DGXX
DGXX · Digi Power X Inc. · Catalyst Calendar · as of August 18, 2026
Gating dependency. The chain runs: Phase 1 ready-for-service, then Cerebras revenue starting, then project financing closing, then Phase 2 at 25 MW, then the fiscal 2027 run-rate. A failure early in the chain invalidates everything downstream, including catalysts dated many months later.
Columbiana Phase 1 ready-for-service (D5) gates the Q4 and FY2026 results (D9), Phase 2 completion (D11) and the FY2027 run-rate validation (D12). The project-level financing announcement (D3) is an explicit contractual condition on Phase 2, so without it the 25 MW expansion does not happen irrespective of construction progress. Roughly two thirds of the fiscal 2027 revenue target depends on tenants not yet announced, tracked as a second colocation customer (D8), which sits outside the chain and has no gating event to watch.
Banded 0–3 / 3–6 / 6–12 months, earliest first. Numbering is gapless and runs straight through the bands, so a boundary never resets the count. An ID marks position in the current ordering and changes between revisions, so catalysts are named by title.
The company expects to open a Silicon Valley office by August 2026 and is building a dedicated engineering team for GPU-as-a-Service. The direct effect is small; the informational value is not. GPU-as-a-Service is an operating business rather than a landlord business and requires engineers, so named senior hires would show the $100M NeoCloudz target is being resourced.
Credible technical leadership from a recognized cloud or GPU operator would partly answer the capability question in governance and key-person control (S5).
Silence, or an announcement consisting of an office address, reads as narrative. G&A rose 60% year on year in Q1 while revenue fell 27%, so headcount worsens that ratio before it helps.
Source: Company release, July 7, 2026. Filed
The entire contracted revenue base depends on this one counterparty. Cerebras listed on May 14, 2026 raising roughly $5.5B and is well capitalized, but on June 23, 2026 it guided to declining gross margin and fell 10% after hours, having already dropped 28% from its debut. Timing is inferred from a single prior quarter because the post-listing reporting calendar is not yet established.
Continued revenue acceleration and reaffirmed capacity expansion would validate that Cerebras needs the Alabama megawatts it has contracted for.
Margin compression plus a roughly $3B expected cash burn across 2026 and 2027 could slow capacity commitments. The full dependency is set out in customer concentration (S2).
Source: Sector reporting, 23 June and May 14, 2026; Cerebras registration statement reporting. Market
Management said on July 7, 2026 that it is putting project-level financing in place, with details expected once definitive documentation is finalised. No deadline was given. This sits in the near band because the capital is needed to hold the Phase 2 timeline, not because a date exists. It is the highest-leverage item in the window: Phase 2 at 25 MW is explicitly conditioned on securing adequate financing, and whether that capital arrives as debt or equity determines the per-share outcome of the whole case.
Non-recourse project debt against the Cerebras lease at a sensible rate would fund Phase 2 without dilution and would represent third-party underwriting of the contract's bankability, a credibility signal no retail holder can replicate.
Silence is information: prolonged delay implies lenders are not comfortable, pushing the company toward the undrawn program and a further raise at a depressed price. Watch for high-coupon converts, attached warrants or equity-linked terms described as non-dilutive. The window narrowed in late July 2026, with credit default swap costs rising across AI infrastructure and CoreWeave 2032 bonds yielding 10.32% at a B rating.
Source: Company release, July 7, 2026; Form 8-K, May 8, 2026 for the Phase 2 financing condition. Filed
Inferred from the 45 day quarterly deadline for the period ending 30 September and the reporting pattern of the two prior quarters. No date has been announced. This is the final scheduled disclosure before 15 December, so construction commentary here is the most reliable public signal on whether Phase 1 lands on time, roughly a month out.
Explicit reaffirmation of 15 December with commissioning detail, and NeoCloudz revenue growing sequentially off the second quarter base.
Cash depletion accelerating as vertical construction consumes capital faster than the roughly $21.6M quarterly pace seen in Q1. Any hedged language on the ready-for-service date. Continued negative gross margin with no inflection.
Source: Statutory filing deadlines; Q1 and Q2 2026 reporting pattern. Estimate
A hard, contractual, publicly stated date, reaffirmed on July 7, 2026. The case resolves here: the company either becomes a revenue-generating AI infrastructure operator or one that missed its own contractual date.
On-time delivery converts a promise into cash flow, starts Cerebras payments, and creates a reference deployment the sales effort can show the next tenant, which is what unlocks a second colocation customer (D8).
Slippage is common even with equipment secured, with commissioning faults and labour the usual causes. The agreement carries service credit, penalty and termination provisions, so delay may cost directly as well as reputationally. Watch for language drifting from December 2026 toward late 2026 or year-end at the Q3 2026 results (D4); that softening is the early warning.
Source: Form 8-K, May 8, 2026; company release, July 7, 2026. Filed
The condition must resolve before the end of the first fiscal quarter of 2027 for the 40 MW target to hold, and capital has to be committed months ahead of that to fund the build, which places the decision point in this band. It determines whether this is a 15 MW company or a 40 MW company: fiscal 2027 colocation guidance of $80–100M assumes a full year of Phase 1 and a partial year of Phase 2.
Financing closed and Phase 2 construction commenced on schedule keeps the full contract value intact and validates the capital plan.
My estimate is that 40 MW of Tier III liquid cooled capacity costs roughly $280–480M at $7–12M per MW. With roughly $95M deployed and $128.1M of cash at June 30, 2026, Phase 1 is fundable and Phase 2 likely is not without outside capital. The company has never disclosed a total project budget, and that absence is itself a flag.
The $7–12M per MW cost band and the $280–480M campus figure derived from it are mine, not company-disclosed. Estimate
Source: Form 8-K, May 8, 2026; company releases, 15 May and July 7, 2026. Filed
A third-party data provider lists an annual meeting on June 15, 2026, but no Form 8-K under Item 5.07 reporting voting results and no 2026 proxy statement exists, and the sweep to August 2, 2026 confirms it. Two pieces of evidence now weigh against that date. The annual report stated that Part III would be incorporated by reference from a proxy filed within 120 days of the year end; instead the company supplied Part III directly by amendment on April 27, 2026, which is the route an issuer takes when the proxy will not arrive in that window. And the previous annual meeting cycle ran to December 2025, when proxy materials and voting results were filed. A June 2026 meeting is therefore doubtful and a later 2026 meeting more likely, but the item stays unresolved rather than upcoming until a filing settles it. It matters more than usual: this would be the first proxy since the company became a domestic filer, so compensation disclosure will be fuller than shareholders have previously seen, against key management remuneration of $2.42M in Q1 alone on $6.79M of revenue, which is roughly 36 cents of every revenue dollar.
Board additions with genuine data center or power engineering operating experience would address the capability question in governance and key-person control (S5).
