DGXX

$1.1 billion contracted, $1.08 million earned, and the legacy business has all but stopped

DGXX · Digi Power X Inc. · Equity Research Report · as of August 18, 2026

Price $3.97 Market
Market cap ~$405M Estimate
Enterprise value ~$263M Estimate
Shares out 101,393,355 Filed
Pricing basis Aug 14, 2026 close Market

Framing

  1. The AI data center segment produced its first revenue in Q2 2026: $1,082,592 of GPU rental. In the same quarter digital currency mining and staking fell 88% to $161,422, so the line that is arriving and the line that is going are both small. Filed 10-Q, Aug 14, 2026
  2. One contract still carries the equity story: a ten-year colocation agreement with Cerebras Systems worth roughly $1.1B for 40 MW at Columbiana, Alabama, with Phase 1 targeted for December 15, 2026. None of it is earned yet. Filed 8-K, May 8, 2026
  3. $128.1M cash and no long-term debt at June 30, 2026, bought with equity: $159.5M of shares were issued for cash in the first half alone, and the count reached 101,393,355 at August 14, 2026. Filed 10-Q, Aug 14, 2026
  4. Management targets a $250–300M annualised run-rate in fiscal 2027 against second-quarter revenue of $6.63M, which is an annualised $26.5M. Filed PR, Jul 7, 2026; 10-Q, Aug 14, 2026

01Snapshot

Key figures, each tagged by provenance. Market data is struck on the August 14, 2026 close except where a row states another date: the session, volume, beta and headcount rows at the foot of this table are still struck at the July 31, 2026 close.

MetricValueBasis
Share price$3.97Market Aug 14, 2026 close. Read from the exchange daily record over 18 sessions with no weekday absent, and agreed to the cent by a second unrelated source on all 18
52-week range$1.86 – $9.20Market Intraday across the 251 sessions to the Aug 14, 2026 close; $1.89 to $8.46 on closes. Low Mar 30, 2026, high May 13, 2026. The window sits wholly after the March 2025 ticker change, so it is a full year of this symbol’s history
Below the 52-week high−57%Estimate $3.97 against the $9.20 intraday high of May 13, 2026
Shares outstanding, subordinate voting101,393,355Filed 10-Q cover, Aug 14, 2026. The balance sheet gives 98,543,358 at June 30, 2026, so 2,849,997 shares were issued in the six weeks between
Proportionate voting shares3,333Filed 10-Q, Aug 14, 2026, unchanged since December 31, 2025
Subordinate shares issuable on their conversion666,600Filed 10-Q, Aug 14, 2026, at 200 for 1
Shares, subordinate plus proportionate as converted102,059,955Estimate 101,393,355 + 666,600
Fully diluted shares108,446,376Filed S-3, Jul 17, 2026
Market capitalization~$405MEstimate $3.97 × 102,059,955 = $405,178,021. Price at Aug 14, 2026 and count at the same date
Enterprise value~$263MEstimate see section 08
Revenue, trailing twelve months$31.7MMarket as of Jul 31, 2026
Revenue, FY2025$34.19MFiled 10-K, Mar 31, 2026
Net loss, FY2025($28.36M)Filed 10-K, Mar 31, 2026
AI data center revenue, Q2 2026$1,082,592Filed 10-Q, Aug 14, 2026, disclosed as GPU rental and appearing for the first time. It was $0 in Q1 2026
Mining and staking revenue, Q2 2026$161,422Filed 10-Q, Aug 14, 2026, against $1,394,740 a year earlier, a fall of 88%
Total revenue, Q2 2026$6,628,607Filed 10-Q, Aug 14, 2026, against $8,111,451 a year earlier
Gross loss, Q2 2026($5,524,064)Filed 10-Q, Aug 14, 2026, against $(92,081) a year earlier, on depreciation of $(4,068,734) against $(1,573,691)
Net loss to common, Q2 2026($14,360,509)Filed 10-Q, Aug 14, 2026. $(0.17) per share on 85,480,992 weighted average shares
Cash$128.1MFiled 10-Q, Aug 14, 2026, at June 30, 2026, against $78.5M at December 31, 2025
Long-term debt$0Filed 10-Q, Aug 14, 2026
Digital assets$14.30MFiled 10-Q, Aug 14, 2026, at June 30, 2026
Property, plant and equipment, net$73.24MFiled 10-Q, Aug 14, 2026, against $23.01M at December 31, 2025, the Columbiana build
Secured power across all sites~400 MWFiled company disclosure
Prior close, Jul 30, 2026$3.80Market Jul 30, 2026
Session change, Jul 31, 2026−3.16%Market $3.68 against a $3.80 prior close
After-hours, Jul 31, 2026$3.61Market Jul 31, 2026
Session volume, Jul 31, 20261.90MMarket against a 4.75M average
Average volume4.75MMarket Jul 31, 2026
Employees17Market Jul 31, 2026
Beta, five year6.10Market Jul 31, 2026
52 week range$1.86–$9.20Market Aug 14, 2026, unchanged from the July 31 window: 251 sessions on either

The figures describe a business being dismantled and a business being built, at the same time. Current financials measure the first; the valuation prices the second. That gap is the subject of this report.

02Business model

Digi Power X owns electrical power capacity and the buildings on top of it, and rents that capacity to operators of compute. It has done so in three successive forms: as Digihost Technology, a self-mining Bitcoin operator from 2017; then as a host for third party miners; and now as a landlord to AI tenants at higher rents per megawatt. Filed

The company renamed to Digi Power X in March 2025 and moved its Nasdaq ticker from DGHI to DGXX on March 18, 2025. It is incorporated in British Columbia, headquartered in Miami, and became a U.S. domestic filer effective January 1, 2026, converting its accounts from IFRS to U.S. GAAP. Filed 10-Q, May 15, 2026

Why power is the asset

The binding constraint on AI capacity is not land or shell but interconnection: a signed agreement with a utility, which takes years to obtain. Operators holding power from the mining era hold the scarce input, which is why the cohort is converting rather than building from nothing.

DGXX holds power assets in Alabama, New York and North Carolina, including a combined cycle plant at North Tonawanda, New York. Across all sites it reports roughly 400 MW of secured power, the strategic asset base even though only a fraction is contracted. The company’s own presentation describes that portfolio as potential rather than secured. Page 7 of the investor presentation published at the company’s investor page and read on August 20, 2026 states a portfolio capable of supporting approximately 409 MW of future AI infrastructure capacity, and page 11 places a potential 409 MW portfolio expansion in a Future Vision column rather than under 2026 or 2027. Estimate Power that is secured and capacity a portfolio could support are not the same claim, and the figure here is carried as filed company disclosure. The two are not reconciled here. The presentation cannot be dated, so it cannot be said either to supersede the filed disclosure or to precede it, and the tag is left as the filing put it rather than changed on the strength of an undated page. At the Columbiana, Alabama campus it reports a dedicated on-site substation complete and a power delivery agreement with Alabama Power secured. Filed

The July 2026 shelf registration describes Columbiana as a 55 MW Tier 3 build, while the Cerebras agreement covers roughly 40 MW. Filed S-3, Jul 17, 2026. That leaves as much as 15 MW at the flagship site under construction and uncommitted: capacity that could host a second tenant without a new site, or sit idle. The company has not addressed the gap. Its own presentation puts a third number on the same site. Page 10 of the investor presentation published at the company’s investor page and read on August 20, 2026 describes the company as currently converting a wholly-owned 70 MW power infrastructure into a Tier III data center in buildout phases. That presentation prints no date on any of its 21 pages, and it is published at a single address that is overwritten in place when the next version ships, with no archive and no version history. It is cited here by the date it was read, never by a date it carries, and a reader following the same address later may be served a different document. Estimate That widens the gap rather than closing it. The filed description is 55 MW, the contract covers 40 MW, and the presentation says 70 MW, and because the presentation carries no date of its own it cannot even be placed in sequence with the July 2026 shelf. The sentence above therefore stands: the gap has not been addressed in any document that can be dated.

Two routes to revenue

DGXX also holds a stake in US Data Centers Inc., which manufactures ARMS, a factory-built modular pod deployable in months rather than years, and houses a URP-1 robotics line. An ARMS 200 unit has run at Tier 3 standards at the Alabama site since May 15, 2026. Filed

Analyst assumption

Ownership of US Data Centers reads inconsistently across documents: 2025 releases described a wholly owned subsidiary, the Q1 10-Q states 51% at March 31, 2026 after a 49% sale for $1.905M, and later material says ~55%. A Form 8-K of March 16, 2026 sets out the structure: a 55% majority stake for DGXX; roughly 35% founder equity held by the co-founding management team including Hans Vestberg, former chief executive of Verizon; a scope limited to manufacturing and distributing ARMS with no site-level revenue participation; all pods, GPUs and site revenue belonging entirely to DGXX; and any equipment sold to DGXX at cost. Filed 8-K, Mar 16, 2026

My reading is that the structure is disclosed and shareholder-protective, and that the residual objection is to the messaging rather than the substance. Describing an entity as wholly owned and later as a diluted majority is drift that costs trust, and the 7 July release reverting to "major shareholder" did not help.

