WhiteFiber

Phase A — Understand the business

Lens 1 · Company Overview

WhiteFiber is a vertically-integrated AI-infrastructure operator — it owns/develops HPC (high-performance-computing) data centers and runs a GPU cloud on top of them. Two revenue lines:

• Cloud services (the "NeoCloud") — renting NVIDIA GPU capacity (H100/H200/B200/GB200) to AI/ML developers and GPU marketplaces for training and inference. FY2025 $68.75M, 87% of revenue ``.
• Colocation / data-center services — leasing Tier-3 rack space and power (hosting third-party hardware). FY2025 $8.91M, 11% of revenue, up from $1.36M — the fastest-growing line ``.
• Other: $1.50M.
• FY2025 total revenue $79.16M, +66% YoY (from $47.64M FY2024) ``.

Corporate structure. A Cayman Islands holding company (HQ 31 Hudson Yards, NYC) carved out of Bit Digital (Nasdaq: BTBT) — a former bitcoin miner pivoting to AI. WhiteFiber IPO'd on Nasdaq Aug 7-8 2025 at $17.00/share, raising ~$183.3M gross including the over-allotment . Bit Digital retained ~71.5% post-IPO — WhiteFiber is a controlled company ``.

The business model in one line: buy cheap, power-rich industrial buildings → retrofit them to Tier-3 liquid-cooled AI halls in ~6 months → fill them with either contracted colo tenants (Nscale, Cerebras) or its own GPU fleet → capture margin at both the real-estate and the compute layer.

Contract structure & payment terms. The economics are prepayment-funded: large customers pay cash up front, which lands as deferred revenue and funds the build. Total deferred revenue $79.6M at 12/31/25 (current $8.0M + non-current $71.6M), up from $30.8M — the $48.8M increase is dominated by a $70.6M prepayment from Nscale ``. Colo contracts are long-dated take-or-pay-style with annual escalators and power as a pass-through; cloud contracts range from multi-year MSAs to week-to-week marketplace orders.

Key customers (see Lens 13 for the concentration risk):

• Nscale — 10-year, ~$865M total contracted-revenue colocation order at NC-1 (Madison NC), signed Nov 2025; 40 MW phase billing expected to start ~June 2026 ``.
• Cerebras — 5 MW IT-load colo tenant at MTL-3, billing CAD 1.4M ($979k) monthly since Nov 2025 on a 5-year contract ``.
• The "Initial Customer" — the legacy whale: 70.7% of FY2025 revenue and 96.6% of FY2024 revenue — now in a service pause / potential termination ``.
• DNA Fund (11.5% of FY2025 rev) — MSA terminated Nov 2025; Boosteroid — cloud-gaming, ROFR on up to 50,000 servers / ~$700M potential.

Suppliers: NVIDIA (Preferred Partner via NPN), Supermicro, Dell (via Advania in Iceland), HPE, and Quanta Cloud Technology (QCT) as the GB200 ODM. Power counterparty: Duke Energy at NC-1 (capacity agreement targeting up to 99 MW, possibly 200 MW long-term).

Lens 2 · Supply Chain

Upstream → WhiteFiber → end customer, named at every link:

1. Silicon / accelerators: NVIDIA (H100/H200/B200/GB200/GB300) — the single hardest chokepoint; WhiteFiber's ability to source GPUs on time is risk-factor #2 in its own 10-K ``.
2. Servers / ODM integration: Quanta Cloud Technology (GB200 NVL72 systems), Supermicro, Dell, HPE. First GB200 NVL72 shipment received April 2025 from QCT.
3. Construction inputs: steel, copper, aluminum, electrical gear, HVAC — sourced partly from Mexico and Canada → directly tariff-exposed (US tariffs on Canadian steel/aluminum/copper raise NC-1 build cost; Canadian reciprocal tariffs raise the Quebec builds) ``.
4. Power: Duke Energy (NC-1, US East-coast corridor); hydroelectric in Quebec (MTL-1/2/3) and Iceland (Blanda Hydro) — the green-power angle is a genuine cost/ESG edge.
5. WhiteFiber — retrofits + operates the halls (MTL-1/2/3, NC-1, Iceland, Atlanta).
6. End buyers: Nscale, Cerebras, Boosteroid, GPU marketplaces (Shadeform), and direct AI/ML developers.

