WYFI

A $923.7 million backlog resting on one site whose billing start remains unconfirmed

WYFI · WhiteFiber, Inc. · Equity Research Report · as of August 18, 2026

Price $25.95 Market
Market cap $1,008M Estimate
Enterprise value $1.18B Estimate
Shares out 38,848,118 Filed
Pricing basis Aug 3, 2026 close Market

Framing

  1. At March 31, 2026 remaining performance obligations were $923.7M, of which $921.0M is colocation traced to a single ten-year agreement with one privately held counterparty, Nscale, at one site. Filed
  2. That site was not earning revenue at the last confirmed disclosure. Billing was originally targeted for April 30, 2026; the May 14, 2026 quarterly report cited a switchgear supply problem and revised this to delivery beginning during Q2 2026 with full revenue contribution in Q3 2026. Filed
  3. Capital expenditure was $344.7M across the first half of 2026, roughly $175M of it in Q2, against total cash and restricted cash of $60.4M at June 30, 2026. The gap is bridged with convertible notes, project debt, and a facility from the majority shareholder. Filed

01Snapshot

Key figures, each tagged by provenance.

MetricValueBasis
Revenue (FY2025)$79.2MFiled 10-K, Mar 26, 2026
Revenue (Q2 2026)$28.8MFiled 10-Q, Aug 12, 2026. $16.5M excluding a $12.3M one-off termination fee; Q1 2026 was $21.9M
Revenue (TTM)$94.6MEstimate FY2025 less H1 2025 plus H1 2026. $82.3M excluding the termination fee
Net loss (FY2025)($24.68M)Filed 10-K, Mar 26, 2026
Net loss (Q2 2026)($14.98M)Filed 10-Q, Aug 12, 2026. $(0.39) per share against consensus of $(0.41). Q1 2026 was $(12.04M)
Adjusted EBITDA (FY2025)$17.3MFiled 10-K, Mar 26, 2026
Cash and restricted cash$60.4MFiled 10-Q, Jun 30, 2026. $56.1M unrestricted. Working capital is a $28.6M deficit, against $85.2M of working capital at Dec 31, 2025
Remaining performance obligations$923.7MFiled 10-Q, Mar 31, 2026
Construction in progress$294.3MFiled 10-Q, Mar 31, 2026
Shares outstanding38,848,118Filed 10-Q cover, as of Aug 10, 2026. From 38,841,201 at Jun 30 and 38,614,216 at May 12
Free float11.1MMarket Jul 31, 2026
Price$25.95Market Aug 3, 2026 close
52-week range$10.51–46.87Market Jul 31, 2026

The equity is priced against two figures rather than on either. FY2025 revenue of $79.2M came from roughly 11 MW of energised capacity. Against that sits $923.7M of contracted future revenue dependent on capacity still being commissioned at the last confirmed disclosure. A market capitalization near $1,008M sits above both, 9% above the second.

Price behavior follows from the float. On July 31, 2026 the stock opened at $27.06, reached $27.63, and closed at $23.70, down 9.71% on the day, with no company announcement. Bit Digital's majority holding leaves a free float near 11.1M shares against 38.6M outstanding, so roughly $288M of stock is available to trade against a $1,008M capitalization. Market

02Business model

What the company sells, to whom, and how it is paid.

WhiteFiber provides high-performance computing data centers and cloud-based GPU services for AI and machine-learning developers. Two distinct businesses sit in one company. Filed 10-Q, May 14, 2026, Note 1.

Colocation

WhiteFiber owns or leases buildings, provisions power, cooling and network, and lets the space to customers who bring their own hardware. Customers pay a recurring monthly fee for space and power. Electricity and property taxes are largely passed through, so the contracts do not carry a power-price position.

The distinguishing tactic is retrofit: acquiring industrial buildings with existing grid connections and converting them rather than building greenfield. The company puts this at roughly $8–10M per gross megawatt and up to 40% cheaper than greenfield. MTL-3 in Quebec was converted from a factory to an operating Tier 3-equivalent facility in about six months.

Cloud

WhiteFiber also owns NVIDIA GPUs and rents capacity by the day. The GPUs are depreciated over five years. During Q1 2026 the company sold 126 H200 GPUs for $26.1M against a $24.3M book value.

Analyst assumption

Management frames the pairing as vertical integration: owned buildings lower the cloud cost base, and cloud workloads absorb space not yet let to third parties. The trade is that one small balance sheet underwrites two risk profiles, long contracted cashflows and short-cycle hardware rental.

Carve-out and the continuing parent relationship

WhiteFiber was carved out of Bit Digital, Inc. (Nasdaq: BTBT), formerly a bitcoin miner and now an Ethereum-focused holding company. Bit Digital contributed its HPC business for 27,043,749 shares immediately before the IPO. WhiteFiber listed on August 7, 2025 at $17.00, raising about $183M gross including the over-allotment exercised on September 2, 2025. Filed Form 8-K exhibit, August 8, 2025.

The parent filed a Schedule 13D on August 8, 2025, and it establishes the opening position. Bit Digital reported 27,043,750 ordinary shares, with sole voting and sole dispositive power over all of them, being approximately 74.3% of the shares outstanding and 74.3% of total voting power immediately after the offering, assuming the over-allotment went unexercised. The count reconciles against the contribution: 27,043,749 shares issued for the HPC business, plus the single subscriber share on incorporation. Both companies are Cayman Islands exempted companies. Filed Schedule 13D, August 8, 2025

The stake has fallen 4.2 percentage points in nine months, from 74.3% at the offering to 70.1% at the May report, and it has fallen through issuance rather than any disclosed sale: no amendment to that Schedule 13D has been filed. Control is not in question at either figure, and the direction is what matters, since each further issuance dilutes the parent alongside everyone else.

Bit Digital held approximately 70.1% as of the May 2026 10-Q, and since May 27, 2026 is also a lender. The relationship is analytically load-bearing rather than historical: it sets the terms of the bridge financing described in section 7, the governance position described in section 10, and the shared corporate functions provided under the Transition Services Agreement.

The company is incorporated in the Cayman Islands, headquartered in New York, and files as a smaller reporting company and emerging growth company, which permits reduced disclosure relative to a larger filer. Financial statements for periods before August 6, 2025 are carve-out combined statements carrying allocated Bit Digital expenses that, in the company's words, may not reflect standalone spending.

03Revenue streams

Two reported segments, moving in opposite directions.

Revenue lineQ1 2026Q1 2025ChangeBasis
Cloud services$16.77M$14.84M+13%Filed 10-Q, May 14, 2026
Colocation services$4.77M$1.64M+190%Filed 10-Q, May 14, 2026
Other (lease interest)$0.38M$0.28M+37%Filed 10-Q, May 14, 2026
Total$21.92M$16.77M+31%Filed 10-Q, May 14, 2026

Cloud is the larger line and the weaker one. It fell sequentially from $19.3M in Q4 2025 to $16.8M in Q1 2026 and to $11.5M in Q2 2026 excluding the termination fee, a third consecutive decline and a 31.5% fall in the quarter. Reported cloud revenue of $23.8M is higher only because the $12.3M fee sits inside it. Management attributed the decline to repositioning capacity toward longer-duration enterprise deployments rather than short-term rentals.

Analyst assumption

That explanation is plausible and would represent a quality improvement if correct. It remains a sequential decline in the segment producing three quarters of revenue, and the file has no independent confirmation either way.

