms 4, the most recent dated June 10, 2026. Transaction codes were not read, so the insider-purchase row keeps its status.

Federal court dockets were searched on August 2, 2026, closing a limitation this report previously named. The CourtListener RECAP index returns twelve matters naming the company as a party, all terminated, the most recent filed March 19, 2024 and closed the following month. None carries a securities nature of suit at any date: the nearest is a 2023 employee-benefits claim in the Northern District of Texas, terminated April 2024, and the two actions against Gryphon Digital Mining were brought by this company, the later closing in March 2025. State-court dockets and Canadian proceedings are outside that index and were not searched.

Counterparty filings were surveyed and disclose nothing on terms. Bitdeer, the counterparty on 30 MW, has not named this company in any SEC filing. Across all filers and forms between 1 May and August 2, 2026 the issuer’s name returns 107 documents; the third-party filings among them are institutional holdings reports, from Vanguard, Citadel, Jane Street, UBS, Wells Fargo, BNP Paribas, Danske Bank and StoneX among others, plus the Endeavor Blockchain Schedule 13D already carried. The only 2026 foreign-issuer hit is a past-directorships list in an unrelated mining company’s circular. The co-mining economics are therefore undisclosed by both sides, which is a checked result rather than an inference from this company’s filings alone.

The Canadian status is older than this report recorded. This report previously stated that the company becomes a reporting issuer in British Columbia, Alberta and Ontario on completion of the arrangement, and so acquired Canadian obligations at closing on June 1, 2026. The inference was wrong even though the sentence it rested on was not. The company’s own definitive merger proxy of April 16, 2026 settles when the status began: its arrangement-agreement representations state that Sphere “is a ‘reporting issuer’ not in default… of each of the Provinces of British Columbia, Alberta, and Ontario”, and separately that Sphere “will continue to be a reporting issuer” in those provinces. A closing document describing what a company becomes on completion is written from the transaction and can describe continuation as readily as creation; the earlier proxy is written from before it and says plainly that the status already existed. Both are accurate from where they stand, and reading the later one first inverted the history, recording an obligation of long standing as newly acquired.

The distinction is not pedantry, because it changes what the gap below means. A missing filing under an obligation two months old is an administrative lag; a missing filing under an obligation the company has carried for years, and represented in a signed agreement that it has met since January 1, 2025, is a different finding. What remains unexamined is not a newly created obligation but a long-standing one, and the unexamined period is correspondingly longer than this report previously implied.

One of those filings has now been read, and it was on EDGAR throughout. Exhibit 99.2 to the Form 8-K of May 21, 2026 is the company’s report of voting results for the 15 May special meeting, addressed to the Ontario, Alberta and British Columbia Securities Commissions and made under National Instrument 51-102 Section 11.3. It is the company’s own Canadian continuous disclosure, obtained by a route the earlier four-route search did not attempt, and being dated ten days before closing it independently corroborates the correction above. It is also where the 15 May consolidation authority in section 7 comes from. An unreachable registry is not the same as an unobtainable document.

The circular was obtained from the issuer’s own copy at cathedra.com, which the circular itself names alongside SEDAR+ as a posting location under the notice-and-access process. It is the primary document rather than secondary coverage of it, and it is a primary source about Cathedra and corroboration about this company, not a substitute for this company’s own disclosure.

Routes attempted, August 2, 2026. For Cathedra’s circular: the SEDAR+ document search and the reporting issuers list remain unqueryable, the search being a session-driven form and the list paginating with client-side script; the issuer copy succeeded. The TSX Venture Exchange bulletins were not retrieved, and the 2 June delisting continues to rest on the company releases of 26 May and June 1, 2026.

The company’s Canadian filings since June 1, 2026 were sought again and not found. SEDAR+ was retried on August 2, 2026 and the failure is now diagnosed: requests to its issuer and document pages are redirected to a bot-detection validation service rather than served, so the registry returns no answer at all. That is a registry not reached, not a registry that answered and held nothing, and the retrieval remains outstanding. The remaining routes returned as before: the investor page carries an SEC filings section and no Canadian equivalent, and the Form 6-K route by which Canadian documents formerly reached the SEC closed in December 2022. Nothing establishes what, if anything, has been filed in the three provinces since June 1, 2026, and no Canadian document appears as an exhibit to any post-closing filing. Whether a business acquisition report is owed for the Cathedra acquisition was not established either, and no claim is made in either direction. The null row stays Not established for that period.

What the retrieval did find, on EDGAR rather than SEDAR+. The substance sought from the circular’s unread appendices was already filed with the closing Form 8-K of June 3, 2026: Cathedra’s audited 2025 and 2024 consolidated statements with the reports of SRCO Professional Corporation at Exhibit 99.2, its unaudited Q1 2026 statements at Exhibit 99.3, and the unaudited pro forma condensed combined financial information at Exhibit 99.4. The exhibit list of that filing had not been enumerated, which is how a combined balance sheet sat unread on EDGAR for two months while this report described one as unavailable.