Any proposal to increase authorised shares or expand the equity incentive pool would foreshadow dilution beyond what the ATM issuance mechanic (S1) already tracks.
Source: 2025 annual meeting precedent; no 2026 proxy statement located as of July 26, 2026. Open
The fiscal 2027 target assumes 90 MW of aggregate colocation, 50 MW beyond the Cerebras agreement, and no tenant has been announced for any of it. This is the difference between a one-contract company and a platform, and it represents up to roughly $100M of the $250–300M target, so its absence caps the achievable outcome however well Phase 1 goes. It stays a dated catalyst rather than a standing condition because it will resolve as a discrete, announceable event.
A named second tenant would be the strongest refutation of the promotional-microcap critique and would reduce the concentration in customer concentration (S2). The shelf registration describes Columbiana as a 55 MW build against a 40 MW contract, implying roughly 15 MW of flagship capacity under construction but uncommitted: a tenant for that needs no new site and no new substation.
Sub-scale operators struggle in competitive tenders while peers contract in gigawatts. Watch for letters of intent presented as contracts: a January 2026 letter of intent covering a 1.3 GW power plant has produced no binding agreement.
Source: Company release, July 7, 2026 for the fiscal 2027 outlook; company news archive, January 7, 2026 for the letter of intent. Open
Inferred: FY2025 was reported on March 31, 2026 and the annual deadline for a smaller reporting company is 90 days after year end. This is the first reporting period containing actual Cerebras colocation revenue, roughly two weeks of it if Phase 1 energised on 15 December, and the first full audited year under U.S. GAAP.
Even a partial quarter at contracted rates lets the run-rate be modelled directly rather than taken from management targets. A clean audit with no going concern language.
First-year conversions are where restatements and control weaknesses surface, and the Q1 narrative already contained a plain error, describing the prior-year comparative as net income of $1,633,261 when the statements show that figure as a loss. Watch the internal control section closely.
Source: Form 10-K filed March 31, 2026; statutory filing deadlines; Form 10-Q narrative, May 15, 2026. Estimate
The conference runs semi-annually, in March in San Jose plus an autumn edition. The March 2026 event ran 16 to 19 March; the March 2027 edition falls inside this window, though exact dates are unconfirmed. It matters indirectly: the roadmap determines the NeoCloudz hardware cycle and the depreciation profile of the existing B200 and B300 fleet.
Roadmap announcements extending the useful life of current-generation hardware would help the economics in NeoCloudz scaling and Vera Rubin deployment (D13).
Faster generational turnover accelerates obsolescence of the fleet bought in 2026. Separately, company conference-attendance releases are marketing rather than catalysts, a recurring feature of the news flow noted in governance and key-person control (S5).
Source: Conference historical schedule; company news archive, January 20, 2026. Market
Completes the Cerebras contract. Full 40 MW at the implied ten year average of roughly $110M a year is the point at which the company would become a profitable operating business rather than a story. Conditional on the financing tracked in the project-level financing announcement (D3) and the Phase 2 financing condition (D6). The company narrows this quarter-end to a month and names an earlier step this row does not carry. Page 10 of its investor presentation published at the company’s investor page and read on August 20, 2026 anticipates Phase 1 at 15 MW live for December and Phase 2 adding 25 MW in March, and page 11 dates the Phase 2 expansion to March 2027. Estimate The timing above does not change. March sits inside the stated quarter, so the two agree, and an undated presentation carries nothing that could be weighed against a filed source anyway. The December step for Phase 1 is recorded here rather than opened as a dated item of its own, for the same reason: a month with no year stated, in a document with no date, is not a timing this calendar can stand behind.
Full deployment on schedule delivers the contracted revenue base and would likely trigger the largest single re-rating in the window.
The most heavily conditioned item here: it requires Phase 1 to succeed, financing to close, and construction to execute a second time. Three sequential conditions, each independently capable of stopping it.
Source: Form 8-K, May 8, 2026; company release, July 7, 2026. Filed
Tested progressively through fiscal 2027, with the first read at first quarter results. This is the figure the current valuation leans on. At an estimated enterprise value of roughly $196M, $250–300M of revenue implies 0.66 to 0.79 times sales, against 6.2 times on trailing revenue of $31.7M. The distance between those two multiples is the whole debate. Management has also guided to $450–500M in 2028 and $800M to $1B in 2029, which sit several dependency layers further out and are not treated as catalysts here.
Approaching even the low end would justify a re-rating from 6.2 times trailing sales toward the sub-1 times forward multiple, amplified by the relatively small share count.
Only the Cerebras portion, guided at $80–100M, rests on a signed contract. Roughly two thirds of the target is uncontracted: up to $100M from unnamed colocation tenants and up to $100M from GPU services.
Source: Company release, July 7, 2026 for the 2027 outlook; valuation figures are analyst calculations. Filed
Targeting a year-end annualised run-rate of up to roughly $100M at about 10 MW of deployed GPU capacity, supported by a $35M Vera Rubin commitment announced June 3, 2026, with delivery following the vendor roadmap. The company states a higher ceiling on a different basis. Page 15 of its 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. 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 This row is not moved by it. The $120M is an undated ceiling for a fleet of that size; the figure above is a dated fiscal 2027 run-rate target at about 10 MW. No date changes and this catalyst does not change state. This is the second-largest component of the fiscal 2027 target and is structurally different from colocation: the company owns the hardware, carries obsolescence risk, and keeps more of the economics. It stays a dated catalyst despite running continuously because it carries a specific management target with a year-end measurement point.
GPU-as-a-Service scales faster than construction because it needs no building, so it can grow while Phase 2 is still being built. It is the one item outside the gating chain.
Owning hardware means depreciation and technology risk. Rental rates for prior-generation hardware compress as new silicon arrives, so the Vera Rubin purchase that helps in 2027 also devalues the fleet bought in 2026. GPU rental is a more competitive market than powered shells.
Source: Company releases, 7 July and June 3, 2026. Filed
Ongoing and undated, most material first. These take “Why undated” in place of timing confidence, because rating a permanent condition for timing certainty is a category error. Standing does not mean secondary: the two most consequential variables here have no date.
Remaining capacity is genuinely uncertain. The Q1 quarterly report records $102.86M raised after 31 March; the prospectus supplement of 8 May states $72,363,650 had been sold under the first $75M supplement before that date. The two do not reconcile and likely span different programs or periods, leaving undrawn capacity under the $175M program somewhere between roughly $37M and $70M. Sales require no contemporaneous announcement, and the agent earns a 3.0% commission.
This is the denominator under every catalyst in this document, and it is now 101,393,355 subordinate voting shares at August 14, 2026, from the quarterly report cover. The balance sheet gives 98,543,358 at June 30, so 2,849,997 shares were issued in the six weeks between, and $159.5M of stock was sold for cash in the first half alone. Cumulative dilution since December 2024 was 199% at July 2, 2026 and is higher now.