Colocation is the smaller line and the one being built. It nearly tripled year over year on the MTL-3 ramp, which began serving Cerebras in Q4 2025 under a 5 MW contract. Colocation also carries the contracted backlog.

Backlog and what it excludes

At March 31, 2026 remaining performance obligations were $923.7M, scheduled for recognition as $51.3M in 2026, $90.7M in 2027, $91.8M in 2028, $93.5M in 2029, $93.5M in 2030 and $502.8M thereafter. Filed 10-Q, May 14, 2026.

Analyst assumption

Two cautions. First, $921.0M of the $923.7M is colocation, so the cloud business contributes almost nothing to contracted backlog because its variable pricing falls outside the disclosure requirement. Second, backlog is a customer commitment rather than cash: it converts only if the building is finished and the counterparty pays across ten years.

Deferred revenue and contract liabilities rose to $144.5M from $79.6M at year end as customers prepaid. That funding is why operating cash flow was positive in a quarter carrying a $12.0M net loss. Filed 10-Q, May 14, 2026.

04Industry & market backdrop

The neocloud category, and the two cracks that opened in it this year.

WhiteFiber sits in the category the market calls neoclouds: companies renting AI compute and the buildings housing it as an alternative to Amazon, Microsoft and Google. The category barely existed before 2024. Synergy Research has forecast the neocloud market approaching $400bn by 2031, naming CoreWeave, Crusoe, Core Scientific, Lambda, Nebius and Nscale as leaders.

Structural demand is real. AI training and inference need power density and cooling that conventional data centers were not built for, and grid connections have become the binding constraint. Control of 54 delivered megawatts is a scarce position.

Crack one: the largest owners are becoming sellers

On July 1, 2026, reporting that Meta was exploring sales of its own surplus GPU capacity under a Meta Compute initiative removed roughly 13–15% of the market value of CoreWeave and Nebius in a single session. Hyperscalers build for peak training demand, hold idle capacity between runs, and carry none of the GPU-collateralised debt the neoclouds rely on. Market

Meta is no longer the only entrant. SpaceX has moved into the market and has reportedly taken capacity deals from Google, Reflection AI and Anthropic. Two well-capitalized entrants in one year moves the question from whether neoclouds can take share from hyperscalers to what happens when the largest infrastructure owners sell capacity directly. Market

Crack two: credit for AI infrastructure has repriced

Sector analysts had flagged tens of billions of dollars of predominantly GPU-collateralised debt across CoreWeave, Nebius, Lambda, Crusoe and Applied Digital maturing between 2026 and 2028, concentrated among a small and correlated lender group. In late July that concern appeared in prices. On July 29, 2026 CoreWeave was reported to be seeking to raise the interest rate on a $2.6bn loan tied to capacity expansion for Anthropic. Its 2032 bond has been yielding around 10.32% at a B rating from S&P. Credit default swap costs for technology borrowers have widened. From their peaks, CoreWeave is down roughly 61%, IREN roughly 47% and Nebius roughly 37%. Market

WhiteFiber's own debt does not mature in the 2026–2028 window, since the convertible runs to 2031. The exposure is immediate anyway: the company is still trying to place permanent project financing for NC-1 into a market that is actively repricing this risk. That is a worse environment than the one in which the facility was originally expected to close. Estimate

Analyst assumption

A useful frame for the sector: the market is pricing contracted capacity, but little of it is switched on. Across the larger names, published activation rates, meaning live capacity as a share of contracted power, have been reported between roughly 4% and 29%. The gap between a signed megawatt and a billing megawatt is where these companies succeed or fail.

05Competitive position

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

PeerComparable metricBasis
CoreWeave (CRWV)Q1 2026 revenue $2.08bn; revenue backlog $99.4bn; more than 1 GW active powerMarket Q1–Q3 2026
Nebius (NBIS)Contracted power above 3.5 GW in Q1 2026; owned power and in-house software layerMarket Q1 2026
WhiteFiber (WYFI)Q1 2026 revenue $21.9M; backlog $923.7M; 54 gross MW delivered at NC-1Filed 10-Q, May 14, 2026
IREN, TeraWulf, Cipher, Applied Digital, Digi Power X, SolunaPower-site converters, several former bitcoin miners; the direct retrofit peer setMarket Q1–Q3 2026

CoreWeave is the reference point for the category and WhiteFiber's quarterly revenue is roughly 1% of it. They do not compete for the same contracts. The retrofit peer set is where WhiteFiber competes directly.

Nscale is a particular case: a European AI infrastructure company that is simultaneously WhiteFiber's largest customer and a named leader in the same category. WhiteFiber lets building space to a competitor's compute business.

Meta and SpaceX are the newest entrants. Neither needs to win colocation contracts to affect the sector; both can sell surplus capacity they already own at whatever price clears.

Analyst assumption

Colocation is partly insulated from that, since a signed ten-year lease on a physical building is harder to displace than a GPU-hour rental. The cloud segment, still roughly three quarters of revenue, competes directly on price for the same workloads.

Where WhiteFiber is differentiated

Retrofit speed is the most credible operational claim: six months from industrial building to operating Tier 3-equivalent facility at MTL-3. Most peers operate buildings or compute; WhiteFiber does both, which is an integration advantage or a focus problem depending on execution. On July 9, 2026 the company reported R&D results for linking two separated data centers into one logical GPU cluster: 111.2 Tbps across 83 km of dark fibre at 0.9 ms guaranteed round-trip latency, with commercial launch targeted for Q3 2026. Filed press release, July 9, 2026. No customer, pricing or contract has been announced. Open

Where it is disadvantaged

Scale and cost of capital. Equity is under $1bn, the convertible carries a 4.50% coupon, and bridge financing from the parent carries 9.5% plus fees. Competitors an order of magnitude larger fund the same megawatts more cheaply, and megawatts are a commodity once energised.

06Financial performance

Multi-period, with the trend stated.

MeasureFY2024FY2025Q1 2025Q1 2026Basis
Revenue~$48M$79.2M$16.77M$21.92MFiled 10-K Mar 26, 2026; 10-Q May 14, 2026
Operating profit+$0.6M−$26.8M+$2.04M−$11.02MFiled 10-K; 10-Q
Net profit+$1.37M−$24.68M+$1.43M−$12.04MFiled 10-K; 10-Q
EPS (basic)+$0.05−$0.78+$0.05−$0.31Filed 10-K; 10-Q
Adjusted EBITDAn/a$17.3M$6.0M$3.0MFiled 10-K; 10-Q
Analyst assumption

FY2024 revenue is derived from the disclosed 65% year-on-year growth rate rather than stated directly in the sources reviewed, and is approximate.

The trend is revenue growth against a widening loss. Revenue grew 31% year over year in Q1 2026 while total operating expenses more than doubled to $32.9M. Depreciation rose to $6.4M from $3.8M as assets came into service, and general and administrative expense rose to $17.8M from $4.2M.

Analyst assumption

The G&A step is the most striking figure in the quarter. It includes $7.3M of share-based compensation, and the company was a carve-out subsidiary a year earlier without public-company equity plans, so the comparison is not like-for-like. Excluding that item, G&A still roughly doubled. Whether this is the fixed cost of public-company operation or weaker cost control is not settleable from the filings to hand.