03Revenue streams

Four reported segments. Three earn today; the fourth, which carries the valuation, earns nothing yet.

SegmentQ1 2026Q1 2025ChangeBasis
Sale of energy$3,716,711$3,427,916+8%Filed 10-Q, May 15, 2026
Colocation services$3,026,908$5,082,795−40%Filed 10-Q, May 15, 2026
Digital currency mining and staking$47,727$765,876−94%Filed 10-Q, May 15, 2026
Tier III AI data centers$0n/an/aFiled 10-Q, May 15, 2026
Total revenue$6,791,346$9,276,587−27%Filed 10-Q, May 15, 2026

Energy is now the largest line and the steadiest, but it lost money at segment level in Q1 2026: segment EBITDA of −$1.68M on cost of revenue of $5.39M against $3.72M of revenue. Filed Selling power below cost is not durable, and management frames it as transitional use of capacity.

Colocation fell 40% because DGXX is emptying a hosting site to rebuild it for AI. Filed This is planned demolition rather than customer loss, but reported revenue deteriorates before it recovers.

Mining is effectively switched off, from $765,876 to $47,727 in a year. The company holds 166 Bitcoin and 1,013 Ethereum, carried at $13.56M against a $17.33M cost basis at March 31, 2026. Filed Holdings revalue through the income statement, so a $3.76M revaluation loss flowed through Q1 2026 results.

Contracted revenue not yet earning

Analyst assumption

Contract unit economics: $1.1B over ten years across 40 MW is roughly $2.75M per MW per year, or about $229,000 per MW per month. Estimate That sits inside the band observed in recent large high density colocation deals, broadly $150k to $250k per MW per month, which is a mild check in the contract's favor rather than proof of it. Revenue is not evenly spread: Phase 1 is 15 MW, so the first full year earns well below the $110M ten year average.

04Industry & market backdrop

Demand for AI compute has outrun the capacity to house it, and the constraint has moved from accelerators to electricity and interconnection queues. Grid operators in several U.S. regions have warned publicly that data center load forecasts are straining planning assumptions. Filed Scarcity of that kind is what gives existing power capacity option value.

Committed capital is large. CoreWeave reported roughly $2.08B of revenue in its May 2026 quarter against a stated backlog near $99.4B. Market Hyperscalers have signed multi-billion dollar capacity contracts with third party providers partly because leasing is expensed over time rather than capitalized on their own balance sheets.

Three currents running against DGXX

Scale consolidation. Contracts are increasingly awarded in gigawatts. Applied Digital said on May 20, 2026 that it had passed 1 GW of contracted capacity. Market The DGXX flagship is 40 MW, roughly 4% of that, and sub-scale operators can be excluded from the largest tenders irrespective of execution quality.

Buyers becoming sellers. On July 1, 2026, reporting that Meta was exploring sales of excess AI capacity under a "Meta Compute" banner took roughly 15% off Nebius and CoreWeave in a single session. Market If the largest buyers of capacity become sellers, landlord pricing power erodes.

Financing as the moat. The differentiator is increasingly the ability to raise project debt cheaply against signed leases. Larger operators can; microcaps often fund with equity, diluting the holders the capital is meant to serve.

That constraint tightened in July 2026. Reporting on 29 July flagged rising credit default swap costs across AI infrastructure, with CoreWeave 2032 bonds yielding 10.32% at a B rating from S&P. Market CoreWeave carries over $25B of debt, Nebius $8.5B, IREN $4B. If lenders are repricing the sector's largest and most established borrowers, a microcap with no debt track record faces a harder market still. Estimate

Cohort equity performance reflects it: neocloud names fell 25–38% in the month to mid July 2026. Market IREN fell 37% despite a five year, $3.4B NVIDIA contract; Applied Digital slid despite a 61% revenue beat. Good news ceasing to be rewarded is the clearest available signal that the sector is being repriced on financing risk rather than on execution.

05Competitive position

Named peers on contracted capacity and anchor tenant, as reported through mid 2026.

CompanyAnchor tenantRelative scaleBasis
Applied Digital (APLD)CoreWeaveOver 1 GW contractedMarket May 20, 2026
IRENMicrosoftMulti-GW power portfolioMarket mid 2026
Cipher Mining (CIFR)AWS, FluidstackMulti-site, GW scaleMarket mid 2026
TeraWulf (WULF)Fluidstack, Core42Several hundred MWMarket mid 2026
Hut 8 (HUT)Anthropic, FluidstackSeveral hundred MWMarket mid 2026
Core Scientific (CORZ)CoreWeaveGW scaleMarket mid 2026
Digi Power X (DGXX)Cerebras40 MW contractedFiled 8-K, May 8, 2026

DGXX is roughly an order of magnitude smaller than most comparables on contracted capacity. Estimate Two readings follow and both hold. The bear reading is that it is sub-scale in a business where scale determines who is invited to bid. The bull reading is that it is small enough for a single 40 MW contract to be transformational, where the same contract at IREN would round to nothing.

The nearest comparables by size are Soluna Holdings (SLNH), WhiteFiber (WYFI) and New Era Energy & Digital (NUAI): the tier where execution risk dominates and valuation moves on contract announcements rather than earnings.

Counterparty quality

Counterparty quality matters more than contract size, and here it is better than the market capitalization would suggest. Cerebras Systems listed on Nasdaq as CBRS on May 14, 2026, raising roughly $5.5B, priced at $185 after an upsized offering, opening at $385 and closing its first session near $311. Market It reported 2025 revenue of about $510M and net income near $237.8M, and holds a large multi-year compute agreement with OpenAI. Market

Analyst assumption

A newly capitalized, profitable, listed counterparty is more creditworthy than a private startup, which removes one common microcap failure mode. Estimate Two qualifications. The DGXX contract was signed on 5 May, nine days before the Cerebras listing. And Cerebras carries its own concentration: a large share of revenue traces to few customers, with OpenAI dominant, so DGXX revenue is exposed to OpenAI spending decisions two steps removed.

06Financial performance

U.S. GAAP. FY2025 from the annual report; quarterly figures from the Q1 filing.

MetricQ1 2026Q1 2025FY2025Basis
Revenue$6.79M$9.28M$34.19MFiled 10-Q, May 15, 2026; 10-K, Mar 31, 2026
Gross profit / (loss)($0.80M)($1.52M)n/aFiled 10-Q, May 15, 2026
G&A expense$4.33M$2.71Mn/aFiled 10-Q, May 15, 2026
Net loss($4.65M)($1.63M)($28.36M)Filed 10-Q, May 15, 2026; 10-K, Mar 31, 2026
Loss per share($0.07)($0.05)($0.64)Filed 10-Q, May 15, 2026; 10-K, Mar 31, 2026
Operating cash flow($6.40M)($10.11M)n/aFiled 10-Q, May 15, 2026
Capital expenditure$15.17M$0.78Mn/aFiled 10-Q, May 15, 2026
Cash$57.81M$0.80M$78.48MFiled 10-Q, May 15, 2026; 10-K, Mar 31, 2026
Digital assets$13.56Mn/a$14.81MFiled 10-Q, May 15, 2026; 10-K, Mar 31, 2026
Long-term debt$0n/a$0Filed 10-Q, May 15, 2026; 10-K, Mar 31, 2026

What the figures show

The operating business does not cover its own cost of revenue. Gross profit is negative in both quarters shown, before any overhead. That is not unusual mid-conversion, but the operating business currently destroys value at the gross margin line.

Overhead is outgrowing revenue. G&A rose 60% year on year while revenue fell 27%. Share-based compensation was $1.35M and key management remuneration $2.42M for the quarter, against $6.79M of total revenue. Filed Roughly 36 cents of every revenue dollar went to management compensation in Q1 2026. Estimate

Amendment No. 1 to the FY2025 annual report, filed April 27, 2026, supplies the Part III disclosure omitted from the original filing. It states chief executive Michel Amar received total 2025 compensation of $12,025,227, roughly 8% of it salary and the remainder bonus, share-based and option-based awards. Filed 10-K/A, Apr 27, 2026

The same amendment discloses president Alec Amar's 2025 total compensation of $7,425,695, so the two together drew $19,450,922. Filed 10-K/A, Apr 27, 2026 It also records directors and executive officers as a group holding 8,879,464 subordinate voting shares, being 11.77% of that class, plus all 3,333 proportionate voting shares, together about 12.55% of total voting power at April 27, 2026. Filed 10-K/A, Apr 27, 2026

Against FY2025 revenue of $34.19M and a net loss of $28.36M, the chief executive alone drew roughly 35% of total annual revenue, and the two named officers together drew 57%. Estimate $19,450,922 / $34.19M. On that comparison one executive's compensation equals roughly 35% of total annual revenue at a company losing money at the gross profit line. Estimate Equity-weighted packages are ordinary in early stage infrastructure and preserve cash, but a package of this size at this revenue scale is an outlier, and it sits in an amendment rather than the original annual report.

Reported earnings are distorted by non-operating items. The Q1 net loss of $4.65M contains a $3.76M crypto revaluation loss, a $2.96M foreign exchange gain and a $0.78M warrant fair value gain. Filed These broadly offset, so the headline loss says little about operations.