Little or no further drawdown would signal that management believes the shares undervalued and that project debt is genuinely coming, preserving equity for existing holders.
The May to July issuance occurred below the $5.16 May average, confirming equity has been the financing route so far. This is the mechanism by which the bear case actually plays out: dilution, not insolvency.
The $37–70M band is inferred from two filings that do not reconcile, not disclosed. The implied $32–41M of drawdown between 15 May and 2 July is derived from the share increase at prevailing prices. Estimate
Source: Form S-3, July 17, 2026; Form 424B5, May 8, 2026; Form 10-Q, May 15, 2026. Filed
Permanent for the life of the contract. One customer represents effectively all contracted future colocation revenue, and that customer in turn derives a large share of its own revenue from few customers, with OpenAI dominant. Shareholders therefore carry OpenAI spending risk two steps removed, with no visibility and no contractual protection. If Cerebras reduces, delays or exits, there is no diversified base to fall back on.
The counterparty is far stronger than a typical microcap anchor tenant: newly listed, raised roughly $5.5B, and reported 2025 net income near $237.8M. That removes one common failure mode outright. A second colocation customer (D8) is the only thing that structurally reduces this exposure.
Cerebras guided to declining gross margin on June 23, 2026 and expects to burn roughly $3B across 2026 and 2027. Note also the sequence: the contract was signed on 5 May, nine days before the Cerebras listing priced.
Source: Form 8-K, May 8, 2026; sector reporting, 23 June and May 14, 2026. Filed
Continuous across the window. The 52 week range of $1.86 to $9.20 is close to a fivefold spread in twelve months at a company whose operating results barely moved. The dominant risk with nothing to do with execution, and it has deteriorated: the cohort fell 25–38% in the month to mid July 2026, with IREN down 37% despite a five year $3.4B accelerator contract and Applied Digital sliding despite a 61% revenue beat. Good news ceasing to be rewarded says the sector is being repriced on financing risk rather than execution, which is the wrong backdrop for a company whose next milestone is a financing announcement.
Sustained capital expenditure growth keeps powered capacity scarce and supports premium multiples across the cohort. Hyperscaler Q2 capital expenditure guidance (R7) is the recurring window in which this becomes observable.
In a genuine sector de-rating, a pre-revenue 40 MW microcap is among the first things sold and among the last bought back, and execution would not protect the share price.
Source: Sector performance data, July 2026; credit spread reporting, July 25 and 29, 2026. Market
The 29 July meeting has occurred: the Federal Open Market Committee held at 3.50–3.75% on a 9–3 vote with all three dissents arguing to raise. Remaining 2026 meetings fall on 15–16 September, 27–28 October and 8–9 December, the first and last carrying projections. The direction has reversed: the June projections point to one quarter point increase by end 2026, futures price two, and the odds of a September hold sat near 42% after the July meeting. Both transmission channels now cut against the company, since rates set the cost of the project debt and long duration speculative equities trade on discount rates at a beta near 6.1.
An easing bias would lower Phase 2 financing costs at exactly the moment the capital is needed and would lift high beta risk assets. This is now the less likely path rather than the base case.
A rise raises financing costs directly into the Phase 2 decision window. Note a timing collision: the December meeting sits six days before Columbiana Phase 1 ready-for-service (D5), so a hawkish decision alongside any slippage would compound in a single week.
Source: Federal Open Market Committee statement, July 29, 2026, and its June 2026 projections. Filed Contemporaneous coverage, July 29 to 30, 2026. Market
Michel Amar and affiliated entities held 8,047,081 shares, 11.5% of the class, at December 31, 2025, since diluted to roughly 6.6% by May 2026. The dual class structure gives proportionate voting shares 200 votes each, and Amar is simultaneously founder, chairman and chief executive. Compensation is the sharpest of the concerns: Amendment No. 1 to the FY2025 annual report, filed April 27, 2026, discloses 2025 chief executive compensation of $12,025,227, roughly 35% of the company's entire FY2025 revenue of $34.19M at a company that lost $28.36M, and it appears only in the amendment rather than the original filing.
The company has attracted senior outside names: Hans Vestberg, former chief executive of Verizon, co-founded the next phase of US Data Centers and holds roughly 35% founder equity in it, a deeper commitment than an advisory title implies. Ajay Gupta joined the board in October 2025.
Insiders have been consistent sellers with no offsetting buying: insider transaction filings show the president and director executing six transactions in roughly eighteen months, all sales and none purchases, netting about 137,500 shares at prices of $2.72 on 2 January, $2.25 on 1 April, $2.26 on 8 April and $3.56 on May 1, 2026 among others, with intent-to-sell notices covering a further 83,000 and no purchase during the drawdown from the $9.20 high. Separately, the Q1 release headlines Adjusted EBITDA of +$1.1M while the segment note in the same filing reports EBITDA of −$3.20M. Promotional material has also associated the company with a well-known former fund manager as a backer; no such holding was found in the ownership filings reviewed, and it should be treated as unverified.
Source: Form 10-K/A, April 30, 2026, Part III; Forms 4, January to May 2026; Schedule 13G/A, December 31, 2025; Form 10-Q and company release, May 15, 2026. Filed
Continuous, crystallising at each quarter end. The company holds 166 Bitcoin and 1,013 Ethereum, carried at $13.56M against a $17.33M cost basis at March 31, 2026 and therefore below cost. Holdings revalue through the income statement, so a $3.76M revaluation loss flowed straight through Q1 2026 results against a $4.65M net loss, distorting the quarters investors most need to read cleanly.
A rally produces reported gains and a modest liquidity cushion, and the position could be monetised to part-fund construction without issuing shares under the ATM issuance mechanic (S1).
Further declines produce headline losses unrelated to the AI business, obscuring operating progress at the worst possible moment.
Source: Form 10-Q, May 15, 2026, Note 3. Filed
Closed items, kept for the record, ordered by resolution date earliest first. Where an item spans dates, the resolution date is the date it completed, not the date it began. A resolved catalyst takes the next free R number and does not carry an earlier number over. These are the track record against which the dated catalysts should be judged.
A dispute over compensation tied to the registered direct financing of July 23, 2025. Announced January 9, 2026, granted final exchange approval on 17 February and closed on February 20, 2026, settled for US$840,000.12 in cash and a five year warrant over 269,231 shares at $2.85. Small in dollar terms and fully resolved, but worth keeping on the record because it explains the otherwise unexplained warrant line in the quarterly report. A dispute with one's own placement agent over fees is a minor governance signal, folded into governance and key-person control (S5), and the same firm also appears as a research provider on the stock.