Gross margin excluding depreciation held at roughly 60.2% against 60.5% a year earlier. Unit economics on the underlying services are not the issue. Filed 10-Q, May 14, 2026.

Cash

Q1 2026 cash flowAmountBasis
Operating activities+$3.2MFiled 10-Q, May 14, 2026
Capital expenditure and deposits−$169.2MFiled 10-Q, May 14, 2026
Proceeds from GPU disposals+$26.1MFiled 10-Q, May 14, 2026
Net convertible note proceeds+$222.1MFiled 10-Q, May 14, 2026
Zero-strike call purchase−$120.0MFiled 10-Q, May 14, 2026
Net change in cash−$38.0MFiled 10-Q, May 14, 2026

Cash and restricted cash fell to $80.1M from $118.3M despite raising $230M of convertible notes in the same quarter. Third-party data services have put trailing-twelve-month free cash flow between negative $220M and negative $245M. Market

The operating cash flow figure repays scrutiny. The $3.2M inflow was produced by $65.0M of customer prepayments offset by a $67.8M build in accounts receivable. Receivables rose from $23.9M to $91.7M, roughly four times a full quarter of revenue, and converted in Q2: accounts receivable stood at $23.2M at June 30, 2026 and six-month operating cash flow was positive $89.1M. Filed 10-Q, May 14, 2026.

The likeliest explanation is milestone and installation billing on NC-1 rather than a collection problem, but it is a large working-capital swing on a small balance sheet. Estimate

Balance sheet at March 31, 2026

LineValueBasis
Total assets$796.3MFiled 10-Q; $651.4M at year end
Property, plant and equipment, net$432.0MFiled 10-Q, May 14, 2026
Of which construction in progress$294.3MFiled Not in service, not earning
Total liabilities$443.7MFiled 10-Q, May 14, 2026
Convertible notes, net of issuance costs$222.3MFiled 10-Q, May 14, 2026
Shareholders' equity$352.6MFiled 10-Q, May 14, 2026
Additional paid-in capital$388.9MFiled Fell from $504.7M

The fall in additional paid-in capital is the accounting effect of the $120.0M zero-strike call purchase rather than an operating loss.

07Capital structure & dilution

Shares, convertibles, the equity plan, and the controlling holding.

InstrumentSizeTerms and basis
Ordinary shares outstanding38,848,118Filed 10-Q cover, as of Aug 10, 2026
Bit Digital holding~70.1%Filed 10-Q, May 14, 2026. 27,043,749 shares issued at carve-out
Free float11.1MMarket Jul 31, 2026
Convertible notes due 2031$230.0MFiled 4.50%, priced Jan 26, 2026, conversion price $25.91
Implied conversion shares~8.9MEstimate Roughly 23% of the current count
Zero-strike call optionn/aFiled Purchased for ~$120.0M; lifts the effective conversion price to approximately $37
Iceland term loan$20MFiled Secured, March 2026
Bit Digital delayed-draw facility$100MFiled 8-K, May 27, 2026. Expandable to $150M

The convertible sits above the current price, so it is not near-term dilutive. That also means it currently functions as straight debt with a 2031 maturity rather than as equity in waiting.

Terms on the parent facility are set out in section 10, since their significance is a governance and cost-of-capital question rather than a share-count one. The facility is debt and does not dilute the share count, but a 9.5% coupon with a minimum 1.1x multiple-on-invested-capital repayment is a drag on equity value whenever drawn.

Analyst assumption

The dilution that matters most is not yet on the table. Because the stock has roughly halved since late June, the number of shares required to raise any given sum has roughly doubled. Dilution risk is inversely related to how well the near-term operational milestones resolve. Separately, a 70.1% holding is a standing overhang: a registered secondary is a distinct event from a primary raise, and no such transaction has been announced. Open

08Valuation

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

MeasureValueBasis
Market capitalization$1,008MEstimate Aug 3, 2026 close, $25.95 × 38.85M shares filed as of Aug 10, 2026
Plus convertible notes, face+$230MFiled 4.50% due 2031
Plus Iceland term loan+$20MFiled Secured, Mar 2026
Less cash and restricted cash−$80MFiled At Mar 31, 2026
Approximate enterprise value$1.18BEstimate Understated: it predates the Q2 and July drawdowns now filed, which include CAD 36.8M on the syndicated RBC facility and $30M on the Bit Digital Capital facility
EV / TTM revenue ($94.6M)12.5xEstimate Trailing, on the enterprise value above. 14.3x on the $82.3M that excludes the Q2 termination fee, which is the multiple the recurring business carries
EV / FY2025 adj. EBITDA ($17.3M)68xEstimate Trailing

Three warnings attach to that table. The cash figure is four months old and the company was consuming cash heavily, so any Q2 draw on the RBC, Iceland or Bit Digital facilities raises enterprise value further. The share count is now seven days newer than the price rather than eleven weeks older: 38,848,118 was filed as of August 10, 2026 and the close is 3 August, so any shares issued between those dates sit outside the count and the market capitalization is understated to that extent. Whether any were issued is not established here, because the equity incentive plan's authorised pool is not disclosed in this document's evidence and share buybacks were not examined. A trailing EBITDA multiple of 68x on a company mid-construction is not a signal in either direction.

The forward case

Analyst assumption

The forward argument runs through the backlog schedule, which implies roughly $90–93M per year of colocation revenue from 2027 through 2030. Adding the Q1 cloud run-rate of about $67M annualised, and the Paris contract at roughly $32M a year if the $160M five-year value is spread evenly, gives 2027 revenue in the region of $170–195M, around 2.2x trailing. On that revenue the enterprise value is roughly 6.0–6.9x sales.

Every forward case assumes NC-1 permanent financing closes at a workable cost. With AI-infrastructure credit repricing, the spread eventually paid is a live variable rather than a rounding error. Interest expense passes straight to the equity, so a wide spread compresses the same forward figures the case depends on. Estimate

Sell-side estimates published this year have been more aggressive. One March 2026 note argued for roughly 5.3x FY2026 EV/EBITDA against peers above 20x, forecasting $139M annualised revenue and $105M EBITDA by end-2026. Market

Analyst assumption

That forecast implies an EBITDA margin near 75%, above anything the company has reported, and assumes NC-1 reaches full contractual capacity on schedule. It is a scenario rather than a base case.

Third-party price targets

The following are the published views of the firms named and are reported here as market data. This document adopts none of them and derives no target of its own.

FirmAnalystDateRatingTargetPrice at publicationBasis and disclosed conflict
BarclaysRaimo LenschowJul 21, 2026Equal Weight$29$27.13Market Raised Jul 21, 2026. Price at publication is the close on July 20, 2026, the last session before the action. The firm did not underwrite the offering of August 8, 2025 and no other role for it appears in the company's filings. What it discloses in its own note is not established.
BTIGGregory LewisJun 24, 2026Buy$50$44.76Market Raised Jun 24, 2026. Price at publication is the close on June 23, 2026. The firm did not underwrite the offering of August 8, 2025 and no other role for it appears in the company's filings. What it discloses in its own note is not established.
NeedhamNot establishedMay 15, 2026Buy$38$29.99Market May 15, 2026. Price at publication is the close on May 14, 2026. Needham & Company, LLC underwrote 2,631,563 of the 9,375,000 shares sold in the offering of August 8, 2025, the second-largest allocation in the syndicate. A data service attributes a $38 Buy to B. Riley Securities dated May 18, 2026 rather than to this firm on 15 May; both underwrote that offering, and which of them struck this target is not established.
Cantor FitzgeraldBrett KnoblauchMay 15, 2026Neutral$27$29.99Market May 15, 2026. Price at publication is the close on May 14, 2026. The firm did not underwrite the offering of August 8, 2025 and no other role for it appears in the company's filings. What it discloses in its own note is not established.
H.C. WainwrightNot establishedMar 27, 2026n/a$20$12.16Market Cut from $34 on Mar 27, 2026. Price at publication is the close on March 26, 2026. The firm did not underwrite the offering of August 8, 2025 and no other role for it appears in the company's filings. What it discloses in its own note is not established.