Two disclosure quality flags

Adjusted against reported EBITDA. The Q1 press release leads with Adjusted EBITDA of +$1.1M. The segment note inside the same quarterly report shows total EBITDA of −$3.20M. Filed Both can be correct under their own definitions, but a headline positive figure alongside a filed negative one is reason to read the reconciliation directly.

Two filings state different digital asset holdings at the same date. The annual report says that at March 31, 2026 the company held approximately 51 bitcoin and 1,010 Ethereum in its custodian account. Filed 10-K, Mar 31, 2026 The quarterly report covering the same date carries 166 Bitcoin and 1,013 Ethereum at a fair value of $13.56M. Filed 10-Q, May 15, 2026 The annual report qualifies its figure as the balance at one named custodian, which may account for the gap, but the two are not reconciled anywhere in either document and this report carries the quarterly figure throughout.

A misstatement in the discussion and analysis. The Q1 2026 narrative describes the prior year comparative as "net income of $1,633,261". The statements in the same document show that figure as a net loss. Filed Small in dollars, but an error in a filed document speaks to internal control.

07Capital structure & dilution

Shares, the at-the-market program, and the instrument overhang. Dilution is the central cost of the strategy and is treated here rather than as one risk among many.

DateSubordinate voting sharesChange vs Dec 2024Basis
Dec 31, 202433,011,600n/aFiled 10-K, Mar 31, 2026
Mar 31, 202536,307,870+10%Filed 10-K, Mar 31, 2026
Dec 31, 202569,427,788+110%Filed 10-K, Mar 31, 2026
Mar 31, 202669,807,452+111%Filed 10-Q, May 15, 2026
May 15, 202690,420,824+174%Filed 10-Q, May 15, 2026
Jul 2, 202698,543,355+199%Filed S-3, Jul 17, 2026

After March 31, 2026 the company issued 19,950,000 shares for $102.86M under an at-the-market program upsized to $175M on May 8, 2026, an average of roughly $5.16 per share and above the current price. Filed 10-Q, May 15, 2026

A further 8,122,531 shares were issued between 15 May and 2 July, roughly 9% dilution in seven weeks, disclosed on the cover of a resale registration rather than announced. Filed S-3, Jul 17, 2026. At prevailing prices that implies roughly $32–41M of further drawdown. Estimate

Analyst assumption

Remaining program capacity cannot be settled from the filings. The Q1 quarterly report records 19,950,000 shares for $102.86M after March 31, 2026. The prospectus supplement of May 8, 2026 states the first supplement covered $75M, of which $72,363,650 had been sold before that date. Filed 424B5, May 8, 2026

The two likely span different programs or periods, the company having run both a Canadian and a U.S. shelf, but on their face they imply materially different remaining capacity: anywhere from roughly $37M to $70M undrawn under the $175M program. Estimate I state the band rather than a point because the evidence does not support a point. The Q2 quarterly report should settle it.

Also disclosed: A.G.P. / Alliance Global Partners earns a 3.0% commission on program sales, and the shares last traded at $6.19 on May 7, 2026, the day before the upsizing. Filed 424B5, May 8, 2026 That is roughly 56% above the 14 August close of $3.97, so the buyers into that program are carrying a loss of about that size. Estimate $6.19 / $3.97 − 1

Instrument overhang

InstrumentSharesTermsBasis
Options3,485,000Weighted average exercise price C$4.90Filed S-3, Jul 17, 2026
Unvested RSUs4,670,946No strikeFiled S-3, Jul 17, 2026
Warrants1,080,475$2.98 average strikeFiled S-3, Jul 17, 2026
Total overhang9,236,4219.1% of shares outstandingEstimate 9,236,421 / 101,393,355
Fully diluted shares108,446,376As stated by the companyFiled S-3, Jul 17, 2026

The RSU pool grew 58% in three months while the option strike moved above the share price. A separate grant around the Q1 report added 650,000 options at C$9.84 and 1,730,000 RSUs vesting in three tranches from 2027 to 2029. Filed PR, May 15, 2026. That strike sits roughly 2.5 times above the current price, so those options are far out of the money, while the RSUs dilute irrespective of where the stock trades.

Who holds the stock

The beneficial-ownership record has three filers since 2025 and they describe three different kinds of holder. Every percentage below is quoted by its filer against the class as it then stood, and each is restated here against the 98,543,355 shares outstanding at July 2, 2026, because the class has grown faster than the filings. The class has grown again since: 101,393,355 at August 14, 2026, so every percentage below overstates the current holding by about 3%.

HolderSharesAs filedOn the current classBasis
Michel Amar, Bit.Management LLC, NYAM LLC, Bit Mining International LLC8,047,08111.5%~8.2%Filed Schedule 13G/A No. 4, Feb 13, 2026, at Dec 31, 2025. Includes 666,600 on conversion and 1,415,000 vested options
Citadel Advisors LLC and affiliates, with Kenneth Griffin3,550,6914.9%~3.6%Filed Schedule 13G, May 12, 2026, at May 5, 2026, passive under Exchange Act Rule 13d-1(c)
Eleven Ventures LLC, Eleven Managers LLC, Hartley Waskoniln/an/aFiled Schedule 13G/A No. 1, Aug 14, 2025: 1,053,536 shares at Jun 30, 2025, reduced to nil on July 7, 2025

Citadel is the largest outside holder and arrived recently. Its position more than doubled from 1,517,083 shares, 3.4%, disclosed in October 2025 to 3,550,691, 4.9%, in May 2026. The figure is worth reading precisely: 4.9% is the last percentage that carries no further obligation, since crossing 5% brings a holder inside the reporting regime that governs Mr Amar. Within the group the shares sit in three pockets, 2,391,705 with the advisory entities, 740,400 with Citadel Securities LLC and 1,158,986 with the Citadel Securities partnerships, which is the shape of a market-making and multi-strategy book rather than a strategic stake. Both filings are passive, and nothing in either states an intention toward the company.

The exit is as informative as the entries. Eleven Ventures took 3,113,636 shares and warrants over an equal number in August 2024, exercised the warrants in tranches, and sold everything: by July 7, 2025 it held none. A financing counterparty that has fully unwound is no longer an overhang, and that is the favorable reading. The unfavourable one is that roughly 6.2 million shares were absorbed by the market inside eleven months, which is part of why the class grew as it did.

Analyst assumption

The restated percentages are this analyst's arithmetic, dividing each filed share count by the 98,543,355 shares outstanding at July 2, 2026. No filer has restated its own figure, and a holder may have traded since its event date, so these show the effect of the growing class rather than a current position. A Schedule 13G reports only holdings above the threshold, so the register below 5% is not visible and no total for insider or institutional ownership can be built from it.

The cash position is accurate and valuable, and it is also shareholders' own capital returned to them as a balance sheet: raised by issuing stock, not generated by the business. Whether that was a good exchange depends entirely on whether the Alabama campus earns a return above the cost of the dilution.

08Valuation

Multiples with their basis and pricing date. There are no earnings to capitalize and the asset base is mid construction, so what follows is arithmetic. No target price is issued.

InputValueBasis
Shares, subordinate plus proportionate as converted102,059,955Filed 10-Q, Aug 14, 2026, being 101,393,355 subordinate voting plus 666,600 issuable on conversion of the proportionate voting shares
Share price$3.97Market Aug 14, 2026 close. Read from the exchange daily record over 18 sessions with no weekday absent, and agreed to the cent by a second unrelated source on all 18
Market capitalization~$405MEstimate $3.97 × 102,059,955 = $405,178,021. Price at Aug 14, 2026 and count at the same date
Less cash($155M)Filed PR, Jul 7, 2026
Less digital assets($13.6M)Filed 10-Q, Mar 31, 2026 figure
Plus long-term debt$0Filed PR, Jul 7, 2026
Enterprise value~$263MEstimate $405M − $128.1M cash at Jun 30, 2026 − $14.3M digital assets
Revenue basisAmountEV / salesBasis
Trailing twelve months, actual$31.7M7.0×Estimate $223M / $31.7M
FY2027 target, low end$250M0.89×Estimate $223M / $250M
FY2027 target, high end$300M0.74×Estimate $223M / $300M

This table is the debate in full. Estimate On current revenue DGXX trades at roughly 6.2 times sales, expensive for a business with negative gross margin. On the company's fiscal 2027 target it trades below 1 times sales, cheap against AI infrastructure peers trading at high single digit to low double digit multiples of forward sales. The market is pricing something between the two, which is a reasonable response to the uncertainty. A 52 week range of $1.86 to $9.20 and a five year beta near 6.1 show how violently that middle position reprices. Market

Analyst assumption

Building 40 MW of Tier III liquid cooled colocation costs in the region of $7–12M per MW for powered shell, electrical plant and cooling, implying roughly $280–480M for the full campus. Estimate This band is mine, not the company's. The company has published a valuation of the same 40 MW, and every input in it is the company's own. Page 10 of the investor presentation published at the company’s investor page and read on August 20, 2026 multiplies an anticipated EBITDA per megawatt of about $1M by a comparable asset multiple of 12.5 times by 40 MW developed to reach an asset valuation of $500M. Estimate Recorded, not used. The multiple is one the issuer selected and applied to itself, which makes the output an estimate however official the document carrying it. And it is a valuation, not a budget, so it does not discharge the open item below: no project cost has been disclosed and the band stays mine. DGXX has deployed roughly $95M year to date and held $128.1M of cash at June 30, 2026, which is plausibly enough to finish Phase 1 at 15 MW but not Phase 2 without outside capital. That is consistent with the agreement conditioning Phase 2 on financing, and with the 7 July release pointing to project level debt once documentation is finalised.