Source: Forms 8-K, 9 January and February 20, 2026. Filed
Uplisted from the TSX Venture Exchange to Cboe Canada on February 27, 2026, and ceased foreign private issuer status with conversion to U.S. GAAP effective January 1, 2026. The company now files quarterly and current reports as a domestic filer rather than the foreign private issuer equivalents, and reports under U.S. GAAP for all periods presented. This is why quarterly segment data exists at all, and why the earnings dates in this document have reliable statutory deadlines. What remains open is that first-year conversions are where restatements surface, and FY2026 will be the first full audited year on the new basis.
Source: Form 10-Q, May 15, 2026, Notes 1 and 2. Filed
A ten year data center colocation and master services agreement for roughly 40 MW at Columbiana, Alabama, with an initial term value near $1.1B and up to roughly $2.5B with a seven year extension, and exclusivity for the tenant at the facility. Announced 5 May and filed May 8, 2026. Transformational and already in the price: the stock ran hard on the announcement and Alliance Global Partners, analyst Brian Kinstlinger, raised its target to $10 from $7 on May 6, 2026, going to $12 on May 15, 2026 and keeping a Buy rating throughout. Market May 15, 2026 What remains open is that the contract is signed but not yet earning, it carries service credit, penalty and termination provisions, and the Phase 2 portion is conditioned on financing. The concentration it created is tracked at customer concentration (S2).
Source: Form 8-K, May 8, 2026; company release, May 5, 2026. Filed
The program was upsized on May 8, 2026, and 19,950,000 subordinate voting shares were issued for gross proceeds of $102.86M after 31 March, disclosed in the quarterly report filed 15 May. The share count rose from 69,807,452 at 31 March to 90,420,824 by 15 May. This is the source of the cash and no debt position, which funds Phase 1 self-construction and is genuinely valuable, but it was bought with roughly 30% dilution in six weeks. Undrawn capacity is now tracked as a standing condition at the ATM issuance mechanic (S1) rather than as a closed event, and those May buyers are underwater at the current price.
Source: Form 10-Q, May 15, 2026; company release, May 8, 2026. Filed
An ARMS 200 pod has run at Tier 3 standards since May 15, 2026, with NeoCloudz running on B200 and B300 hardware since the second quarter, backed by a $20M Supermicro purchase on January 12, 2026, a roughly $19.6M twenty-four month GPU rental agreement on April 20, 2026 and a $35M accelerator commitment on June 3, 2026. De-risking that counts: the company has shown it can physically run AI workloads rather than only announce an intention to. What remains open is that no dollar figure has ever been disclosed, so first AI revenues could be $50,000 or $5M. The Q2 2026 results (R8) are where this is quantified for the first time.
Source: Company releases, 7 July, 3 June, 20 April and January 12, 2026. Filed
Commitments are in place for every major long-lead Phase 1 item, including critical electrical and switchgear infrastructure, with the on-site substation complete, a power delivery agreement with the utility secured, and crews erecting the building shell. This retires the most common cause of data center schedule slip, since transformer and switchgear lead times have derailed several peer projects, and it raises the odds on Columbiana Phase 1 ready-for-service (D5). What remains open is that commitments in place is not the same as delivered and installed: commissioning, labour and integration risk are all still live.
Source: Company release, July 7, 2026. Filed
The question was whether continued escalation or a signal of digestion arrived, and the answer is escalation by a wide margin. Discrete quarterly capital expenditure, differenced from the year-to-date figures each issuer files in its own cash-flow statement:
| Issuer | Quarter to Jun 30, 2026 | Year earlier | Change |
|---|---|---|---|
| Microsoft | $35.80B | $17.08B | +110% |
| Alphabet | $44.92B | $22.45B | +100% |
| Amazon | $54.21B | $32.18B | +68% |
| Meta | $30.12B | $16.54B | +82% |
| Total | $165.05B | $88.25B | +87% |
These are cash actually spent, not guidance. Filed Each figure is a discrete quarter derived by differencing consecutive year-to-date amounts from the issuer's own Form 10-Q or 10-K, read through the SEC's XBRL company-concept endpoint on August 13, 2026. Estimate The differencing is arithmetic on filed figures, not a model. Amazon reports the line under a different tag from the other three, and a search on the common tag returns nothing for it after 2017.
What it means here, and the limit of it. DGXX holds no hyperscaler relationship and trades on AI infrastructure sentiment, with a five year beta near 6.1, so the direction matters more than the level. Spending that doubles reinforces the scarcity case for powered capacity. It does not validate this company's execution, and the sector still discriminates between contracted and uncontracted capacity, which is the distinction DGXX has yet to answer with a second named customer.
Source: SEC XBRL company-concept endpoint, us-gaap PaymentsToAcquirePropertyPlantAndEquipment and, for Amazon, PaymentsToAcquireProductiveAssets, read August 13, 2026. Filed
AI revenue arrived as a figure rather than a phrase, at $1,082,592 of GPU rental, and cash closed the quarter at $128.1M with no long-term debt after $159.5M of share issuance in the half.
It arrived alongside a collapse elsewhere. Mining and staking fell 88% to $161,422, total revenue fell 18%, and the gross loss went from $(92,081) to $(5,524,064) as depreciation more than doubled on the new build.
The question this catalyst existed to settle is answered: the AI data center line is $1,082,592, disclosed as GPU rental and appearing for the first time. That is inside the range the analysis carried and toward its lower middle.
The quarter said something larger than the question asked of it. 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, so the legacy business has effectively stopped rather than declined. Total revenue fell to $6,628,607 from $8,111,451. The gross loss widened sixty-fold to $(5,524,064) from $(92,081), driven by 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). The net loss to common shareholders was $(14,360,509), or $(0.17) per share on 85,480,992 weighted average shares.
Reading this quarter only as the one that quantified AI revenue would be accurate and misleading. The transition is now visible in both directions at once.
Source: Form 10-Q for the quarter ended June 30, 2026, filed August 14, 2026, and the results release on Form 8-K of the same date. Filed
Every category reviewed appears as a row, including those that produced catalysts, which read across by title and ID. A category checked and found empty is recorded rather than omitted.