Price at publication is the close on the last trading day before the action, which is what the author could see when the target was struck. Read down that column rather than across the targets: the same five figures were struck against prices from $12.16 to $44.76, so a $20 target set on 27 March was a 64% premium to the market of the day while a $50 target set on 24 June was a 12% premium. Against the 3 August close of $25.95 the range runs from 23% below to 93% above, and the two firms furthest apart on target were the two closest to the market when they wrote.

An aggregator publishes a consensus of $33–36 as reported. The figure carries no publication date, so it cannot be placed against a session and no price at publication attaches to it. A service read on August 4, 2026 publishes a different aggregate, $38.11 across nine analysts and dated July 21, 2026. Neither is a center struck in this document. Market

One of the five underwrote the company's own offering. Needham & Company took 2,631,563 of the 9,375,000 shares sold on August 8, 2025, behind B. Riley Securities on 4,785,000, with Macquarie Capital, Roth Capital Partners, Craig-Hallum and Clear Street completing the syndicate. Barclays, BTIG, Cantor Fitzgerald and H.C. Wainwright do not appear in it. Stated as a distribution rather than as a superlative: the one disclosed underwriting relationship sits in the middle of the target range, not at either end, so it does not sort the views. What each firm discloses in its own note is not established, no note having been read.

The count above is a floor rather than the whole coverage set. A data service read on August 4, 2026 carries 2026 actions from two further firms, B. Riley Securities and Craig-Hallum, both of them underwriters of the same offering, and the company's own statement of who follows it was not read. Two services also disagree about the aggregate, at $33 to $36 undated and $38.11 across nine analysts dated July 21, 2026. Both are reported as those services published them and neither is averaged with the other: where providers disagree about the figure and about how many contributors stand behind it, no meaningful consensus is available, and this document computes no center of its own and derives no target.

B. Riley Securities was a joint bookrunner on the IPO, publishes research on the stock, and in May 2026 acquired a $20M portion of the related-party term loan to WhiteFiber's subsidiary. Published research on a company of this size is not necessarily independent of its financing relationships. Filed 8-K, May 27, 2026.

09Growth drivers

What would have to happen, and what is already contracted.

The capacity ladder

Megawatts are the operative unit. Three figures set the frame, and they are not equivalent in standing.

RungCapacityStandingBasis
Energised and earning, end 2025~11 MWProduced the $79.2M of FY2025 revenueFiled 10-K, Mar 26, 2026
Company target, gross, end 2026~76 MWA stated target, not a commitmentFiled 10-K, Mar 26, 2026
Pipeline under review~1,500 MWSites being evaluated. No contracts, financing or power agreements attachedOpen
Analyst assumption

The distance from the first rung to the second is the investment case. The distance from the second to the third is the story supporting the multiple. Conflating the three is the most common error in reading this company.

Contracted and announced drivers

NC-1 reaching full contracted capacity. The 40 MW Nscale deployment at Madison, North Carolina is the dominant driver: approximately $865M over ten years with 3% annual escalators, structured as a modified gross lease with electricity passed through. Duke Energy has delivered the initial 54 gross MW. This item drives essentially all of the backlog. Filed 8-K, December 18, 2025.

NC-1 expansion beyond 40 MW. The site carries a 99 MW capacity agreement with Duke Energy, about one million leasable square feet, and management believes it could eventually support up to 200 MW. Nscale holds priority notification rights on new capacity, and the December 2025 announcement referenced potential expansion toward roughly double the initial deployment by end-2027. Filed 8-K, December 18, 2025.

The Paris contract. A five-year agreement announced May 21, 2026 with an investment-grade technology customer, over $160M total value, using NVIDIA systems in third-party French capacity, with service expected to commence July 2026 and supported by twelve months of customer prepayments. Filed 8-K, May 21, 2026. No confirmation that service has commenced or that the associated project financing closed as expected in June 2026. Open

MTL-3 ownership. The purchase option was exercised on January 14, 2026 and completed on May 8, 2026, converting a leased revenue-generating asset into an owned one with expansion optionality. Filed 10-Q, May 14, 2026, Note 6.

Cerebras. MTL-3 supports Cerebras wafer-scale systems under a 5 MW contract and completed its first full quarter of operations in Q1 2026. Filed 10-Q, May 14, 2026.

Project Redwood. The cross-data-center networking architecture, if commercialised in Q3 2026 as targeted, would be a differentiated product. Commercially unproven. Open

10Risks

Severity-ranked, most severe first.

Customer concentration is close to total Severe

Roughly $921M of $923.7M in backlog rests on one agreement with Nscale, a privately held company whose balance sheet investors cannot examine. WhiteFiber describes the arrangement as backed by an investment-grade hyperscaler offtake, but that end customer is not named in the disclosures reviewed. The credit chain runs through a party investors cannot see to another they are not told the identity of. A default, delay or renegotiation changes the character of the backlog figure entirely. Filed 10-Q, May 14, 2026.

The majority owner is also the lender Severe

On May 27, 2026 WhiteFiber announced a $100M delayed-draw term loan facility, expandable to $150M, from Bit Digital Capital, a subsidiary of the 70.1% shareholder. Per Bit Digital's own 8-K the terms are 9.5% annual interest, stepping to 8.0% only once NC-1 Phase I is substantially complete and at least 80% leased at market rates; a 3% original issue discount; a 0.50% commitment fee on undrawn capacity; and a minimum 1.1x multiple-on-invested-capital repayment per advance at maturity. It is guaranteed by White Fiber Operating Partnership LP and secured by equity in the NC-1 holding company. Filed Bit Digital 8-K, May 20, 2026.

Analyst assumption

Those are bridge terms negotiated between affiliated parties rather than arm's-length bank pricing. Bit Digital funded its side by drawing $50M from Galaxy Digital on May 20, 2026 under a crypto-collateralised loan, so WhiteFiber's bridge liquidity sits one step from digital-asset collateral values. A minority holder has limited ability to influence any of it.

Permanent NC-1 financing has not closed Severe

The December 2025 Nscale announcement said the company was in advanced discussions with lenders and expected to formalise a construction credit facility in early Q1 2026. By late May 2026 it described itself as still advancing non-dilutive permanent financing solutions while using a related-party bridge. No announcement of a closed permanent facility was found as of August 3, 2026, a slipped milestone of roughly two quarters on the most important financing in the company. Open

Schedule slippage at NC-1 Severe

Billing on the first 20 MW was targeted for April 30, 2026 and the second for May 30, 2026. The May 14, 2026 report disclosed a medium-voltage switchgear supply issue and revised this to delivery beginning during Q2 with full revenue contribution in Q3. No public confirmation that billing has commenced was found. The Q2 report, expected around 9–August 13, 2026, is where this settles. Open

Cash consumption against available liquidity Moderate

Roughly $175M of quarterly capital expenditure against $60.4M of cash and restricted cash, and a working capital deficit of $28.6M. The company depends on continued access to capital markets to finish what it has started. There is no going-concern warning in the filings, and contracted backlog and prepayments genuinely help, but the margin for error is thin. Filed 10-Q, May 14, 2026.