Treat the band as a sanity check rather than a figure. The company has not disclosed a total project budget, and the absence is itself worth noting.

Third-party coverage

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. Two of these actions carry a date the day after the market was shut: September 2, 2025 follows the Labor Day closure and the weekend before it, and November 17, 2025 follows a weekend.

FirmAnalystDateRatingTargetPrice at publicationBasis and disclosed conflict
Weiss Ratingsn/aJul 17, 2026Sell, D−, reiteratednone$3.72Market The close on July 16, 2026, the last session before the action. A quantitative ratings service rather than broker research: model-derived, no target, no named analyst. It is the most recent action on this issuer and the only negative one.
Alliance Global PartnersBrian KinstlingerMay 15, 2026Buy$12, raised from $10$7.22Market The close on May 14, 2026, the last session before the action. Raised alongside the first quarter results, citing time-to-power as the sector bottleneck. The target was struck against a close about 83% above the current price. Disclosure about this issuer not established.
Alliance Global PartnersBrian KinstlingerMay 6, 2026Buy$10, raised from $7$5.11Market The close on May 5, 2026, the last session before the action. Raised immediately after the Cerebras announcement, and the shares rose 22% on the day the note landed, so the session before it is the one the target was struck against. Superseded nine days later by the row above, and kept because the pair shows how fast the target moved. Disclosure about this issuer not established.
H.C. WainwrightKevin DedeNov 17, 2025Buy$5.00$3.21Market The close on November 14, 2025, the last session before the action, which fell on a Monday. Not revisited through any of the events since, so its currency is not established. Disclosure about this issuer not established.
iA FinancialNot establishedSep 2, 2025Not established$3.50$2.20Market The close on August 29, 2025, the last session before the action: September 1, 2025 was a market holiday, so the preceding Friday supplies the close. Reported by the service, which records a target without a rating. Eleven months old and predating the Cerebras agreement and the name change, so its currency is not established. Disclosure about this issuer not established.

Coverage runs to four firms and one ratings service, not two. The three published targets that carry a number are $12, $5.00 and $3.50, struck across eleven months on entirely different facts: the $12 postdates the Cerebras agreement, the $5.00 predates it, and the $3.50 predates the change of name. An average of them would describe nothing, and none is struck here. The current price of $3.97 sits above one of the three, $3.50, and well below the most recent.

The most recent action is the negative one, and it is not broker research: a quantitative service reiterated a Sell grade on July 17, 2026. It carries no target and names no analyst, and is reported for what it is rather than counted alongside the analysts.

A service separately publishes a consensus of $4.00 over three analysts with two Buy and one Sell. That is reported as a figure that service publishes, not as a center struck in this document. This document adopts none of these targets and issues none of its own.

What is not established. No note was read, so what any of these firms discloses about this issuer is unknown. The analyst at iA Financial is not named by the service, and no rating accompanies its target. Every target here is cited to the service reporting it rather than to the firm.

09Growth drivers

The path from $31.7M to the fiscal 2027 target is not gradual. It steps as megawatts energise, so the schedule is the growth case.

StageCapacityTimingStatus and basis
Live today~0 MW AIJul 2026One ARMS 200 pod live since May 15, 2026 plus the NeoCloudz B200/B300 fleet, producing first AI revenues at an undisclosed amount. Filed
Columbiana Phase 115 MWDec 15, 2026Ready-for-service target. Self funded, substation complete, long lead electrical and switchgear secured at July 7, 2026, shell under vertical construction. Filed 8-K, May 8, 2026
Columbiana Phase 240 MWEnd Q1 FY2027Full Cerebras deployment. The additional 25 MW is conditioned on securing adequate financing, the single most important conditional in the case. Filed 8-K, May 8, 2026
Aggregate colocation ambition90 MWFY202750 MW beyond the Cerebras agreement for up to ~$200M of colocation revenue. No tenant announced. Roughly 15 MW of it may sit at Columbiana, described as a 55 MW build against a 40 MW contract. Open

Management has also stated targets beyond fiscal 2027, disclosed on the Q1 call: an annual run-rate of $450–500M in 2028 and $800M–$1B in 2029, contingent on financing and execution. Filed These are recorded for completeness. They sit several dependency layers beyond a company whose AI segment reported its first $1,082,592 in the second quarter of 2026, and I attach little weight to them. Estimate

Secondary drivers

Of the $250–300M fiscal 2027 target, only the Cerebras portion, guided at $80–100M, rests on a signed contract, and that portion still needs Phase 2 financing. The remainder, up to ~$100M from additional colocation, up to ~$100M from GPU services and ~$12M from energy, depends on tenants not yet announced. Estimate Roughly two thirds of the target is uncontracted. The company states a materially higher ceiling for the GPU business, on a different basis. Page 15 of the investor presentation published at the company’s investor page and read on August 20, 2026 gives roughly 4,000 Blackwell GPUs of hosted capacity against up to about $120M of potential annualised revenue, repeated on page 13. Estimate The bases differ and that is the whole of it: the $120M is an undated ceiling for a fleet of that size, while the figure above is a dated fiscal 2027 run-rate target. Neither document has carried any GPU unit count until now, and 4,000 is the first the company has put on it.

10Risks

Severity ranked, most severe first.

Customer concentration Severe

One customer represents effectively all contracted future colocation revenue, with no diversified base to absorb a reduction, delay or exit. Cerebras in turn depends heavily on OpenAI, so the chain runs from DGXX through Cerebras to OpenAI, with no visibility and no contractual protection at the far end. Cerebras guided to declining gross margin in its first post-listing results on June 23, 2026 and its shares fell on the news. Market

Phase 2 financing Severe

Phase 2 is conditional on capital not yet raised. Without project debt on acceptable terms the alternatives are equity at a depressed price or a scaled back deployment. Filed The risk has grown since the Q1 filings: with AI infrastructure credit spreads widening through July 2026, terms available to a first time sub-scale borrower are likely worse than when the intent was signalled on 7 July. Estimate At the Q1 combined pace of operating burn plus capital expenditure, roughly $21.6M per quarter, $155M funds about seven quarters, but capital expenditure accelerates during vertical construction so that figure flatters the true runway. Estimate

Execution and schedule Severe

December 15, 2026 is a hard, publicly stated date. Data center construction slips on transformer lead times, switchgear delivery, commissioning faults and labour. Long lead items are secured, which reduces the risk without eliminating commissioning risk. The agreement contains service credit, penalty and termination provisions, so delay may carry direct financial consequence. Filed 8-K, May 8, 2026

Dilution Severe

The share count roughly tripled in nineteen months to 98,543,355, and 9.4% of shares sit in options, RSUs and warrants. Revenue growth per share will lag revenue growth in aggregate. The mechanism deserves emphasis: the 8.1M shares issued between 15 May and 2 July were disclosed in a resale registration filed for an unrelated purpose rather than announced. Filed Program sales require no contemporaneous disclosure, so an investor tracking releases and quarterly filings alone will understate the share count between reporting dates.

Governance, insider selling and promotional communications Moderate

Insider transaction filings show president and director Alec Amar executing six transactions over roughly eighteen months, every one a sale and none a purchase, a net disposal of about 137,500 subordinate voting shares. Individual sales include 27,500 shares on January 2, 2026 at $2.72, 1,800 on 1 April at $2.25, 25,700 on 8 April at $2.26 and 27,500 on 1 May at $3.56. Intent to sell notices covering a further 83,000 shares have been lodged. Across the last twelve months insiders as a group were net sellers. Filed Forms 4, Jan to May 2026

This cuts against the usual defense of a founder-led microcap, that interests are aligned through ownership. Estimate Two qualifications in fairness: the amounts are modest against Alec Amar's remaining ~1.37M direct shares, and routine personal sales evidence nothing improper. But there is no offsetting purchase at any point, including through the drawdown from the $9.20 high. Michel Amar's own holding is reported on two bases, and the difference between them is arithmetic rather than movement. The 8,047,081 shares recorded at December 2025, 11.5% of the class then, comprise 5,338,937 subordinate voting shares held directly and through Bit.Management LLC and NYAM LLC, 626,544 held through Bit Mining International LLC, 666,600 issuable on conversion of 3,333 proportionate voting shares, and 1,415,000 issuable on fully vested options. Remove the last two and the figure is 5,965,481, which is the roughly 5.97M reported at May 2026. The gap of 2,081,600 is exactly those two components, so the share count did not fall and nothing was sold. The percentage did fall, from 11.5% to 6.6%, and that part is real, but it is a denominator effect: the class grew from 70,094,385 to 90,420,824 while the holding stood still. Estimate

Two things follow for the reader. Alec Amar's steady selling and Michel Amar's static holding are different facts and should not be read as one trend. And on the 98,543,355 shares outstanding at July 2, 2026, the December basis restates to about 8.2% and the direct basis to about 6.1%, so every ownership percentage in the beneficial-ownership record is quoted against a smaller class than exists today. Estimate

The company also issues releases at a high cadence, several announcing intentions, letters of intent or conference attendance rather than binding commitments. The chief executive's background is chiefly in fashion and licensing rather than power engineering, which cuts both ways given he has assembled a technical team and secured a nine figure contract. The July 7, 2026 operational update was met with a share price decline of roughly 7%, which suggests the market has become inured to announcement flow. Market

An unverified association Moderate

Promotional material around this stock, including a podcast and third party write-ups, has associated DGXX with Peter Lynch, the former Fidelity Magellan manager, as a shareholder or backer. At least one critical commentator argues this refers to a different individual of the same name. I could not verify any such holding in the ownership filings reviewed, which relate to Michel Amar and affiliated entities at 8,047,081 shares, 11.5% of the class, as of December 31, 2025. Filed 13G/A. Treat the association as unverified absent a filed ownership document. That a case is marketed on a celebrity endorsement is itself information about the shareholder base.