Chokepoints / single-source dependencies: (a) GPU allocation from NVIDIA — without timely chips the cloud fleet can't grow; (b) Duke Energy power delivery at NC-1 — the entire $865M Nscale ramp is gated on Duke energizing 40 MW (Apr 2026 target) then 99 MW; a power slip = a revenue slip; (c) cross-border tariff exposure on the construction bill of materials. The supply chain is generic to the neocloud cohort except for the retrofit-in-constrained-power-markets angle, which is genuinely differentiated.

Lens 3 · Competitive Advantages (moats)

The real edge is cost-of-buildout and speed, not technology. Management claims ~$8-10M build-out cost per gross MW vs. a ~$13M industry average, and ~6-month retrofit build time vs. ~1/3-to-1/2 of the greenfield timeline ``. That comes from buying existing power-rich industrial buildings (an encapsulation plant, a Unifi textile mill) rather than developing greenfield — sidestepping the power-interconnect queue that is the binding constraint for the whole sector.

Durable-moat assessment, honestly:

• Scale / first-mover: weak — WhiteFiber is sub-scale against CoreWeave, Nebius, IREN, Applied Digital. ~76 MW targeted by Q4-2026 vs. peers measured in GW.
• Switching costs: moderate-and-growing — a 10-year Nscale colo lease and a 5-year Cerebras lease create real stickiness; physical colo tenants don't move easily.
• Power access as the moat: this is the strongest claim. In a market where power, not capital, is the bottleneck, a team that can source brownfield sites with in-place power and energize them fast has a structural edge. The 1,500 MW pipeline "under management review" is optionality on that skill ``.
• Vertical integration: owning both the hall and the GPUs lets it capture margin at two layers and redeploy GPUs across customers (it redeployed the paused Initial Customer's GPUs to three other customers) — genuine operational flexibility.
• IP / network effects: essentially none. This is an infrastructure/real-estate business with a cloud overlay, not a software moat.

Bargaining power: weak over NVIDIA (price-taker on the scarce input), weak-to-moderate over hyperscale-adjacent customers (Nscale is a $14.6B counterparty that can dictate terms), stronger over power utilities only where it has locked agreements. Net: a cost-and-speed operator, not a wide-moat compounder.

Lens 4 · Segments

No formal multi-segment EBITDA disclosure in the 10-K beyond the revenue/cost split (the company effectively runs one HPC business with two revenue lines). By revenue line ``:

*Gross profit = revenue − cost of revenue (exclusive of D&A) ``. Both lines run ~61% cost-of-revenue gross margin before D&A; after the $23.4M D&A and the $52.5M G&A the company is operating-loss-making.

Geography: operations span Canada (Quebec — MTL-1/2/3), Iceland (Blönduós), the US (NC-1 Madison NC; Atlanta), with Japan/Australia shells. The revenue today is Canada+Europe-weighted; the US (NC-1/Nscale) is the FY2026 growth engine.

Trend & cause: Colocation is accelerating hard (MTL-3 came online Nov 2025 with Cerebras; NC-1/Nscale starts Q2-2026) and is the structurally higher-quality, longer-duration revenue. Cloud is growing but lumpy and concentration-exposed. The mix shift from volatile cloud toward contracted colo is the bull's core argument.

---

Phase B — Measure performance

Lens 5 · Earnings Result (latest print — Q1 2026, filed 2026-05-14)

`` unless noted:

• Revenue $21.92M, +31% YoY (from $16.77M). Beat consensus $21.62M ``.
• Cloud $16.8M (+13% YoY — decelerating), Colocation $4.8M (+190% YoY) as MTL-3 ramps, Other $0.38M.
• Operating loss −$11.02M (vs +$2.04M operating income Q1-2025) — a swing into loss driven by G&A exploding to $17.77M from $4.24M (SBC of $7.3M is the bulk), and the new interest expense $2.0M from the January convertible.
• Net loss −$12.04M; EPS −$0.31 — MISSED the −$0.28 consensus ``.
• Margins: cost-of-revenue gross margin held (~60% on cloud, ~59% on colo); the loss is below the gross line — operating deleverage from growth spend, not a unit-economics break.
• Adjusted EBITDA $3.0M, DOWN from $6.0M Q1-2025 `` — the single most important number on the print: adjusted profitability fell ~50% while revenue rose 31%. This is a company spending ahead of a contracted ramp.
• Balance-sheet flags: cash $75.8M (down from $114.4M at YE — capex is burning the IPO proceeds); AR ballooned to $91.7M and deferred revenue +$65.0M in the quarter — the Nscale prepayment + billing cycle is washing through working capital. Working capital $124.9M.
• Capital-structure event: the $120M zero-strike call option purchased with the convertible vaporized $120M of additional paid-in capital, dropping total equity to $352.6M (from $482.5M) ``.
• Market reaction: shares reacted cautiously to the EPS miss but the stock is up huge over the trailing year (52-wk range $10.51-$40.75) — the market is trading the NC-1/Nscale inflection, not the current P&L ``.