Financing-market conditions have tightened Moderate

The credit market for AI infrastructure repriced in late July 2026, as set out in section 4. This is the market into which the company must sell NC-1 project debt. The risk is no longer only that the facility is late: the terms available now may be worse than those contemplated when the deal was announced, and continued delay accrues 9.5% plus a 1.1x repayment multiple on the related-party bridge. Estimate

Dilution Moderate

The $230M convertible notes carry an initial conversion price of $25.91, four cents below the current price, and they imply roughly 8.9M shares, about 23% of the current count, if converted. The zero-strike call structure lifts the effective conversion price to approximately $37, which reduces without eliminating the exposure. Filed 10-Q, May 14, 2026.

The cloud fleet is a depreciating asset Moderate

GPUs are depreciated over five years while the hardware generation cycle is shorter. The segment's revenue declined sequentially in Q1 2026. Selling 126 H200 GPUs mid-quarter at a small gain is prudent asset management and also a reminder that this half of the business is a fleet requiring continual capital refresh. Filed 10-Q, May 14, 2026.

Insider reporting was delinquent Moderate

On July 28, 2026 WhiteFiber filed four Form 4 insider reports. Two of them, for CEO Samir Tabar (41,982 RSUs) and CFO and director Erke Huang (41,982 RSUs), report transactions dated March 31, 2026. Exchange Act Section 16(a) requires a Form 4 within two business days. These were filed roughly four months late. Filed Forms 4, July 28, 2026.

Both officers named in that delinquency have since changed roles. On July 30, 2026, effective 1 August, Erke Huang resigned as chief financial officer, principal financial and accounting officer and director, moving to senior adviser and non-voting board observer while remaining chief financial officer of the parent. Justin Zhu succeeds him, having resigned his own Bit Digital posts, where he had been vice president of finance since 2021 and, from July 2025, principal financial officer and chief accounting officer. His terms are $450,000 a year to July 31, 2028, restricted units worth $100,000 on signing, and three months' salary plus a pro-rated bonus on termination without cause or on a change of control. Filed Form 8-K, August 3, 2026

The board turned over at the same time, and the committee arithmetic is worth watching. David Andre resigned from the board and from the audit, compensation and nominating committees. Samir Tabar, chief executive since February 2025 and previously chief executive of the parent, joined the board. Michael Rulf joined at $150,000 a year for the board and the compensation and nominating committees, not the audit committee; separately his own company holds a professional services agreement with the issuer, terminable on thirty days' notice. Pruitt Hall continues on all three committees, his fee rising from $150,000 to $270,000, and his professional services agreement was terminated in connection with continued audit committee membership, the board confirming he meets the Nasdaq listing standard and the heightened audit-committee independence test. The pattern is consistent: the director with a consulting arrangement sits off the audit committee and the director on it gave his up. What the filing does not name is a third audit-committee member to replace Mr Andre, and the exchange requires three.

Analyst assumption

Nothing here indicates wrongdoing: these were non-discretionary RSU vestings under an equity plan rather than open-market trades, and the economic event was unchanged by the reporting date. A four-month delay at a company less than a year into public life is nonetheless an internal-controls signal. Delinquent filings of this kind must be disclosed under Regulation S-K Item 405 in the annual report or proxy, so the matter should surface again in the FY2026 10-K. It sits against an otherwise clean record: both periodic reports were filed on time, with no NT 10-K or NT 10-Q notification anywhere in the complete index of 73 filings, searched August 3, 2026.

The parent filed a comparable batch the same evening. Sam Tabar is CEO of both companies and Erke Huang is CFO of both. A shared compliance function is the likeliest explanation, consistent with the Transition Services Agreement under which Bit Digital still provides certain corporate functions, and it means a lapse at the parent propagates directly into WhiteFiber's filing record.

Sector and market risk Moderate

Neocloud multiples moved violently on the Meta Compute reports of July 1, 2026 and have not settled. WhiteFiber shares fell from roughly $46 in late June to $23.70 on July 31, 2026, a decline of roughly half in five weeks, with no company-specific announcement identified. Short interest was reported at a record high in May 2026, and the stock appeared on lists of the largest borrow-rate increases. Market A heavily shorted stock with an 11.1M share float and binary near-term catalysts will move more than the underlying business changes, in both directions.

Energy prices, inflation and the rate path Moderate

US electricity prices rose about 6.3% in the twelve months to January 2026, more than double headline inflation at the time. Colocation contracts pass electricity through, so that segment is largely insulated; the cloud segment is not, since electricity is a direct cost of revenue there and cloud remains roughly three quarters of revenue. Market

At the July 29, 2026 meeting the Federal Reserve held its target range at 3.50–3.75% on a 9–3 vote, with three members dissenting in favor of a hike. Market pricing has moved to rate increases rather than cuts in 2026, driven by energy-led inflation associated with the US–Iran conflict. Market

Analyst assumption

For a company whose value rests on contracted cashflows starting in 2027 and which must raise project debt now, a rising-rate path is a double charge: a higher discount rate on future revenue and a higher coupon on the financing required to deliver it. Published estimates of current crude prices conflict, some placing Brent below $75 and others above $100, so the energy input is genuinely uncertain rather than settled.

Other disclosed risks Lower

The company's forward-looking risk list flags GPU supply timing, integration of Enovum, executive-team dependence, power price volatility, tariffs and export restrictions particularly concerning Canada, AI-specific regulation, and its limited history as an independent public company. Filed 10-Q, May 14, 2026.

11Bull / base / bear

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

Bear

Commissioning problems push billing into late 2026. The permanent facility either does not close or closes on terms transferring most of the project economics to lenders, a path that became more plausible in late July when AI-infrastructure credit repriced and a much larger peer was reported to be raising the rate on its own loan, its junk-rated 2032 bond yielding around 10.32% at a B rating from S&P. The related-party bridge is drawn heavily at 9.5% plus a 1.1x repayment multiple, and the company returns to equity markets at a depressed price. Cloud revenue keeps declining sequentially, so the segment funding overhead shrinks while G&A holds at its new level. In the worst version something goes wrong at Nscale and the $921M backlog stops anchoring the valuation. There is no earnings floor beneath the stock; the 52-week low of $10.51 indicates how far it has traded down before.

Requires: further NC-1 slippage, or a financing close on punitive terms, or a counterparty failure at Nscale.

Base

NC-1 begins billing during Q2 to Q3 2026 with a modest delay against the original target. Colocation revenue steps toward the roughly $90M annual run-rate the backlog schedule implies for 2027. Cloud stabilises near current levels rather than growing. The Paris contract commences and contributes. Permanent project financing closes at a workable spread, replacing the bridge. 2027 revenue lands around $170–195M with adjusted EBITDA margins in the 40–55% range as the colocation mix rises: profitable at the EBITDA line, still loss-making at the net line while depreciation and interest run heavy. The company becomes a small contracted infrastructure business and the equity trades on execution.