Interest rate direction has reversed Moderate

At the July 29, 2026 meeting the Federal Open Market Committee held at 3.50–3.75% on a 9–3 vote, with all three dissents arguing to raise. Filed Federal Open Market Committee statement, Jul 29, 2026 The June dot plot points to one quarter point increase by end 2026 and futures price two. This bites twice: a capital intensive builder seeking first time project debt faces a higher cost of capital, and a high beta pre-revenue equity faces multiple compression. Estimate Any model of this company built on falling rates needs revisiting.

Sector sentiment Moderate

A broad AI infrastructure reversal would hit this name harder than larger peers, and execution would not protect it. The cohort fell 25–38% in the month to mid July 2026 with contract wins and earnings beats failing to arrest the move. Market

Residual crypto exposure Low

$13.56M of Bitcoin and Ethereum revalue through earnings each quarter, injecting volatility unrelated to the AI business, and the holdings are carried below cost at $13.56M fair value against a $17.33M cost basis at March 31, 2026. Filed 10-Q, May 15, 2026

Structural and listing Low

Dual U.S. and Canadian listing and regulatory regimes; a dual class structure in which proportionate voting shares carry 200 votes each; a recent conversion from IFRS to U.S. GAAP whose first full audited year is still ahead. Filed 10-Q, May 15, 2026

11Bull / base / bear

Conditional scenarios with their preconditions, carrying no probabilities. They state what must be true, not what might be.

Bear

One of three failures. Construction slips past December 2026, triggering service credits or penalties and delaying all revenue. Project financing fails to close, a risk raised by AI infrastructure credit spreads widening through July 2026, forcing Phase 2 into equity at a depressed price or abandonment. Or Cerebras, facing its own margin pressure and OpenAI concentration, renegotiates, delays or reduces. Meanwhile the legacy business keeps shrinking, cash depletes into construction, and the program is used at falling prices. The stock retests the lower part of its 52 week range. The mitigant is real: with no long-term debt, no lender can force an outcome, so the realistic bear case is severe dilution and a long wait rather than insolvency. Estimate

Requires: a schedule slip, a financing failure, or a counterparty change.

Base

Phase 1 arrives on time or slips a quarter. Cerebras revenue begins in 2027 and Phase 2 proceeds on a mix of debt and further equity, diluting existing holders. The additional 50 MW and the full $100M GPU run-rate do not arrive on the stated timeline. Fiscal 2027 revenue lands below the $250M floor, perhaps in the $100–160M region, still a step change on $31.7M. Estimate The company becomes a small operating AI colocation business rather than a story, and valuation settles between today's trailing multiple and the peer forward multiple, with the outcome turning on how much dilution occurred along the way.

Requires: Phase 1 delivering roughly on time.

Bull

Phase 1 energises on schedule in December 2026 and Cerebras revenue begins. Project debt closes on reasonable terms, funding Phase 2 without further equity. Delivery becomes a reference for the next tenant and DGXX signs part of the additional 50 MW. NeoCloudz scales on the Vera Rubin and B300 fleet. Revenue approaches the lower half of the $250–300M target, at which the current ~$263M enterprise value is under 1.1 times sales for a contracted operator against peers at multiples of that. Estimate The re-rating would be large, though smaller per share than before the recent issuance.

Requires: on-time delivery, Phase 2 financing, and new tenants.

12Research summary

Digi Power X is an option on a real trend priced with real uncertainty. Stripped of narrative: a microcap that owns power capacity, has converted it into one large signed contract with a newly listed and currently profitable counterparty, has funded the first construction phase from its own cash, and has roughly five months until the date on which the case either proves itself or unravels.

What is verifiably strong. No long-term debt, $128.1M of cash at June 30, 2026, a signed ten year contract worth about $1.1B, long lead equipment secured and a substation built. At this size those are not trivial, and several microcap peers with louder narratives have none of them.

What is verifiably weak. Revenue is falling, gross margin is negative and widened to a $(5,524,064) loss in the second quarter, management compensation consumed roughly a third of quarterly revenue, and the share count rose 199% in the nineteen months to July 2026 and again since, to 101,393,355 at August 14, 2026. The AI segment has now reported its first revenue, $1,082,592, against a legacy mining line that fell 88% to $161,422 in the same quarter. The distance between release framing and filed figures, Adjusted EBITDA of +$1.1M alongside filed EBITDA of −$3.2M, warrants scepticism toward promotional communications generally.

What is not knowable today. Whether Phase 1 meets 15 December. Whether project debt closes and on what terms. Whether the uncontracted two thirds of the fiscal 2027 target finds tenants. Analysis cannot resolve these in advance; disclosure will.

What to watch, in order of weight.

1. Q2 2026 results on August 14, 2026: the first quarter with disclosed AI revenue, plus updated cash, capital expenditure and share count. 2. Any project level financing announcement, its terms and whether debt or equity, which determines the Phase 2 dilution question. 3. Construction language on Phase 1, watching for softening from December 2026 toward late 2026 or early 2027. 4. A second colocation tenant, the difference between one contract and a platform. 5. Program usage and a reconciliation of the $102.86M against $72,363,650 discrepancy. 6. Any insider purchase, which would be the first in eighteen months. 7. A named tenant for the roughly 15 MW of uncommitted Columbiana capacity.

The company cannot be valued on current financials, because those financials describe a business it is actively dismantling. It can only be assessed on whether a specific construction project completes on a specific date and is financed on acceptable terms.

13Null categories

Categories examined that produced nothing. An absent section is ambiguous; an explicit null is information.

CategoryStatusBasis
AI segment revenueNoneThe Tier III AI data center segment reported $0 in Q1 2026. First AI related revenues are described but unquantified. 10-Q, May 15, 2026
DividendsNoneNo distribution appears in any period presented in this document. Capital is being consumed by construction and raised by equity issuance. 10-K, Mar 31, 2026
Share buybacksNoneThe company is a net issuer throughout the periods presented, with an at-the-market program active. 10-Q, May 15, 2026
Total project budgetNoneNo project budget appears in the filings and releases listed above, which is why the cost band in section 08 is an analyst estimate. 10-K, Mar 31, 2026; 10-Q, May 15, 2026; S-3, Jul 17, 2026
Insider purchasesNoneNo purchase appears in insider transaction filings over roughly eighteen months. Forms 4, Jan to May 2026
Peter Lynch shareholdingNot establishedNo such holding appears in any of the six beneficial-ownership filings on this issuer since 2025, which are by Michel Amar and affiliates, by Citadel Advisors and Kenneth Griffin, and by Eleven Ventures and Hartley Wasko. Those report holdings above the beneficial ownership threshold only, so a smaller position would not appear and absence cannot be established from them. Also named in what was not checked. 13G/A, Dec 31, 2025
Late filing notificationsCovered aboveTwo Form NT 20-F notifications have been filed, on May 2, 2023 and May 1, 2024, both under the Digihost Technology name. The 2024 notice states the annual report could not be filed on time because the auditor was still completing its audit of the 2023 statements, which places the delay on the accounts rather than on administration. Established against the complete filing index for this issuer, 242 filings, searched August 4, 2026. No NT 10-K, NT 10-Q or NT 40-F appears.
Schedule 13D positionsNoneNo Schedule 13D of any kind appears across the complete filing index for this issuer, 242 filings, searched August 4, 2026, so no activist position is on record. The beneficial-ownership record consists of five Schedule 13G and 13G/A filings since 2025 and seven earlier ones under the previous form designation. The five are not all on the same footing: Citadel and Eleven Ventures file under the passive-investor exemption in Exchange Act Rule 13d-1(c), while both of the Amar statements are filed under Exchange Act Rule 13d-1(d), which is open to a holder whose position predates the registration of the class and rests on no certification of passivity.
Analyst coverageCovered aboveCoverage is tabled in section 08 and runs to four firms and one quantitative ratings service, three of them carrying a target. A consensus drawn from three contributors struck eleven months apart carries little averaging benefit, and none is computed here.
Price target issued hereNoneNo target price is issued in this document. Third party targets are reported in section 08 for context only.