| Category | Status | Basis |
|---|---|---|
| Earnings | Covered above | Q2 2026 results (R8), Q3 2026 results (D4), Q4 and FY2026 results (D9). |
| Product launches | Covered above | NeoCloudz scaling and Vera Rubin deployment (D13); resolved as NeoCloudz live and ARMS 200 pod operating at Tier 3 (R5). |
| Major contracts | Covered above | A second colocation customer (D8); resolved as the Cerebras colocation agreement executed (R3). |
| Macro events | Covered above | Hyperscaler Q2 capital expenditure guidance (R7); standing as the Federal Reserve policy path (S4) and AI infrastructure sector sentiment (S3). |
| Industry conferences | Covered above | NVIDIA GTC, GPU roadmap and hardware cycle (D10). |
| Investor days | None | No capital markets day appears in the current reports and company releases listed in sources through July 7, 2026. A site visit around Phase 1 energisation would be a natural move, but nothing is on the calendar. |
| Management changes | None | No departure or pending change appears in the current reports and company releases listed in sources through July 7, 2026. Appointments during 2026 include a chief technology officer leading the ARMS 200 platform and NeoCloudz, and Hans Vestberg, who co-founded the next phase of US Data Centers; Ajay Gupta joined the board in October 2025. Amar remains founder, chairman and chief executive, with Paul Ciullo as chief financial officer. What is absent is more informative: no insider has bought stock in eighteen months, tracked at governance and key-person control (S5). |
| Lawsuits | None | No securities class action or shareholder investigation was located as of July 26, 2026, checked against the current reports listed in sources and contemporaneous reporting. Absence of evidence is not evidence of absence, so the search is stated rather than the conclusion. Several microcap peers in this cohort do carry active suits, so the absence is a relative positive. The only litigation item is the settled H.C. Wainwright compensation dispute (R1). Also recorded, and correcting what was carried here before: two Form NT 20-F late-filing notifications have been filed, on May 2, 2023 and May 1, 2024, both under the Digihost Technology name, so the annual report has been late twice. The 2024 notice attributes the delay to the auditor still completing its audit of the 2023 statements, which places it 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. |
| Regulatory decisions | None | No decision is pending in the current reports and company releases listed in sources: colocation and power sales carry no product approval regime. Permitting and interconnection are handled at project level and are already secured for Phase 1 through the utility agreement. |
| Dividends | None | No dividend is recorded in the annual and quarterly reports listed in sources, and the company states it does not intend to pay one. Capital is being consumed by construction, so a distribution would be irrational here. Holders can only realize a return by selling. |
| Share buybacks | None | No repurchase program appears in the annual and quarterly reports or the releases listed in sources. The company is a net issuer with undrawn at-the-market capacity, tracked at the ATM issuance mechanic (S1). Expect dilution rather than accretion. |
| Mergers and acquisitions | None | Nothing pending in the current reports and company releases listed in sources. A January 2026 letter of intent covering a 1.3 GW power plant has produced no binding agreement, so letters of intent are treated as non-events until definitive documents are filed. |
| Index inclusion and coverage | Not established | No index constituent list was consulted, so membership cannot be confirmed either way. A market capitalization near $365M sits below major index thresholds on the market data listed in sources, and sell-side coverage runs to two contributors, so a stated consensus carries little averaging benefit. Also named in what was not checked. |
Sits across every other catalyst rather than beside them. A calendar implicitly assumes the denominator is fixed; here it is not, and it moved again between builds.
Good news arriving after heavy dilution produces a different per-share outcome from the same news arriving before it. The ATM issuance mechanic (S1) covers the standing mechanism; this section carries the arithmetic.
| Component | Shares or amount | Note |
|---|---|---|
| Share count, Dec 2024 | 33,011,600 | Starting point for the dilution series Filed |
| Share count, Dec 2025 | 69,427,788 | Up 110% Filed |
| Share count, May 15, 2026 | 90,420,824 | Up 174% Filed |
| Share count, Jul 2, 2026 | 98,543,355 | Up 199% in nineteen months Filed |
| Issued 15 May to Jul 2, 2026 | 8,122,531 | Roughly 9% dilution in seven weeks, surfacing on the cover of a resale registration filed for an unrelated purpose Filed |
| Implied drawdown in that period | ~$32–41M | At prevailing prices Estimate |
| Undrawn program capacity | ~$37–70M | Two filings do not reconcile; band rather than point. At the 14 August close of $3.97 that spans roughly 9M to 18M additional shares Estimate |
| Options | 3,485,000 | Weighted average C$4.90, plus a separate 650,000 at C$9.84 Filed |
| Unvested RSUs | 4,670,946 | Pool grew 58% in three months; a further 1,730,000 vest 2027 to 2029 Filed |
| Warrants | 1,080,475 | $2.98 average strike Filed |
| Total instrument overhang | 9,236,421 | 9.4% of shares outstanding Estimate |
| Fully diluted shares | 108,446,376 | As stated by the company Filed |
If project debt does not close, Phase 2 must be equity funded. Against an estimated $280–480M total project cost and $128.1M of cash at June 30, 2026, that gap is large enough to move the share count by a further order of magnitude.
When Columbiana Phase 1 ready-for-service (D5) or Phase 2 completion (D11) resolves favorably, check the share count in the same filing before concluding what it is worth per share, and check filings unrelated to the milestone, because that is where the last increase appeared. The bear case here is not bankruptcy: with no long-term debt, no lender can force an outcome. It is severe dilution and a long wait, and this overlay is where that risk lives.
What would prove this read wrong, stated in advance. Each trigger is settleable from a document; an unquantified threshold is not a test. Every threshold below is one this document already states.
| # | If this happens… | …this was wrong | Status |
|---|---|---|---|
| 1 | The company does not confirm Phase 1 ready-for-service on or before December 15, 2026. | The High timing confidence on Columbiana Phase 1 ready-for-service (D5), and the gating chain that the Q4 and FY2026 results (D9), Phase 2 completion (D11) and the FY2027 run-rate validation (D12) depend on. | Untriggered |
| 2 | Any Form 4 records an open-market purchase of shares by an officer or director. | The claim in governance and key-person control (S5) that there has been no insider buying at any point in roughly eighteen months. | Untriggered |
| 3 | The Q2 2026 quarterly report discloses undrawn at-the-market capacity outside the $37M to $70M band. | The band stated in the ATM issuance mechanic (S1) and carried into the dilution and capital overlay. | Untriggered |
| 4 | A colocation agreement with a counterparty other than Cerebras is filed or announced for any company site. | The concentration claim in customer concentration (S2) that one customer represents effectively all contracted colocation revenue. | Untriggered |
| 5 | Phase 2 financing closes as non-recourse project debt with no equity or equity-linked component. | The assessment in the ATM issuance mechanic (S1) that equity has been the financing route, and the downside framing in the project-level financing announcement (D3). | Untriggered |
| 6 | A Form 8-K under Item 5.07 reports voting results for a 2026 annual meeting. | The unresolved status recorded for the annual general meeting and proxy statement (D7). | Untriggered |
| 7 | An NT 10-K or NT 10-Q is filed. | The clean filing record recorded in the null categories and in methodology. | Untriggered |
| 8 | Reported fiscal 2027 revenue is below $250M. | The low end of the target the valuation leans on, tested at the FY2027 run-rate validation (D12). | Untriggered |
Every ID in one table, gapless within each class, matching the cards above exactly in IDs, count and order. Links point at title slugs so they survive renumbering.