Requires: NC-1 billing by Q3 2026, permanent financing closed, cloud flat rather than declining.

Bull

The 40 MW deployment comes online cleanly and Nscale exercises its priority rights toward roughly double the initial size by end-2027, taking NC-1 toward its 99 MW Duke Energy agreement and potentially beyond. The retrofit model, six months from factory to Tier 3 at $8–10M per MW against greenfield costs, proves repeatable at two or three further pipeline sites, funded increasingly by project-level debt and customer prepayments rather than equity. Project Redwood commercialises in Q3 2026 and gives the cloud business a product no one else sells. The company compounds contracted megawatts at a falling cost of capital, and the current enterprise value of roughly $1.18 billion is applied to a materially larger contracted base.

Requires: clean NC-1 commissioning, a Nscale expansion order, repeatable site acquisition, and Redwood reaching commercial launch with a customer.

Analyst assumption

All three cases are constructions from the disclosed facts rather than forecasts, and no probabilities are assigned. With a single dominant contract and an unresolved financing, the outcome distribution is closer to binary than to a smooth curve, and a weighted average would misrepresent it.

12Research summary

WhiteFiber is a company under a year into public life, built from a carve-out, still 70% owned by its former parent, that has signed an $865M ten-year contract and converted industrial buildings into AI-grade capacity faster than most peers, and has not yet been paid for the part that matters.

The comparison that frames everything: $79.2M of FY2025 revenue from roughly 11 MW, against $923.7M of contracted future revenue dependent on capacity still being commissioned at the last confirmed disclosure. A market capitalization near $1,008M sits above both, 9% above the second.

Three questions determine which figure the valuation comes to resemble, and all are answerable within months rather than years: whether NC-1 started billing and when; whether permanent project financing closed and at what cost; and whether the cloud segment stopped shrinking.

The strongest positives are a contracted backlog with disclosed year-by-year recognition, roughly 60% gross margins on services, customer prepayments funding part of the build, an owned flagship site with 99 MW of utility capacity secured, and a retrofit cost structure cheaper and faster than greenfield. The most serious concerns are near-total dependence on one contract with one private counterparty whose own end customer is undisclosed, a permanent financing that was due two quarters ago and has not closed, bridge debt from the controlling shareholder on terms a minority holder had no say in, and a cash burn requiring continuous capital-market access at the moment those markets are repricing AI infrastructure risk upward.

No buy, sell or hold recommendation is issued here, and none should be inferred. What the file supports is a view of which facts are settled, which are targets, and which remain unproven.

What to watch next

WhenWhat
~9–13 AugResolved August 12, 2026. Colocation was flat at $4.7M, NC-1 billing had not commenced, and the $91.7M receivable converted, closing at $23.2M.
Q3 2026Project Redwood commercial launch, targeted. A named customer would be the first validation.
Any 8-KClosing of permanent NC-1 project financing, and the spread at which it prices.
Any 8-KDrawdowns on the Bit Digital facility. Heavy usage would signal permanent financing is still unavailable.
OngoingParis service commencement, and whether the June 2026 project financing closed.
OngoingNscale expansion toward double the initial 40 MW, referenced for end-2027.
27–29 AugJackson Hole symposium. Chair Warsh speaks, watched for signals on the rate path.
15–16 SepFOMC, described as the first meaningful test of whether the hold continues.

13Null categories

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

CategoryStatusBasis
Analyst coverageCovered aboveSection 8 carries published targets from Barclays, BTIG, Needham, Cantor Fitzgerald and H.C. Wainwright, with dates, plus aggregator consensus.
DividendsNot establishedNot examined in this document. Also named in what was not checked.
Share buybacksNot establishedNot examined in this document. Also named in what was not checked.
Securities class actionNot establishedNot examined in this document. Also named in what was not checked.

Each of these three is a gap in this document's evidence rather than a finding that the category is empty. Absence of evidence is not evidence of absence.

14Methodology & sources

Pricing basis

Market figures are struck at the August 3, 2026 close. Balance-sheet figures are as of March 31, 2026 and are four months old; the company was consuming cash rapidly and has facilities available, so the enterprise value shown understates rather than overstates. The as-of stamp is the build date and differs from the pricing basis.

Provenance tags

TagMeaning
FiledStated in an SEC filing or company press release, with form and date cited.
EstimateDerived, modelled or inferred here rather than reported by the company.
OpenExpected or targeted but unconfirmed, with nothing filed either way.
MarketPrice, float, third-party targets and peer market data, stamped with the close date.

Sources

WhiteFiber Form 10-Q for the quarter ended March 31, 2026 (filed May 14, 2026); Form 10-K for the year ended December 31, 2025 (filed March 26, 2026); Forms 8-K dated August 8, 2025, December 18, 2025, May 21, 2026, May 27, 2026 and June 10, 2026; four Forms 4 filed July 28, 2026 (Tabar, Huang); company press releases dated August 8, 2025, December 18, 2025, March 26, 2026, May 14, 2026, May 21, 2026, May 27, 2026 and July 9, 2026; Bit Digital Form 8-K dated May 20, 2026 and press release dated May 27, 2026; Q1 2026 earnings call transcript, May 14, 2026; market data as of the August 3, 2026 close; Federal Reserve FOMC statement and implementation note, July 29, 2026; sector and credit-market reporting from Synergy Research and trade press, Q2 to Q3 2026.

What was not checked

A Form 8-K of August 17, 2026 and the company's announcement the same day of an agreement to acquire two North Carolina development sites were not read here, and nothing in this document is revised for them. Dividends, share buybacks and securities litigation were not examined in the preparation of this document. Their rows in section 13 read Not established for that reason, and nothing here should be read as a finding that those categories are empty.

Peer credit data, including CoreWeave bond yields and the reported loan repricing, comes from secondary market reporting rather than from primary filings reviewed here. Energy-price sources conflict on where crude currently sits. Items tagged Open are targets or expectations for which no confirming public source was found as of August 3, 2026; absence of confirmation is not evidence of failure.

The Nscale end customer is not named in the disclosures reviewed, so the credit quality behind the largest contract cannot be independently assessed here.

Section 7 is thinner than the heading suggests. The equity incentive plan's authorised pool, the share count referenced by the zero-strike call, and the premium at which the convertible priced are not disclosed anywhere in this document's own evidence, so no figure is given for them. The section carries the instruments themselves, their terms where stated, and the resulting exposure, and stops there rather than presenting figures this document did not establish.

Known limitations carried forward

The qualification dropped from the valuation caveat is restored. The caveat below the forward case had lost a supporting parenthetical when the surrounding sentence was recast, leaving the section stating that credit had repriced without the specifics that showed it. The bear case now carries both: the peer raising the rate on its own loan, and its junk-rated 2032 bond yielding around 10.32% at a B rating. The figure is restored at the precision the industry section already establishes rather than at the rounded 10.3% the lost clause used, since the two are one quantity and the file should state it once.

The enterprise value now reads on one basis. The bull case carried roughly $1.06 billion, struck at the 27 July close and left behind when the file was repriced, against approximately $1.09B in the valuation section at the 31 July close. The bull case is restated to $1.09 billion, which is the valuation section's own figure at the pricing basis this document uses throughout. The companion label defect, a sourcing line naming the 27 July close against text priced off 31 July, was corrected earlier and section 8 names 31 July.