14Methodology & sources

Pricing basis and tagging

Market figures are struck on the August 14, 2026 close at $3.97: the exchange reports that session closed, and volume of 6,783,273 shares corroborates it at about 1.96 times the mean of the nine preceding sessions, which is consistent with the quarterly report having been filed that day. Every material claim carries a provenance tag, defined in the table below. Tags are not color coded, because a filed fact can be bad news.

Provenance tags

TagWhat it asserts
FiledStated in an SEC filing or company release, cited by form and date. Also covers a dated official publication by a named non-SEC issuer, with the issuer named in the source line. The extension is used here for the Federal Open Market Committee statement of July 29, 2026 and its accompanying projections.
EstimateDerived or inferred here. The arithmetic is shown.
OpenExpected but unconfirmed. Nothing filed either way.
MarketPrice, volume, float, 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.

Sources

Filings and company releases are primary. Market and sector data are used for corroboration and for figures that only market sources carry.

SourceDateUsed for
Form 10-Q, quarter ended Mar 31, 2026May 15, 2026Q1 financials, segment data, share counts, digital assets, warrants, options, RSUs, related party compensation
Form 10-K, year ended Dec 31, 2025Mar 31, 2026FY2025 revenue $34.19M, net loss $28.36M, loss per share $(0.64)
Form 10-K/A, Amendment No. 1Apr 27, 2026Part III: 2025 chief executive compensation of $12,025,227; 72,826,680 subordinate voting shares at that date; director, related party and auditor fee disclosure
Form 8-KMar 16, 2026US Data Centers structure: 55% stake, ~35% founder equity, ARMS only scope, equipment at cost
Form 8-KMay 8, 2026Cerebras colocation and master services agreement: phases, dates, financing condition, contract value
Form 424B5May 8, 2026Program upsizing to $175M; $72,363,650 sold under the first supplement; 3.0% commission; $6.19 close on 7 May
Form S-3Jul 17, 2026Resale registration for 380,590 warrant shares; 2 July share count of 98,543,355; option, warrant and RSU schedule; 108,446,376 fully diluted; 55 MW Columbiana description
Forms 4, Alec AmarJan to May 2026Sales dated 2 Jan, 2 Feb, 1 Apr, 8 Apr, 9 Apr and May 1, 2026, with no offsetting purchases
Schedule 13G/A, No. 3May 15, 2025Michel Amar, Bit.Management LLC and NYAM LLC, 6,465,414 shares, 17.5% of the class on 36,974,470 then outstanding
Schedule 13G/A, No. 1Aug 14, 2025Eleven Ventures LLC, Eleven Managers LLC and Hartley Wasko, 1,053,536 shares, 2.6%. The same filing states the holding fell to nil on July 7, 2025
Schedule 13GOct 17, 2025Citadel Advisors LLC and affiliates with Kenneth Griffin, 1,517,083 shares, 3.4% of the class on 44,707,644 then outstanding. Filed passive under Exchange Act Rule 13d-1(c)
Schedule 13G/A, No. 4Feb 13, 2026Michel Amar and affiliates, 8,047,081 shares, 11.5% of the class on 70,094,385 then outstanding
Schedule 13GMay 12, 2026Citadel Advisors LLC and affiliates with Kenneth Griffin, 3,550,691 shares, 4.9% of the class on 72,826,680 then outstanding. Filed passive under Exchange Act Rule 13d-1(c)
Form NT 20-FMay 2, 2023 and May 1, 2024Late-filing notifications for the 2022 and 2023 annual reports, both under the Digihost Technology name. The 2024 notice attributes the delay to the auditor still completing its audit of the 2023 statements
Company releases12 Jan, 20 Apr, 5 May, 8 May, 15 May, 3 Jun, Jul 7, 2026Supermicro purchase, SubQ AI agreement, Cerebras announcement, program upsizing, Q1 results and grants, Vera Rubin commitment, operations update
Third party research, Alliance Global Partners6 May and May 15, 2026Target raised to $10 from $7, then to $12 from $10, Buy maintained, analyst Brian Kinstlinger, as reported by the distributing service
Third party research, H.C. WainwrightNov 17, 2025Target of $5.00, analyst Kevin Dede, as reported by the distributing service
Market and sector dataAug 14, 2026 and earlierPrice, beta, 52 week range, trailing revenue, peer capacity and performance, Cerebras listing and results, credit spread reporting, Federal Open Market Committee statement of July 29, 2026

What was not checked

The filing record was swept to August 14, 2026 and re-read on August 18, 2026, covering the annual, quarterly, current, registration, prospectus, ownership and insider transaction families in both the domestic and foreign private issuer forms. It returned five filings after the shelf registration of July 17, 2026: the current report of August 4, 2026 setting the second quarter reporting date, and on August 14, 2026 the quarterly report, a current report and two amended Schedule 13G filings. It was run against the issuer central index key rather than the ticker, which matters for a company that renamed in March 2025. Two corrections came out of it. The ownership family had been read as a single founder position; it holds seven filings since 2025 by at least three unrelated parties, two of them original Schedule 13G filings rather than amendments. The two lodged on August 14, 2026 were not read, so whether they add a fourth filer is not established. And two Form NT 20-F late-filing notifications exist, on May 2, 2023 and May 1, 2024, against the statement carried here that none had ever been made. Both were re-established against the complete index, which now holds 246 filings. Market data is repriced to the August 14, 2026 close. Specifically not reached: the two amended Schedule 13G filings of August 14, 2026; any Canadian continuous disclosure lodged on SEDAR+ that does not have a U.S. counterpart; and any register of beneficial ownership below the threshold at which a Schedule 13G becomes reportable, which is why the shareholding named in section 13 reads Not established rather than None. An unexamined area is a gap rather than a clean bill.

Known limitations carried forward

Two defects are identified and not resolved in this version. Three carried at the previous revision have been closed: the figures owed to the source are restored to the body, the proportionate voting share count is now cited to a filing, and the third party target is attributed and dated.

Remaining program capacity is a band, not a figure. The text at issue reads: "anywhere from roughly $37M to $70M undrawn under the $175M program." The Q1 quarterly report and the 8 May prospectus supplement do not reconcile, and resolving them requires the Q2 quarterly report, which has not been filed. Until then the band stands.

The Columbiana cost band is an analyst estimate with no company figure behind it. The text at issue reads: "in the region of $7–12M per MW for powered shell, electrical plant and cooling, implying roughly $280–480M for the full campus." No project budget has been disclosed. No disclosed figure exists to replace the band.

15Document log

Newest first. One entry per revision date, consolidated as net change since the previous entry. The original build entry is preserved as written.

August 26, 2026 Latest
Undated investor presentation read · priced off Aug 14, 2026 close · $3.97 · ~$405M cap · not repriced

The company’s investor presentation was read for the first time, and it carries no date. Not on its cover and not on any of its 21 pages. It is published at a single address that is overwritten when the next version ships, with no archive and no version history, so it is cited throughout by the date it was read, August 20, 2026, and never by a date it carries. Everything taken from it is recorded rather than used, and nothing in this report is repriced or restamped on the strength of it.

The flagship capacity gap now has three numbers rather than two. The shelf describes Columbiana at 55 MW, the Cerebras contract covers 40 MW, and the presentation says the company is converting a wholly-owned 70 MW. Because the presentation cannot be dated it cannot be placed in sequence with the shelf, so the existing statement that the company has not addressed the gap stands, now qualified to say that it has not addressed it in anything datable.

The portfolio figure is described differently by the company than by this report. Roughly 400 MW is carried here as secured power, on filed company disclosure. The presentation calls approximately 409 MW capacity the portfolio is capable of supporting, and files it under a future-vision heading. The tag is not changed: the underlying company disclosure has not been re-read against the presentation, and a provenance tag is not moved on the strength of an undated page. The disagreement is recorded instead.

The company has published a $500M valuation of the same 40 MW this report costs at $280 to $480M to build. It is the issuer’s own EBITDA per megawatt times a comparable multiple the issuer selected for itself, so it is recorded beside the cost band and not used. It is a valuation and not a budget, so the open item about the missing project cost is not discharged by it.

August 18, 2026
Priced off Aug 14, 2026 close · $3.97 · ~$405M cap · not repriced
Correction

Four statements still dated this report's market data to the 4 August close. The sources table row, the 52-week window, the price beside the discount to the 52-week high, and the sentence reading "Market data is repriced to the August 4, 2026 close" all survived the reprice recorded below, which moved the masthead, the price and the capitalization to the 14 August close of $3.97. All four now read August 14. No figure moved: the 4 August close was $3.95 against $3.97 on 14 August, and the discount to the 52-week high is 57% on either.

The 52-week window was recomputed rather than re-dated. Across the daily record the window to the 31 July close, the window to the 4 August close and the window to the 14 August close each hold 251 sessions, an intraday range of $1.86 to $9.20, a close range of $1.89 to $8.46, the low on March 30, 2026 and the high on May 13, 2026. Only the date on which the window ends was stale.