Maps the earlier numbering to the current IDs, so earlier log entries remain resolvable.
| Was | Now | Catalyst |
|---|---|---|
| 1 | R7 | Hyperscaler Q2 capital expenditure guidance |
| 2 | D1 | Silicon Valley office opening and GPU engineering hires |
| 3 | R8 | Q2 2026 results, first disclosed AI revenue |
| 4 | D2 | Cerebras quarterly results |
| 5 | D3 | Project-level financing announcement |
| 6 | D4 | Q3 2026 results |
| 7 | D5 | Columbiana Phase 1 ready-for-service at 15 MW |
| 8 | D6 | Phase 2 financing condition satisfied, or not |
| 9 | D7 | Annual general meeting and proxy statement |
| 10 | D8 | A second colocation customer |
| 11 | D9 | Q4 and FY2026 results, first Cerebras revenue |
| 12 | D10 | NVIDIA GTC, GPU roadmap and hardware cycle |
| 13 | D11 | Phase 2 completion, full 40 MW deployment |
| 14 | D12 | FY2027 run-rate validation |
| 15 | D13 | NeoCloudz scaling and Vera Rubin deployment |
| S1 | S1 | Equity issuance mechanic, the at-the-market program |
| S5 | S2 | Customer concentration and the counterparty chain |
| S4 | S3 | AI infrastructure sector sentiment |
| S3 | S4 | Federal Reserve policy path |
| S6 | S5 | Governance, key-person control and promotional communications |
| S2 | S6 | Bitcoin and Ethereum revaluation |
| C3 | R1 | H.C. Wainwright compensation dispute settled |
| C6 | R2 | Structural transitions: uplisting and U.S. GAAP conversion |
| C1 | R3 | Cerebras colocation agreement executed |
| C2 | R4 | At-the-market program upsized to $175M |
| C4 | R5 | NeoCloudz live and ARMS 200 operating at Tier 3 |
| C5 | R6 | Phase 1 long-lead equipment fully secured |
| Tag | What it asserts |
|---|---|
| Filed | Stated 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. |
| Estimate | Derived or inferred here. The arithmetic is shown. |
| Open | Expected but unconfirmed. Nothing filed either way. |
| Market | Price, volume, float, published targets and ratings. Stamped with the close or publication date. |
| Level | Means |
|---|---|
| High | Date is company-announced, contractual, or statutorily fixed. |
| Medium | Date inferred from filing cadence or a stated deadline window. |
| Low | Event expected within the window with no schedule at all. Could move by a quarter or more. |
An item is dated if it will resolve as a discrete, announceable event, even where no schedule exists; it is a standing condition if it never resolves. The project-level financing announcement (D3), the annual general meeting and proxy statement (D7) and a second colocation customer (D8) all lack schedules but stay dated, because Low confidence for want of a schedule is a coherent timing statement whereas not applicable is not.
Two exceptions are stated rather than applied silently. Hyperscaler Q2 capital expenditure guidance (R7) stays dated although AI infrastructure sector sentiment (S3) covers the ambient exposure, because reporting dates are real and clustered and the figures are discrete disclosures. NeoCloudz scaling and Vera Rubin deployment (D13) stays dated although it runs continuously, because it carries a specific management target with a year-end measurement point.
Confidence rates timing, never outcome. A High-confidence catalyst can be a coin flip; a Low-confidence one can be near-certain in direction.
Impact is independent of confidence. A High-impact, Low-confidence item is the most important kind here: it matters enormously and could land any time. The project-level financing announcement (D3) is exactly that.
Defects identified but not resolved, with the text at issue quoted and the reason it remains open.
One assertion in this document has no falsification test, and the omission is deliberate rather than an oversight. The text at issue, in AI infrastructure sector sentiment (S3), reads: “In a genuine sector de-rating, a pre-revenue 40 MW microcap is among the first things sold and among the last bought back, and execution would not protect the share price.”
Every trigger in the falsification tests is settleable from a document because every threshold it uses is one this document already states. This claim is different: making it settleable requires a relative-performance threshold, saying how far the stock would have to outperform the cohort, over what window, during a drawdown of what depth. No such figure appears anywhere in this document or its sources, so choosing one would be new analysis rather than a restatement of work already done. It is therefore recorded as failed rather than passed with a note, and it stays open until the threshold can be derived from evidence. The governing requirement is that every trigger must name an observable event whose occurrence can be settled from a document, and an unquantified threshold does not meet it.
The earlier version of this document is no longer available as markup, and only a text extraction of it survives. The figure set is therefore recoverable and was used for the comparison above, so every figure claim in the preceding note rests on that extraction. The markup is not recoverable, so no structural comparison against the earlier version is possible for this document, and any defect visible only in its markup would go undetected here. The companion research report is unaffected: its earlier version survives in full. The seven figures this extraction showed were missing have since been restored to the body of this document and their deferral removed, so the comparison it supported is closed.
Primary filings and company releases first, with form type and date.
| Source | Date | Used for |
|---|---|---|
| Form 10-Q, quarter ended Mar 31, 2026 | May 15, 2026 | Segment revenue, share counts, digital assets, instrument schedules, structural transitions |
| Form 10-K, year ended Dec 31, 2025 | Mar 31, 2026 | FY2025 revenue and loss, annual reporting cadence for the earnings inference |
| Form 10-K/A, Amendment No. 1 | Apr 27, 2026 | Part III: 2025 chief executive compensation of $12,025,227 |
| Forms 8-K | 9 Jan, 20 Feb, 16 Mar, May 8, 2026 | Settlement announcement and closing, US Data Centers structure, Cerebras agreement |
| Form 424B5 | May 8, 2026 | Program upsizing, $72,363,650 sold under the first supplement, 3.0% commission |
| Form S-3 | Jul 17, 2026 | 2 July share count, instrument schedule, fully diluted figure, 55 MW Columbiana description |
| Forms 4 | Jan to May 2026 | Insider sales with no offsetting purchases |
| Schedule 13G/A | Dec 31, 2025 | Founder holding of 8,047,081 shares, 11.5% of the class |
| Company releases | 12 Jan, 20 Apr, 5 May, 8 May, 15 May, 3 Jun, Jul 7, 2026 | Contract announcements, Q1 results and grants, operations update and fiscal 2027 outlook |
| Third party research, Alliance Global Partners | 6 May and May 15, 2026 | Target raised to $10 from $7, then to $12, Buy maintained, analyst Brian Kinstlinger, as reported by the distributing service |
| Market and sector data | Aug 14, 2026 and earlier | Pricing basis, beta, 52 week range, peer performance, credit spreads, counterparty listing and results, Federal Open Market Committee statement of July 29, 2026 |
The filing record was swept to August 14, 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. The sweep was run against the issuer central index key rather than the ticker, which matters for a company that renamed in March 2025. Two Form NT 20-F late-filing notifications exist, on May 2, 2023 and May 1, 2024, and the ownership family now holds seven filings since 2025; the two lodged on August 14, 2026 were not read, so whether they add a fourth filer is not established. Established against the complete index, which now holds 246 filings. Categories and ranges not reached: the two amended ownership filings of August 14, 2026; any 2026 proxy statement or current report under Item 5.07, neither of which exists, which is why the annual general meeting and proxy statement (D7) is unresolved; Canadian continuous disclosure lodged on SEDAR+ without a U.S. counterpart; the counterparty's own post-listing filings beyond the single quarter already reported; and any index constituent list, which is why index inclusion reads Not established rather than None. Two figures could not be settled from the filings that were read: undrawn at-the-market capacity, where the quarterly report and the prospectus supplement disagree, and the total Columbiana project budget, which has never been disclosed. An unexamined area is a gap, not a clean bill.