FY2025 full-year for reference ``: revenue $79.16M (+66%); operating loss −$26.82M (vs +$0.63M FY2024); net loss −$24.68M; EPS −$0.78; operating cash flow +$45.66M (positive, deferred-revenue-funded); capex −$268.4M; Adjusted EBITDA $17.3M, down from $21.9M FY2024.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on disk (transcripts/ empty) and WYFI has only been public since Aug 2025, so the call history is thin — Q3-2025, Q4/FY-2025, and Q1-2026. From the public record ``:

• Consistent management focus: the NC-1 "$865M Nscale inflection" is the drumbeat across every communication — management is explicitly telling the Street to "look past Q4 noise" to the NC-1 ramp ``.
• Tone shift: from an IPO-era "diversified neocloud" story toward a "contracted-backlog, capacity-delivery" story — the narrative has pivoted from cloud-fleet growth to colo-build execution as the Initial Customer dispute surfaced.
• What they stopped saying: the Initial Customer is no longer framed as a growth anchor — it's now a wind-down/redeployment line. The emphasis moved to Nscale + Cerebras.
• New language Q1-2026: convertible-funded "development pipeline," "1,500 MW under review," and Adjusted-EBITDA discipline — classic build-phase framing.

Lens 7 · Comps (neocloud cohort)

Read: On a trailing-sales basis WYFI is ~18x ($1.49B ÷ $82.7M TTM) — richer than the large-caps trade on forward sales. On forward sales (~$139M FY26E) ~10.7x it sits roughly in line with NBIS/IREN/APLD and above CRWV. The premium is defensible only if the NC-1/Nscale backlog converts and FY27 revenue steps up sharply. Dividend yield: 0% across the cohort (all reinvesting). 5-yr avg ROE: n/a — WYFI has <1yr of public history and negative equity returns; not a comparable metric for a pre-profit builder.

Lens 8 · Stock-Price Catalysts

WYFI has <11 months of trading history (since Aug 7 2025), so the ">5% move over 5 years" frame compresses to "what's moved it since IPO." `` throughout:

• IPO Aug 7-8 2025 at $17.00 → ran to a 52-wk high of $40.75; 52-wk low $10.51. The stock has roughly 3.7x'd off its low — a violent re-rating ``.
• Nov 2025 — $865M Nscale contract announced: the single biggest fundamental catalyst, reframed the whole thesis around NC-1.
• Jan 2026 — $230M 4.50% convertible (upsized from $210M): funded the build but introduced the zero-strike-call equity hit and interest drag.
• Q4/FY-2025 print (Mar 2026) — revenue +61% in Q4, $865M deal reiterated; "look past Q4 noise."
• Q1-2026 print (May 14 2026) — revenue beat, EPS miss; Needham PT $36→$38 (Buy); Cantor $13→$27 (Neutral) — the Cantor doubling is notable ``.
• Jun 16 2026 — Bit Digital originated a $100M delayed-draw term loan to a WhiteFiber subsidiary; WYFI jumped ``.

Pattern: the market reacts to contracted-capacity news and financing (Nscale, the BTBT loan) far more than to the GAAP print. This is a backlog-and-funding story — the tape rewards de-risking the build, not quarterly EPS.