The filing account was ten days behind this report's own log. It said the record was swept to August 4, 2026, that one filing had appeared since the July shelf registration, that the index held 242 filings, and that no second-quarter filing had been reached because none was due, while the entry below records those results as filed on August 14 and this report carries their figures. Counted from the submissions index today: 246 filings; five since the shelf registration of July 17, 2026, being the current report of August 4 and, on August 14, the quarterly report, a current report and two amended Schedule 13G filings; and seven ownership-family filings since 2025 rather than five, two of them originals. The two lodged on August 14 were not read.

No newsroom feed is recorded upstream for this issuer, so the sweep above establishes that nothing was filed since August 14 and cannot establish that nothing was announced.

Eight rows at the foot of the snapshot are still struck at the 31 July close and are left there. Prior close, session change, after-hours, session volume, average volume, employees, beta and a second 52-week range each carry their own date, so each is true as it stands. What was wrong was the lead-in above them, which said all market data was struck on the 14 August close; it now says which rows are not. Restriking them needs session data this revision did not fetch, and re-dating them without recomputing would make them false.

Not repriced, and the stamp moves without the basis. The 17 August session closed at $3.88; this report stays on the August 14, 2026 close of $3.97 and a capitalization of ~$405M. Two clauses that tied the basis to the stamp are removed rather than rewritten, because 17 August has since closed: the snapshot lead-in called 14 August the last completed session before this stamp, and section 14 called it the most recent completed session. Every figure in both sentences stands.

The 52-week range was dated two ways in one table, and the block below it cannot be restruck. The snapshot carried the range at the top dated August 14 and again at the foot dated July 31, both reading $1.86 to $9.20. Measured across the daily record the window is identical on every basis this document has used: to the 31 July, 4 August and 14 August closes it is 251 sessions, $1.86 to $9.20 intraday and $1.89 to $8.46 on closes, with the low on March 30, 2026 and the high on May 13, 2026. The date moves and the figure does not, and the lead-in above the table, which names the session, volume, beta and headcount rows as the ones still at the 31 July close, is now exactly true of what remains. Those remaining rows stay: the prior close, the session change and the session volume reproduce exactly from the daily record and could be restruck, but the average volume matches no window of it, beta is a provider figure with no source here, and an after-hours price for a past session is not retrievable at all. Moving three rows of four would split a block that is uniform and correctly dated.

August 17, 2026
Priced off Aug 14, 2026 close · $3.97 · ~$405M cap

Second-quarter 2026 results were filed on August 14, 2026, and the AI data center segment produced revenue for the first time: $1,082,592, disclosed as GPU rental. The segment was $0 in the first quarter, so the question this coverage has carried since the Cerebras agreement is answered, at a figure toward the lower middle of the range the analysis allowed for.

The same quarter showed the legacy business effectively stopping rather than declining, and the reading here is widened to say so. Digital currency mining and staking fell to $161,422 from $1,394,740, which is 2.4% of revenue against 17.2% a year earlier. Total revenue fell to $6,628,607 from $8,111,451. The gross loss widened to $(5,524,064) from $(92,081), on depreciation of $(4,068,734) against $(1,573,691) as the Columbiana build entered service, and general and administrative expenses reached $(9,493,911) from $(3,890,501). Net loss to common shareholders was $(14,360,509), or $(0.17) per share on 85,480,992 weighted average shares.

Figures moved as follows. Cash ~$155M at July 3 → $128.1M at June 30, now on a filed balance sheet rather than a release. Digital assets $13.56M → $14.30M. Shares outstanding 98,543,355 → 101,393,355, the count now taken from the quarterly report cover at August 14 rather than from the July shelf registration. As-converted count 99,209,955 → 102,059,955. Market capitalization ~$392M → ~$405M. Enterprise value ~$223M → ~$263M. Property, plant and equipment is added at $73.24M against $23.01M at the year end.

Repriced from the August 4, 2026 close of $3.95 to the August 14, 2026 close of $3.97, the last completed session. Eighteen sessions were read with no weekday absent, a second unrelated source agreed on all eighteen to the cent, and the feed reproduced both the $3.95 close of August 4 and the $3.80 close of July 30 that this document already carried.

A share count difference worth stating rather than smoothing. This document has carried 98,543,355 subordinate voting shares, sourced to the shelf registration of July 17 as a July 2 figure. The quarterly report's balance sheet gives 98,543,358 at June 30, three higher. Both are filed figures at stated dates and neither is corrected here; the current count of 101,393,355 supersedes both.

Correction

Every price shown beside a third-party target was struck on the action date rather than on the session before it. The price beside a published 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 note is out and sometimes moves because of it. All five rows move: Weiss Ratings, July 17, 2026, $3.77 → $3.72; Alliance Global Partners, May 15, 2026, $7.54 → $7.22; Alliance Global Partners, May 6, 2026, $6.24 → $5.11; H.C. Wainwright, November 17, 2025, $3.70 → $3.21; iA Financial, September 2, 2025, $2.37 → $2.20. The table stated no basis for the column at all, so the basis is now stated above it rather than left to be inferred.

The largest of those moves changes what the row says. The $10 target of May 6, 2026 was struck against a close of $5.11, not $6.24: the shares rose 22% on the day the note landed, so the same-day close carried the reaction to the Cerebras announcement that the target was responding to. Read on the right session the target sat 96% above the market rather than 60%.

The second quarter reporting date is filed rather than estimated. A current report of August 4, 2026 states that results for the quarter ended June 30, 2026 will be released on August 14, 2026, with a conference call at 8:30 a.m. Eastern. The date was previously carried as a 13 to 14 August window inferred from the statutory deadline and the prior year, and the companion calendar moves with it. Its provenance moves from market data to a filing.

Repriced to the 4 August close. The session ended at 16:00 Eastern, stamped by the exchange, on volume of 3,455,185 shares, about 1.02 times the mean of the nine preceding sessions. $3.85 → $3.95, capitalization ~$382M → ~$392M, enterprise value ~$213M → ~$223M, and the three enterprise value to sales multiples 6.7× / 0.85× / 0.71× → 7.0× / 0.89× / 0.74×. The discount to the 52-week high narrows 58% → 57%. The 52-week range of $1.86 to $9.20 is unchanged across 251 sessions, with the low on March 30, 2026 and the high on May 13, 2026.

Two figures were stated against a close they were not struck at. The loss carried by buyers into the at-the-market program read 68% against the 31 July close while the arithmetic beside it divided by the 3 August one, and the snapshot lead-in named the 31 July close where the table beneath it named 3 August. Both now read the 4 August close, and the loss is 57%. Separately, the count of published targets the current price sits above read two and is one: the shares are above $3.50 and below both $5.00 and $12.

Filing sweep re-run to August 4, 2026 against Central Index Key 0001854368, across the complete index, which now holds 242 filings and does not paginate. The one filing since the previous sweep is the current report above. Two Form NT 20-F notifications stand and no Schedule 13D appears, both re-established against the larger index.

August 3, 2026
Priced off Aug 3, 2026 close · $3.85 · ~$382M cap

Third-party coverage moves from prose to a table, and the set is larger than the prose said. The two firms already described keep their figures exactly: Alliance Global Partners at $12 from May 15, 2026 and H.C. Wainwright at $5.00 from November 17, 2025, with the intermediate raise to $10 on 6 May now shown as its own row. Added from a search on August 3, 2026: a $3.50 target from iA Financial dated September 2, 2025, and a Sell grade reiterated by a quantitative ratings service on July 17, 2026, which is the most recent action on this issuer and the only negative one. Each row now carries the close on its publication date, so the reader can see that the $12 was struck at $7.54 and the $3.50 at $2.37.

Correction

The account of who owns this company was wrong, not merely short. This report described the ownership record as “Schedule 13G/A filings for Michel Amar and affiliates” and stated that “only Schedule 13G/A filings appear”. Both are false. Five beneficial-ownership filings have been made since 2025 by three unrelated filers, and two of them are original Schedule 13G filings rather than amendments. The largest outside holder, Citadel Advisors and Kenneth Griffin at 3,550,691 shares and 4.9% as at May 5, 2026, appeared nowhere in this file, nor did Eleven Ventures, whose 1,053,536-share position was disclosed in the same filing that recorded its reduction to nil on July 7, 2025. The register now sits in capital structure, with each holding restated against the current class.

The founder’s holding was described as falling when it had not. This report recorded that Michel Amar’s position fell from 8,047,081 shares to roughly 5.97M “chiefly through dilution rather than sales”. Both figures are correct and they are the same holding on two bases: the difference of 2,081,600 is exactly the 666,600 shares issuable on conversion of proportionate voting shares plus 1,415,000 issuable on vested options. Nothing was sold and the count did not move. The fall from 11.5% to 6.6% is real and is a denominator effect.

The clean-filing record was not clean. This report stated that no NT 10-K, NT 10-Q, NT 20-F or NT 40-F had been filed at any point and that no filing had ever been late. Two Form NT 20-F notifications exist, on May 2, 2023 and May 1, 2024, both under the Digihost Technology name, the 2024 one because the auditor was still completing its audit of the 2023 statements. Established against the complete index of 241 filings. The companion claim that no Schedule 13D exists survives the same search and stands.