Newest first. One entry per revision date, consolidated as net change since the previous entry. The original build entry is never removed or rewritten.
The company’s investor presentation was read for the first time, and it carries no date on any of its 21 pages. It sits at a single address that is overwritten when the next version ships, with no archive, so it is cited by the date it was read, August 20, 2026, and never by a date it carries. No catalyst changes state, no date moves, and nothing is repriced.
Two rows gained the company’s own figures and neither was retimed. The GPU scaling row now carries the presentation ceiling of roughly 4,000 GPUs against about $120M, which is an undated capacity ceiling and not a dated run-rate, so it does not displace the fiscal 2027 target the row tracks. The Phase 2 row now carries the presentation’s March 2027 month and its December step for Phase 1. The March sits inside the quarter already stated, so the timing is unchanged, and the December step is recorded rather than opened as an item because a month with no year, in a document with no date, is not a timing this calendar can stand behind.
This document cited a Form 10-K/A by a date on which nothing was filed. A source line read "Form 10-K/A, April 27, 2026, Part III". Digi Power X filed no document of any kind on April 27, 2026, and holds one Form 10-K/A, filed April 30, 2026. The date is corrected to the date of filing. The document cited is unchanged and the Part III reference stands.
The sources table dated this document's market data to the 31 July close. The row reading "Market and sector data, Jul 31, 2026 and earlier" survived two reprices, to the 3 August close and then to the 14 August close, and the second of those is the basis on the masthead. It now reads August 14, 2026 and earlier. Nothing was repriced and no figure moved: the 31 July close was $3.68, the 4 August close $3.95 and the 14 August close $3.97, which is the price this document carries.
One estimate still cited the session before the last reprice. The undrawn program capacity row read "At the 4 August close of $3.95"; it now reads the 14 August close of $3.97. The band it derives, roughly 9M to 18M additional shares, is unchanged on either close and is not restated.
The entry below states both that this document was repriced and that it was not. It records the move from the August 4 close to the August 14 close, and ends with a paragraph reading "Not repriced. Market figures remain on their previous basis." That paragraph, and the assessment above it, belong to the revision of 13 August, which the entry absorbed without keeping its date; the same is true of the statement that nothing had been filed since the August 4 Form 8-K, which the entry itself contradicts three paragraphs earlier. The reprice is the operative statement, and the masthead, the price and the exchange record all agree on it. The entry is left exactly as written, because a past entry is a record of what it said.
The filing record was re-read today. The submissions index holds 246 filings and nothing has been filed since the quarterly report, the current report and the two amended Schedule 13G filings of August 14, 2026. No newsroom feed is recorded upstream for this issuer, so this establishes that nothing was filed and cannot establish that nothing was announced. Digi Power X also files on SEDAR+, which is a third record this project does not read.
Not repriced, and the stamp moves without the basis. The 17 August session closed at $3.88 and this document stays on the August 14, 2026 close of $3.97. No catalyst resolved, none was added and no expected date moved.
The headline counted fifteen dated catalysts and there were fourteen. The count was correct when this calendar was first built on July 26, 2026 and stopped being correct on August 4, 2026, when one item resolved and the total fell to fourteen. It stood through two subsequent revisions. The headline is rewritten and no longer carries a count of the document's own contents.
One timing was malformed. The third-quarter results card read "~12–Nov 16, 2026", which states neither a single date nor a range that keeps one month. It reads "~Nov 12 to 16, 2026".
Second-quarter results were filed on August 14, 2026 and the catalyst written to settle the AI revenue question resolved. The answer is $1,082,592, disclosed as GPU rental and appearing for the first time. It moves to the resolved sequence, the dated sequence closes up from fourteen to thirteen, and resolved rises from seven to eight.
The quarter said more than the question asked of it, and the reading here is widened accordingly. Mining and staking fell to $161,422 from $1,394,740, so the legacy business has effectively stopped rather than declined: it 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), driven by depreciation of $(4,068,734) against $(1,573,691) as the Columbiana build entered service. Net loss to common shareholders was $(14,360,509), or $(0.17) per share.
Repriced. The basis moves from the August 4, 2026 close of $3.95 to the August 14, 2026 close of $3.97, the last completed session. The series was read across eighteen sessions with no weekday absent, and a second unrelated source agreed on all eighteen to the cent. The August 4 close of $3.95 and the July 30 close of $3.80 that this document already carried were both reproduced by the feed.
Hyperscaler Q2 capital expenditure resolved, and it resolved to the upside case. The item asked whether escalation continued or digestion appeared. Across Microsoft, Alphabet, Amazon and Meta the June quarter carried $165.05B of capital expenditure against $88.25B a year earlier, up 87%. These are cash figures from each issuer's own cash-flow statement, differenced from consecutive year-to-date amounts, not guidance commentary. Filed It moves to the resolved sequence.
Why it sat unresolved. Its timing read "Late Jul to early Aug 2026", which carries no parsable day, so the elapsed scan counted it among the timings it cannot read and this ticker never surfaced for it. A window that closes is not a date that passes, and nothing in the apparatus sees the difference.
Resolving the first dated catalyst shifted every later one down by one on August 13, 2026. Links point at title slugs, so a reference by title remains resolvable. The map below runs from the earlier identifiers to the current ones, so a reference by identifier resolves too.
| Was | Now | Catalyst |
|---|---|---|
| D1 | R7 | Hyperscaler Q2 capital expenditure guidance |
| D2 | D1 | Silicon Valley office opening and GPU engineering hires |
| D3 | D2 | Q2 2026 results, the first quarter with disclosed AI revenue |
| D4 | D3 | Cerebras quarterly results, counterparty read-through |
| D5 | D4 | Project-level financing announcement |
| D6 | D5 | Q3 2026 results, the last checkpoint before Phase 1 |
| D7 | D6 | Columbiana Phase 1 ready-for-service at 15 MW |
| D8 | D7 | Phase 2 financing condition satisfied, or not |
| D9 | D8 | Annual general meeting and proxy statement |
| D10 | D9 | A second colocation customer, the 50 MW with no name |
| D11 | D10 | Q4 and FY2026 results, first Cerebras revenue |
| D12 | D11 | NVIDIA GTC, GPU roadmap and hardware cycle |
| D13 | D12 | Phase 2 completion, full 40 MW deployment |
| D14 | D13 | FY2027 run-rate validation, the $250 to 300M test |
| D15 | D14 | NeoCloudz scaling to ~10 MW and Vera Rubin deployment |
Assessed August 13, 2026. The submissions index was read in full: nothing filed since the August 4 Form 8-K, which this calendar already reflects at the Q2 results item. The issuer newsroom could not be read: no feed is recorded upstream for this company, so this assessment establishes that nothing was FILED and cannot establish that nothing was announced. Digi Power X also files with the Canadian authorities on SEDAR+, which is a third record this project does not currently read.