Both internal inconsistencies recorded below are now resolved, and the entries are kept so the correction remains auditable against what was previously carried. Both figures in this entry are superseded by the reprice completion of August 18, 2026 recorded in the document log: the enterprise value reads $1.18B on the August 3 capitalization, and this entry describes the file as it stood on the 31 July basis.

Enterprise value in the bull case. Section 11 reads: "the current enterprise value of roughly $1.06 billion is applied to a materially larger contracted base." Section 8 gives an enterprise value of approximately $1.09B at the July 31, 2026 close. The $1.06B figure was struck at the July 27, 2026 close and was not updated when the file was repriced. The text is preserved as written rather than altered at the time. The governing requirement is that every figure carry the pricing date it was struck at, and that a document use one pricing basis throughout.

Valuation source label. The source carried a sourcing label reading "Market data Jul 27, 2026 close" against a valuation section whose text prices off the July 31, 2026 close of $23.70. The masthead, the body and the document log all establish 31 July as the pricing basis, so the label was the stale element and section 8 now names 31 July. The discrepancy is recorded rather than passed over.

15Document log

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

August 20, 2026 Latest
Priced off Aug 3, 2026 close · $25.95 · $1,008M cap · not repriced
Correction

This document cited two Forms 8-K that were never filed. The sources list named Forms 8-K dated March 26, 2026 and May 14, 2026. Those are the dates WhiteFiber filed its Form 10-K and its Form 10-Q, both cited correctly elsewhere in the same list, and both also the dates of a company press release. No Form 8-K was filed on either date. The two dates are removed from the Form 8-K list; the five that remain are as filed, and no other source in the list changes.

August 18, 2026
Priced off Aug 3, 2026 close · $25.95 · $1,008M cap · enterprise value completed to $1.18B
Correction

This report stated two prices for itself, and two capitalizations. The masthead carried $25.95 at the August 3, 2026 close and the snapshot carried $23.70 at the July 31 close; one paragraph put the capitalization near $915M and the next near $1,008M. The reprice of 3 August reached the masthead and one table row and stopped there. The basis is unchanged and remains the August 3 close; what moves here is everything that reprice never reached.

The enterprise value did not sum, and four multiples were computed off the figure that did not. The table gives a capitalization of $1,008M, notes of $230M, an Iceland term loan of $20M and cash of $80M, which is $1,178M; the row read $1.09B, which is the enterprise value at the superseded $915M capitalization. The value now reads $1.18B in the masthead, in section 08 and in the bull case, and the multiples move with it: EV to trailing revenue 11.5x to 12.5x, to revenue excluding the termination fee 13.2x to 14.3x, to FY2025 adjusted EBITDA 63x to 68x, and to 2027 revenue of $170 to 195M, 5.6 to 6.4 times to 6.0 to 6.9 times.

The composition of that enterprise value is not restated, and the row still says why. Cash is taken at $80M at March 31 while the snapshot carries $60.4M at June 30, and the debt side omits the Q2 and July drawdowns. The row records that it understates on both counts. Only the capitalization term moved, because only the capitalization was repriced.

Two claims inverted rather than aged, and the arithmetic forces both. A capitalization of $1,008M is above the $923.7M of remaining performance obligations it was said to sit between, by 9%. And the 2031 notes convert at $25.91 against a basis of $25.95, so the stock is four cents above the conversion price where two sentences said it was below. The economic reading is unchanged and is not restated: the zero-strike call lifts the effective conversion price to about $37, which this document already says in two places.

What did not move. The free float and 52-week range rows are market data struck at the 31 July close and carry that date. Three narrative sentences date the 31 July session and its $23.70 close and are correct as history. The correction record in this section is kept as written, with one sentence added recording that this reprice supersedes both figures in it.

Not absorbed, and named instead. A Form 8-K of August 17, 2026 and the company's announcement the same day of an agreement to acquire two North Carolina development sites were not read. Both are named in what was not checked. The submissions index was re-read on August 18, 2026 and the issuer newsroom was reached.

August 12, 2026
Priced off Aug 3, 2026 close · $25.95 · $1,008M cap

Q2 2026 was filed on August 12, 2026. Revenue $28.8M against a consensus near $19.4M and a loss of $0.39 per share against $(0.41). On the headline it beat on both lines.

$12.3M of that revenue is a one-off termination fee from the Initial Customer. Excluding it, revenue is $16.5M, below the consensus it appears to have beaten and down 24.6% sequentially from $21.9M; cloud excluding the fee falls from $16.8M to $11.5M, a third consecutive decline. The Company states the position itself: the lost monthly GPU revenue "was substantially offset by $12.3 million of termination fee revenue recorded." The fee was amended after quarter-end to $15.7M and remains outstanding.

The Initial Customer relationship ended rather than being renegotiated. That customer was approximately 70.7% of 2025 revenue. $12.5M of preserved receivables was collected in full and $2.2M was written off. Against the loss, new customer orders contracted in the six months and through the filing date total $635.8M over six-month to three-year terms.

The receivable question this report raised is answered, and answered well. Accounts receivable fell from $91.7M at March 31 to $23.2M at June 30, and six-month operating cash flow was positive $89.1M. That was the single largest open item in the Q1 reading and it converted.

The balance sheet went the other way. Working capital swings from $85.2M at December 31, 2025 to a deficit of $28.6M. Cash and restricted cash fall to $60.4M from $80.1M at March 31, against first-half capital expenditure of $344.7M. A $5.0M impairment of capitalised software is taken. Adjusted EBITDA is positive at $5.5M.

Four post-quarter events are recorded, none of which reached an 8-K. On August 10 the Company purchased 55,105 SAIHEAT Limited shares for ~$1.0M at $18.15 and surrendered its $1.0M Canopy Wave SAFE the same day; four members of Bit Digital’s management, including its Chief Financial Officer, invested $0.5M each in that same PIPE in their personal capacity, twice the Company’s own outlay between them. On July 30 a subsidiary leased data centre space in Sydney, Australia, commencing Q4 2026 on a 59-month term at AUD 488k a month, parent guaranteed, in a business whose cloud revenue is described as generated from Iceland. On July 6 Enovum entered a syndicated RBC credit facility, with the prior agreement repaid in full on July 15 and CAD 36.8M drawn. On July 27 and 31 the Company drew $20M and $10M on the Bit Digital Capital facility.

The share count is now newer than the price rather than older. 38,848,118 as of August 10, 2026 against a August 3 close, where the previous revision carried a May 12 count against a July 31 close and said so. Capitalization moves $1,002M → $1,008M on the count alone. The price was not restruck.

The Central Index Key is recorded here for the first time. The sweep is run against 0002042022 rather than against the ticker, which does not survive a rename. The index holds 78 filings from January 6, 2025 to August 12, 2026, with no supplementary index, against 73 to August 3 at the previous entry.

August 4, 2026
Priced off Aug 3, 2026 close · $25.95 · $1,002M cap

The coverage table now carries the analyst, the date, the price at publication and the disclosed conflict. The six rows and their ratings, targets and bases are unchanged; four columns were added to them. Price at publication is the close on the last trading day before each action, and it runs from $12.16 to $44.76 across a set of targets running $20 to $50, so the premium each firm was asking for at the time varies far more than the targets alone suggest.