The beneficial-ownership record was re-read in full and it has three filers, not one. Five Schedule 13G and 13G/A filings have been made since 2025: Michel Amar and affiliates in May 2025 and February 2026, Eleven Ventures with Hartley Wasko in August 2025, and Citadel Advisors with Kenneth Griffin in October 2025 and May 2026. The register now appears in capital structure with every holding restated against the 98,543,355 shares outstanding at July 2, 2026, because each filer quotes its percentage against the class as it then stood and the class has grown faster than the filings. Citadel at 4.9% is the largest outside holder and sits one tenth of a point below the level that would bring it inside the regime governing the founder. Eleven Ventures is recorded as an exit rather than a holding: 1,053,536 shares at June 30, 2025 and nil from July 7, 2025, having taken 3,113,636 shares and an equal number of warrants in August 2024.

Two late-filing notifications were found where the file asserted none. Both are Form NT 20-F, filed May 2, 2023 and May 1, 2024 under the Digihost Technology name, and they are recorded in the filing history alongside the ownership filings. The 2024 notice attributes the delay to the auditor still completing its audit of the 2023 statements, so the annual report has been late in two of the last four years for a reason resting on the accounts.

A 52-week range is carried for the first time, $1.86 to $9.20 intraday across the 251 sessions to the 31 July close, with the low on March 30, 2026 and the high on May 13, 2026. The window sits wholly after the March 2025 change of symbol, so it is a full year of this ticker rather than a truncated one. At $3.85 the shares are 58% below the high.

Repriced to the August 3, 2026 close. The 3 August session has closed, so the basis moves July 31, 2026 → August 3, 2026 and the price $3.68 → $3.85, up 4.62% on the day. Market capitalization moves ~$365M → ~$382M on the unchanged 99,209,955 shares including proportionate voting as converted, enterprise value ~$196M → ~$213M on unchanged cash and digital assets, and the three enterprise value to sales multiples 6.2× / 0.79× / 0.66× → 6.7× / 0.85× / 0.71×. The discount to the 52-week high narrows 60% → 58%, and the loss carried by buyers into the at-the-market program narrows 68% → 61%. The as-of stamp was already 3 August and does not move.

August 2, 2026
Priced off Jul 31, 2026 close · $3.68 · ~$365M cap
Correction

A third party target was described as a consensus. The valuation section previously read that "the one year consensus sits near $12 from very few contributors". The $12 is not a consensus: it is Alliance Global Partners' own target, raised from $10 on May 15, 2026. The only other published target is H.C. Wainwright's $5.00 of November 17, 2025. The sentence has been replaced with both targets, each attributed to the firm that issued it and dated, and with the observation that averaging two figures struck eight months apart on different facts would describe nothing.

Filing sweep performed and reached August 2, 2026 across the annual, quarterly, current, registration, prospectus, ownership and insider transaction families in both the domestic and foreign private issuer forms. It returned nothing after the shelf registration of July 17, 2026. Two late-filing notifications do exist, both Form NT 20-F, on May 2, 2023 and May 1, 2024. Not repriced: the last close remains Jul 31, 2026 at $3.68, since 31 July was a Friday. Market capitalization, enterprise value and every multiple are therefore unchanged.

Backlog closed. Six figures owed to the record are restored to the body rather than to a quotation: the 31 July session change of −3.16% against a $3.80 prior close, the $3.61 after-hours price, session volume of 1.90M against a 4.75M average, and a headcount of 17, all now carried in section 01; and the founder holding of roughly 5.97M shares restored beside the 6.6% that survived alone. The comparison lost with an em dash is restored and recomputed rather than reinstated at its old value: $6.19 on May 7, 2026 is roughly 68% above the 31 July close of $3.68. The proportionate voting share count of 3,333, previously traceable to no source, is cited to the shelf registration of July 17, 2026, which also states the 666,600 subordinate shares issuable on its conversion, so the share reconciliation in section 01 now shows a filed input rather than an unexplained one. Three deferrals are removed; two remain, being the at-the-market capacity band and the Columbiana cost band, both of which still need a figure the filings do not carry.

Added from the sweep. President Alec Amar's 2025 total compensation of $7,425,695, so the two named officers together drew $19,450,922, or 57% of FY2025 revenue against 35% for the chief executive alone. Directors and executive officers as a group held 11.77% of the subordinate class plus all proportionate voting shares, about 12.55% of voting power at April 27, 2026. Both from the Part III amendment of April 27, 2026. A third disclosure quality flag is added: the annual report puts holdings at March 31, 2026 at approximately 51 bitcoin and 1,010 Ethereum at one named custodian, while the quarterly report covering the same date carries 166 Bitcoin and 1,013 Ethereum at $13.56M fair value, and the two are nowhere reconciled.

August 1, 2026
Priced off Jul 31, 2026 close · $3.68 · ~$365M cap
Correction

Insider alignment. The governance discussion previously leaned on founder ownership as evidence of alignment without examining the insider record. Forms 3 and 4 were never checked in the original build. They show President and director Alec Amar making six transactions in roughly eighteen months, all sales and none purchases, netting about 137,500 shares, with no insider buying at any point, including through the drawdown from the $9.20 high. The governance section now carries the actual record.

Refreshed against a Form S-3 filed July 17, a systematic sweep of every EDGAR form type, and macro data through the July 29 FOMC. Changes are stated as the net difference from the original build. No new 8-K, Q2 pre-announcement, proxy filing or litigation has appeared; Q2 earnings still track to August 13, 2026.

Share count, dilution and valuation. Share count 90,420,824 → 98,543,355 (Form S-3, July 2 figure); cumulative dilution since December 2024 restated +174% → +199%. The 8.1M-share increase surfaced only on the cover of a resale registration filed for an unrelated purpose, a disclosure mechanism now flagged explicitly in the risk section. Repriced $3.98 → $3.68 (Friday July 31 close); market cap ~$360M → ~$365M; enterprise value ~$194M → ~$196M; multiples 6.1× / 0.78× / 0.65× → 6.2× / 0.79× / 0.66×. Total shares including proportionate voting as converted are 99,209,955. ATM capacity ~$72M → a $37–70M range: the May 8 Form 424B5 reports $72,363,650 sold under the first supplement, which does not reconcile with the 10-Q's $102.86M; presented as an unresolved discrepancy rather than a point estimate, pending the Q2 10-Q. A.G.P. commission of 3.0% added. Overhang schedule replaced: options 3,485,000 at C$4.90, RSUs 4,670,946, warrants 1,080,475, being 9,236,421 shares, 9.4% of the count, against a stated 108,446,376 fully diluted. RSU pool grew 58% in three months. Separate C$9.84 option grant and 1,730,000 RSUs vesting 2027–29 added.

Governance and disclosure. CEO compensation added: the 10-K/A of April 27, 2026, a Part III amendment never read in the original build, discloses 2025 CEO compensation of $12,025,227, roughly 35% of FY2025 revenue at a company that lost $28.36M. New callout in the financial performance section. US Data Centers rewritten: the original build flagged the shifting ownership percentages as a disclosure inconsistency; a Form 8-K dated March 16, 2026 sets out the structure explicitly, being a 55% stake, ~35% founder equity held by the USDC management team including Hans Vestberg, ARMS-only scope with no site-level revenue participation, and equipment sold to DGXX at cost. The criticism now addresses messaging drift rather than substance. Clean-filing record confirmed: no Schedule 13D, and no NT 10-K, NT 10-Q, NT 20-F or NT 40-F at any point, so no filing has ever been late. Recorded in sources.

Macro, sector and operations. Interest rate direction reversed: the July 29 FOMC held 9–3 with all three dissents favoring a hike, and the dot plot points to an increase by year-end. New subsection in risks; anyone modeling this company on falling rates should revisit that input. Sector and credit deterioration added: neocloud peers fell 25–38% in a month, with contract wins and earnings beats failing to arrest it; AI-infrastructure credit spreads widened materially, with CoreWeave's 2032 bonds at 10.32% on a junk B rating, which sharpens the Phase 2 financing risk. Columbiana is a 55 MW build: against a 40 MW Cerebras contract, implying ~15 MW of flagship capacity under construction but uncommitted; portfolio-wide secured power of ~400 MW added to the business model section. Longer-dated guidance recorded: management targets $450–500M in 2028 and $800M–$1B in 2029, included for completeness with little weight attached at this stage.

July 26, 2026 (original build)
Priced off Jul 24, 2026 close · $3.98 · ~$360M cap

Report constructed from the Q1 2026 Form 10-Q (filed May 15), the FY2025 Form 10-K (March 31), Forms 8-K through July 7, and company press releases through the July 7 operations update. Priced off a $3.98 reference (July 24 close), a share count of 90,420,824, a market capitalization of roughly $360M and an enterprise value of ~$194M, giving multiples of 6.1× trailing and 0.78× / 0.65× on the FY2027 target range.

Structure: ten numbered sections from business model through final summary, with inline Fact / Estimate / Risk tagging throughout, a megawatt ladder for the growth path, and three scenario cases. No rating issued.