Not repriced. Market figures remain on their previous basis.
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 Q2 2026 results (D2) carried a 13 to 14 August window inferred from the statutory deadline and the prior year; it now carries the filed date, and its provenance moves from market data to a filing. Timing confidence was already High and stays there. The item does not move in the ordering.
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. The undrawn at-the-market band of $37M to $70M spans roughly 9M to 18M shares at that close where the figure carried read 10M to 19M, which was struck at the 31 July close rather than the one it named.
Filing sweep re-run to August 4, 2026 against Central Index Key 0001854368. The complete index now holds 242 filings and does not paginate, and the current report above is the only filing since the previous sweep.
Third-party coverage tabled in the companion report. A search on August 3, 2026 found the set to be four firms and one quantitative ratings service rather than the two the earlier reading described. The most recent action is a Sell grade reiterated on July 17, 2026 by that ratings service. No catalyst changed class and no identifier moved.
The clean-filing record was not clean. This calendar recorded that no NT 10-K, NT 10-Q, NT 20-F or NT 40-F had ever been filed, and drew from it that no filing had ever been late. Two Form NT 20-F late-filing notifications exist, on May 2, 2023 and May 1, 2024, both under the Digihost Technology name. The 2024 notice attributes the delay to the auditor still completing its audit of the 2023 statements, so it rests on the accounts rather than on administration, and the annual report has been late in two of the last four years. Established against the complete filing index for this issuer, 241 filings. The companion claim that no Schedule 13D exists survives the same search and stands.
The ownership record has three filers, not one. Beneficial ownership was treated here as a single founder position. Five filings have been made since 2025 by three unrelated parties: Michel Amar and affiliates, Citadel Advisors with Kenneth Griffin at 4.9% as at May 5, 2026, and Eleven Ventures, which reported 1,053,536 shares at June 30, 2025 and nil from July 7, 2025. This bears on the governance and key-person item, where a register described as one holder understated both the institutional presence and the turnover in it.
No catalyst changed state. The filing record was swept to today by issuer central index key rather than by ticker, which matters for a company that changed symbol in March 2025, and nothing has been filed since the shelf registration of July 17, 2026. The 3 August session has closed, so the file is repriced: the basis moves July 31, 2026 to August 3, 2026 and the close $3.68 to $3.85, up 4.62% on the day. The horizon band boundaries do not move, being a convention this document adopted for dividing near from far rather than a measurement, and the as-of stamp was already 3 August, so no figure computed from it changed.
An unattributed, undated price target. The Cerebras colocation agreement executed (R3) read that "a covering broker raised its target to $10 from $7", naming neither the firm nor the date. A target is a claim about a moment and market data only where both are given. The firm is Alliance Global Partners, analyst Brian Kinstlinger, and the raise was made on May 6, 2026, with a further raise to $12 on May 15, 2026. Both are now stated on this document's own evidence rather than taken from the companion report.
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. No catalyst moved. Counts are unchanged at 15 dated, 6 standing and 6 resolved.
Backlog closed. Seven figures owed to the record are restored to the body rather than to a quotation: the prior-year comparative of $1,633,261 in the Q4 and FY2026 results (D10); the four insider sale prices of $2.72, $2.25, $2.26 and $3.56 in governance, key-person control and promotional communications (S5); the $6.79M revenue denominator and the 36 cents in every revenue dollar it produces, in the annual general meeting and proxy statement (D8); the 42% odds of a September hold in the Federal Reserve policy path (S4); and the 10M to 19M share span in the dilution and capital overlay, recomputed at the 31 July close. Both remaining deferrals on those figures are removed. The note recording that the earlier version of this document survives only as a text extraction is kept, since that remains true.
Analysis updated. The annual general meeting and proxy statement (D8) now carries the reason no proxy exists: the annual report said Part III would come from a proxy filed within 120 days of the year end, and the company instead supplied Part III by amendment on April 27, 2026, which is what an issuer does when the proxy will not arrive in that window. The prior meeting cycle ran to December 2025. A June 2026 meeting is therefore doubtful and a later 2026 meeting more likely, though the item stays unresolved until a filing settles it.
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 below 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.
One correction to a claim that was wrong. The insider buying claim was removed from governance, key-person control and promotional communications. The card read "Amar has been a net buyer historically." The Form 4 record shows the opposite: President and director Alec Amar made six transactions in roughly eighteen months, all sales, none purchases, netting about 137,500 shares, with no insider buying at any point, including through the drawdown from the $9.20 high. Replaced with the actual record. Forms 3 and 4 were never examined in the original build.
Share count, dilution and pricing. Share count 90,420,824 → 98,543,355, being the Form S-3 July 2 figure; cumulative dilution since Dec 2024 restated +174% → +199%. The 8.1M-share increase surfaced only on the cover of a resale registration filed for an unrelated purpose. Independently confirmed by market data showing 98.54M outstanding. 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; the range is deliberate 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. Added the separate C$9.84 option grant and 1,730,000 RSUs vesting 2027–29. Repriced $3.98 → $3.68, the Friday July 31 close. Market cap ~$360M → ~$365M; EV ~$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, an addition error of mine corrected along the way.
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. US Data Centers structure clarified: a Form 8-K dated March 16, 2026 sets out the 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. Vestberg is a co-founder of that entity, not merely a senior advisor. A CTO appointment was also added. The annual general meeting item was reclassified from "not yet announced" to unresolved: a third-party source lists a June 15, 2026 annual meeting, but no proxy and no 8-K under Item 5.07 reporting voting results exist. 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.
Macro, sector and operations. Fed 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. The card previously framed rate risk around whether easing would arrive; that framing was wrong and has been rewritten. Sector deterioration quantified: 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 financing downside. Columbiana is a 55 MW build against a 40 MW Cerebras contract, implying ~15 MW of flagship capacity under construction but uncommitted, the cheapest possible home for a second customer; portfolio-wide secured power of ~400 MW noted. Longer-dated guidance recorded: management targets $450–500M in 2028 and $800M–$1B in 2029, recorded for completeness and not treated as catalysts.
Calendar constructed from the Q1 2026 Form 10-Q (May 15), the FY2025 Form 10-K (March 31, amended April 27), 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, and a market capitalization of roughly $360M. Structure: 15 dated catalysts across three horizons, 6 standing conditions, 6 completed events, plus dilution overlay, null categories, summary and methodology.