One covering firm underwrote the company's own offering. Needham & Company, LLC took 2,631,563 of the 9,375,000 shares sold on August 8, 2025, the second-largest allocation behind B. Riley Securities. Barclays, BTIG, Cantor Fitzgerald and H.C. Wainwright are not in that syndicate. The relationship sits in the middle of the target range rather than at either end.

Two attributions are not settled and the table says so. A data service attributes a $38 Buy to B. Riley Securities dated May 18, 2026 where this document records Needham on 15 May, and both firms underwrote the offering, so which struck the target is not established. The analyst behind the H.C. Wainwright action is not established either. The aggregate row carries no publication date, and a service read on August 4, 2026 publishes a different figure, $38.11 across nine analysts dated July 21, 2026; both are reported and neither is averaged with the other.

The filing sweep was re-run and the record is unchanged in substance. The complete index holds 73 filings from January 6, 2025 to August 3, 2026 and carries no supplementary index. No late-filing notification of any kind exists across it. The Form 8-K of August 3, 2026 reporting the change of Chief Financial Officer, effective 1 August, was already carried and is unaltered.

No rotation away from the colors that carry meaning here was needed.

Known gaps. The company's own statement of who follows it was not read, so whether any covering firm has ended coverage is not established, and a service shows 2026 actions from two firms this table does not carry. What each firm discloses about this issuer in its own note is not established. The pricing basis is unchanged at the 3 August close, no session having completed since.

August 3, 2026
Priced off Aug 3, 2026 close · $25.95 · $1,002M cap
Correction

Market capitalization was tagged as published market data and is computed in this document. The masthead figure and the valuation-table figure both read Market. Each is the 31 July close multiplied by the filed share count, so each now reads Estimate. The tag asserted that a data provider had published the figure rather than that this document derived it.

The parent's beneficial-ownership filing exists and this report did not carry it. The ownership record was described as resting on periodic reports and the parent's own filings rather than a Schedule 13D. Bit Digital filed a Schedule 13D on August 8, 2025 reporting 27,043,750 ordinary shares, sole voting and dispositive power, approximately 74.3% of the shares outstanding and of total voting power at the offering. It is captioned in full rather than abbreviated, which is why a search for the abbreviated caption returned nothing. The count reconciles against the contribution: 27,043,749 shares for the HPC business plus the subscriber share. Read against the 70.1% carried from the May report, the parent has been diluted 4.2 points in nine months with no amendment filed and no disclosed sale.

Two carried defects are closed. The qualification dropped from the valuation caveat is restored to the bear case, the peer raising the rate on its own loan and its junk-rated 2032 bond near 10.32% at a B rating, stated at the precision the industry section already carries rather than at the rounded figure the lost clause used. And the bull case is restated from roughly $1.06 billion to $1.09 billion, which is the valuation section's own enterprise value at the 31 July close, so the document now reads on one pricing basis throughout.

The board and the finance function turned over on 1 August. Erke Huang left as chief financial officer and director, staying on as adviser and board observer and remaining chief financial officer of the parent; Justin Zhu succeeded him at $450,000 a year to July 2028, having resigned his own posts at the parent. David Andre left the board and all three committees. Samir Tabar, chief executive, joined the board, and Michael Rulf joined the compensation and nominating committees but not the audit committee, while holding a professional services agreement through his own company. Pruitt Hall stays on all three committees at $270,000, up from $150,000, and gave up his own services agreement to do so. The filing names no third audit-committee member in Mr Andre's place.

The mark has changed since the note was written, so the description is updated, but it still yields no usable hue.

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 $23.70 → $25.95, up 9.49% on the day and recovering most of the 9.71% fall of 31 July. Market capitalization moves $915M → $1,002M on the unchanged 38.61M share count, and the tradable float rises $260M → $288M. The filing record was swept again on August 3, 2026 by issuer central index key against the complete index of 73 filings, which carries no overflow. The newest item is the Form 8-K of that date, and its nine exhibits were enumerated rather than the form type ticked off: the substance is not one appointment but a governance refresh, carrying the advisory agreement for the departing chief financial officer, the incoming officer's employment and indemnification agreements, director and indemnification agreements for the chief executive and for Michael Rulf, an amendment to a third director's agreement, and a professional services agreement with Rasterwerk LLC, a company of Mr Rulf's, which is a related-party arrangement with a sitting director rather than board pay.

August 1, 2026
Priced off Jul 31, 2026 close · $23.70 · $915M cap

Revised against the 28 July publication.

Pricing. Repriced to the 31 July close of $23.70 (down 9.71% on the day, from a $27.06 open, $27.63 high). Market cap approximately $915M; enterprise value approximately $1.09B; EV/TTM revenue 12.6x → 12.9x; EV/2025 adj. EBITDA 61x → 63x; forward EV/2027e sales 5.5–6.2x → 5.6–6.4x. 52-week range corrected to $10.51–46.87. Free float of approximately 11.1M shares against 38.6M outstanding added to Section 0.

Filings. Form-by-form EDGAR review across 8-K, 6-K, 10-Q, 10-K, 20-F, 40-F, S-1, S-3, 424, 13D, NT 10-K, NT 10-Q, NT 20-F, NT 40-F, Form 3 and Form 4. Correction: the 28 July version stated that the most recent SEC filing was the 10 June 8-K. That was wrong; it relied on a third-party aggregator that was stale. Four Form 4s were filed on July 28, 2026, two of them delinquent by roughly four months against the Section 16(a) two-business-day deadline. Added to Section 9 as a governance risk and noted in Section 11. A performance milestone met on June 30, 2026 is disclosed in those filings; the milestone itself is unspecified and no inference is drawn from it. The periodic-report position (8-K, 10-Q, 10-K) was correct as originally stated and is unchanged; no NT filing exists anywhere in the company's history.

Industry and competitors (Sections 4–5). "Crack one" retitled and expanded for SpaceX's entry alongside Meta. "Crack two" rewritten from a forecast refinancing wall into observed credit repricing: a peer reported on 29 July as raising the rate on a $2.6bn loan, its 2032 bond near 10.32% at a B rating, widening technology CDS. Peer drawdowns added. Meta and SpaceX added to the competitor set.

Risks (Section 9). Two new subsections: insider reporting delinquency, and energy prices, inflation and the rate path. The 29 July FOMC held at 3.50–3.75% on a 9–3 vote with three dissents favoring a hike, with market pricing having moved from cuts to increases. Financing-market risk separated out as its own subsection. Sector risk updated for the current drawdown and float.

Elsewhere. Bear case sharpened for the credit environment (Section 10). Valuation caveat added on financing cost (Section 7). Jackson Hole 27–29 August and the 15–16 September FOMC added to the watch list (Section 11).

Unchanged. No new 8-K, 10-Q or press release since 10 June and 9 July respectively. NC-1 billing commencement, the Paris go-live and the NC-1 permanent financing all remain unconfirmed and stay tagged Open. All balance-sheet figures remain as of March 31, 2026.

July 28, 2026 (original build)
Priced off Jul 27, 2026 close · $23.15 · $894M cap

First published. Research report covering business model, revenue streams, industry backdrop, competitors, financial performance, valuation, growth drivers, risks, bull/base/bear cases and research summary. Built from SEC filings and company press releases through July 27, 2026, priced off that day's close of $23.15 (market cap approximately $894M, EV approximately $1.06B). Evidence tagged Filed / Estimate / Open throughout. No buy, sell or hold recommendation.