---

Phase C — Judge people & books

Lens 9 · Management

• Sam Tabar — CEO. Also CEO of Bit Digital (the 71.5% parent) — a dual-role, shared executive spending "not more than ~30% of his time" on Bit Digital under the Transition Services Agreement ``. Background in capital markets/legal; the architect of the BTBT→AI pivot.
• Erke Huang — CFO & Director. Also CFO of Bit Digital — same dual-role structure. Concentrating both the CEO and CFO seats across parent and subsidiary is a material governance flag (conflicts on capital allocation, related-party financing, transfer pricing all run through two people who sit on both sides).
• Operating bench: Billy Krassakopoulos (Enovum founder, 20+ yrs DC development — the retrofit engine) and Thomas SanFilippo (CTO, 40 yrs IT/infrastructure). The genuine data-center operating talent came in via the Enovum acquisition (Oct 2024) — that, not the Bit Digital lineage, is the real operating DNA.
• Track record: the Enovum team has demonstrably delivered ~75 MW of retrofit projects over their careers ``; the corporate/Bit Digital side has a track record of pivots (bitcoin mining → ETH staking → AI infra) — execution-on-the-current-thesis is unproven beyond ~1 year.
• Skin in the game: Bit Digital (and thus its insiders) own ~71.5%; alignment with minority holders is via the parent, not direct. SBC is heavy — $19.2M FY2025, $7.3M in Q1-2026 alone — diluting minorities while the parent's stake is fixed.
• Capital allocation: aggressive growth-capex ($268M FY2025), an IPO + a convertible + an Iceland term loan + a related-party BTBT loan in <12 months. Disciplined on site selection (only builds against customer commitments) but financially complex and dilutive; ROE/ROIC are negative and not yet a fair test.
• Archetype: professional-manager / financier-led at the top, founder-operator in the field. For a build-phase infra company that's workable, but the shared-executive overhang means the parent's interests can dominate.

Lens 10 · Forensic Red Flags

Acting as a forensic analyst on the FY2025 10-K + Q1-2026 10-Q:

• Revenue recognition & deferred revenue: the $70.6M Nscale prepayment sits in deferred revenue and converts to revenue as 40 MW is energized — watch the timing; any Duke Energy power slip pushes recognition right. Sales-type lease accounting (net investment in lease $13.9M) and the colo-vs-cloud split deserve scrutiny but look conventional.
• Cash flow vs. earnings divergence: operating cash flow is positive (+$45.7M FY25) while GAAP is a loss — because of customer prepayments (deferred revenue +$48.7M), not operating strength. This is a financing-via-customer-prepayment model; if the prepayment spigot (Nscale) slows, OCF inverts fast. In Q1-2026, OCF was distorted by AR +$67.8M vs deferred-rev +$65.0M — a wash that will reverse.
• Receivables outrunning revenue: AR jumped to $91.7M (Q1-26) against $21.9M quarterly revenue — a >4x AR/quarterly-rev ratio. Much is the Nscale billing cycle, but collection risk is real — the company already had to chase $7.3M from the terminated DNA Fund (only $2.1M collected) ``.
• SBC flattering non-GAAP: Adjusted EBITDA adds back $19.2M of SBC (FY25) — without it the company is barely EBITDA-positive. Heavy reliance on SBC to bridge to a "profit" metric.
• Goodwill / intangibles: $20.1M goodwill + $12.8M intangibles from Enovum — modest, low impairment risk for now.
• The zero-strike call option: the convertible was structured with a $120M zero-strike call that hit equity directly — an unusual, complex structure (effectively a prepaid share-settlement/hedge). It means the headline "$230M raised" delivered only ~$102.5M of net cash ``. Investors must not read the gross convert as liquidity.
• Smaller-reporting-company / EGC status — reduced disclosure; no auditor attestation on internal controls (Sarbanes-Oxley 404(b) not yet required). Audit fees are tiny ($307k) for the balance-sheet size — normal for a newly-public EGC but worth noting as the company scales.

Regulatory findings (required sub-section):

• SEC Litigation Releases / AAERs: None. regulatory/regulatory-findings.md reports total_sec_findings: 0 across SEC EDGAR EFTS (LR + AAER) for the 2021-06-20 → 2026-06-20 window ``.
• 10-K Item 3 (Legal Proceedings): the company states it is not presently party to any material litigation ``.
• Non-SEC enforcement (FTC/DOJ/FDA/CFPB): web search surfaced no material enforcement actions or consent decrees against WhiteFiber ``.
• Verdict: No material regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER), web search, and 10-K Item 3 as of 2026-06-20. The real "legal" overhang is commercial, not regulatory: the Initial Customer termination negotiation (40%-of-remaining-fees early-termination claim) and the related-party governance with Bit Digital.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (EPS, three fiscal years — FY2026 / FY2027 / FY2028)

Built bottom-up from FY2025 actuals + Q1-2026 + the Nscale ramp + Street consensus. Output ; consensus anchors . Forecast NOT logged to forecast.ts (watchlist/unattended rule).