What remains unread. On Cathedra’s SEDAR+ profile: any material change reports. Its audited 2025 and 2024 statements are no longer outstanding, having been obtained from Exhibit 99.2. In the circular itself, retrieval truncated at roughly page 30 and the appendices remain unread, including the full fairness opinion and the sections concerning Sphere and the combined company. Appendix I is no longer a gap, since Exhibit 99.4 carries the same substance in the acquirer’s own presentation.

EDGAR full-text search was queried successfully on August 2, 2026, closing a gap this report previously recorded as unreachable. The earlier attempt failed against the public interface, which is a client-side application; the underlying index answers directly. Results are reported under counterparty filings above. The index reaches filing contents from 2001 onward and does not cover the paper era.

The municipal record was located but not read. The Hopkinsville zoning proceeding’s primary documents sit with Community Development Services, the joint city and county planning agency, and comprise a text amendment as introduced on June 22, 2026, a plan review text amendment carrying edits as of July 17, 2026, a separate Hopkinsville Electric System proposed amendment, and Data Centers Memo and Resolution 2026-03, together with video of the July 27, 2026 hearing. They are published as scanned images with no text layer, and no page-rendering tool was available, so their operative wording is unread and no threshold from them is carried in section 10. This is a narrower gap than before, when the documents had not been identified: they are now named and dated, and only the text is outstanding.

Grid operator data was not pulled. The Kentucky and Tennessee sites sit in the Tennessee Valley Authority footprint, which is not an organized market with a public interconnection queue of the kind ERCOT, PJM, MISO or NYISO publish, and the Iowa site is a managed hosting arrangement rather than a company interconnection. No queue position, capacity auction result or curtailment record was retrieved for any site, so the 53 MW rests on company disclosure and the circular reconciliation rather than on independent confirmation.

Press and social examined and substantially empty. Searches on August 2, 2026 across trade press, aggregator news lists, social trackers and meme-stock screens found no paid promotion, no disclosed compensation arrangement, no coordinated posting and no appearance on social-momentum screens. Coverage of the Schedule 13D and the proposed rename was located and is ordinary filing-derived reporting rather than promotion. This is the result of those searches on that date, not a general assurance. The 23 and 24 July sessions traded 5.55M and 6.51M shares against a preceding baseline nearer 150,000 to 300,000, and nothing located explains that volume beyond the activist accumulation subsequently disclosed on 31 July; the absence of a promotional finding is not an account of what moved the stock.

One social source could not be retrieved. The self-reported Alpine Fox position carried in the section 7 register is the only line in this report resting on a statement by its own subject. The originating post was not retrievable on August 2, 2026, and no SEC filing discloses the position. It is retained with Social provenance and is relied on for nothing.

Market data providers disagree materially and none was reconciled. Against the Google Finance close used here, TradingView showed a market capitalization near $14.90M and a beta of negative 0.24 where this file carries 3.48; Macrotrends showed a latest close of $1.92 dated May 28, 2026; Investing.com showed a 9 July quote. The capitalization spread is a share-count lag of the kind described in section 7. The beta conflict is unresolved and the figure should be treated as indicative only.

Figures not independently verified. The 100 MW+ expansion pipeline cited in merger materials appears in no primary document reviewed and is excluded from the valuation metrics rather than credited. Cathedra’s site leases, contract book and any escrow terms are not visible in any filing to date. Peer enterprise values were not pulled, so no peer multiple is shown.

The dated gap closed. The Q2 2026 Form 10-Q was filed on August 14, 2026, the statutory deadline, and it is the first document showing combined-entity financials and the post-merger cash position. Balance-sheet figures below are now taken from it at June 30, 2026 rather than being pre-merger standalone until it arrives.

Verification path

The following would independently confirm or overturn the central claims here, in descending order of value. The Q2 2026 Form 10-Q was read on filing, beginning with the going-concern footnote and liquidity section rather than the revenue line, since runway is the governing variable. The going-concern warning is reaffirmed in the same terms and the liquidity position is thinner than the first quarter, which is treated below as the governing fact of the quarter. Check whether any power purchase or interconnection agreement has appeared, without which the AI and HPC language remains positioning. Recompute shelf headroom as one-third of public float less trailing twelve-month sales, which governs whether the facility can be drawn at all. Track the share count on each 10-Q cover and in every prospectus supplement, since these update between quarterly reports and data providers lag them. Watch for Schedule 13D amendments from Endeavor Blockchain, particularly any move toward board nominations or a stake above 10%. Follow the Hopkinsville City Council zoning decision, which governs whether the 50 MW expansion is buildable. And confirm the outcome of the 24 August vote, noting that Board approval to proceed is discretionary even where shareholders approve.

Known limitations carried forward

One defect is closed and one remains open.

15Document log

Newest first. The original build entry is never removed or rewritten.