Revenue:

• FY2026E ~$137-140M — Street consensus ``. Driver: NC-1/Nscale 40 MW billing from ~June 2026 + MTL-3/Cerebras full year + cloud redeployment. Roughly +75% YoY.
• FY2027E ~$240-300M ``. Wide band — entirely gated on Duke Energy power delivery.
• FY2028E ~$350-450M ``.

EPS path (≈38.6M shares, rising ~3-5%/yr on SBC dilution; converts add ~$10.4M/yr interest):

• FY2026E EPS ≈ −$0.85 to −$1.10 . Consensus implies positive EPS only by **Q1-2027** .
• FY2027E EPS ≈ −$0.10 to +$0.40 ``.
• FY2028E EPS ≈ +$0.60 to +$1.50 ``.

Base / bull / bear:

• Base: FY26 −$1.00, FY27 +$0.15, FY28 +$0.90. NC-1 energizes roughly on schedule; Nscale performs; one incremental site.
• Bull: FY27 +$0.50, FY28 +$1.80. Duke delivers 99 MW early, Boosteroid expands, pipeline converts; the multiple holds → meaningful upside from $38.
• Bear: FY27 −$0.40, FY28 −$0.10. Duke power slips, Nscale renegotiates/slows, converts + BTBT loan pile leverage on a still-loss-making P&L; the ~18x trailing multiple compresses hard.

The forecast that matters most is binary, not EPS: does Duke Energy energize NC-1's 40 MW phase on schedule so Nscale billing starts in Q2-2026 as guided? That single event resolves most of the FY26-27 spread.

Lens 12 · Bull vs Bear

Bull case. WhiteFiber is a cost-advantaged retrofit operator ($8-10M/MW vs $13M, ~6-month builds) attacking the one binding constraint in AI infra — power-ready capacity — at a moment of "well-documented surge in demand." It has already converted that into an $865M, 10-year contract with Nscale (a $14.6B Nvidia/Microsoft-backed counterparty) plus a 5-year Cerebras lease, and runs positive operating cash flow funded by customer prepayments. The mix is shifting from volatile cloud toward sticky, escalator-protected colo. A 1,500 MW pipeline is free optionality on a proven retrofit skill. At ~$1.49B it's a fraction of CRWV/NBIS/IREN/APLD — if it executes NC-1 and lands even one more large site, FY27-28 revenue could 3-4x and the stock re-rates with the cohort.

Bear case (permanent-impairment risks).

1. Customer concentration is existential, twice over. The Initial Customer was 70.7% of FY25 revenue and is in termination; the replacement concentration is Nscale — swap one whale for another. If Nscale (itself a young, cash-burning private) slows, renegotiates, or fails to ramp, the entire bull thesis breaks.
2. The multiple already prices the inflection. ~18x trailing / ~10.7x forward sales on a company with negative GAAP earnings, Adjusted EBITDA going down YoY, and EPS that misses. The market has paid up front for NC-1; the print can only disappoint relative to that.
3. Power-delivery dependency. The whole FY26-27 ramp is gated on Duke Energy energizing NC-1 on a utility's timeline — outside management's control. A 1-2 quarter slip cascades through revenue, covenants, and sentiment.

Pre-mortem (18 months out, thesis broke): Duke delayed NC-1 to late-2026; Nscale (under its own funding pressure) renegotiated the $865M order; the Initial Customer termination cost more than the 40% fee implied; the converts' interest + the $100M BTBT related-party loan turned a build-phase loss into a financing squeeze; SBC kept diluting; the stock round-tripped from $38 toward the high-teens as the sales multiple compressed to the cohort's forward range on lower-than-hoped revenue.

Contrarian view (what the market is refusing to see): the market is treating WYFI as a mini-CoreWeave when its real comparable is a leveraged, single-contract, parent-controlled specialty REIT-with-a-GPU-overlay. The genuine, under-appreciated asset is the Enovum retrofit team's power-sourcing skill — but that's being valued as if the $865M Nscale contract is already de-risked revenue rather than a power-gated promise. Multiples are too high for the current risk; they're a bet on flawless execution.

Lens 13 · Devil's Advocate (short-seller)

The short case writes itself, and the tape agrees: short interest is 23.3% of float, up 711% over 12 months ``.

• Where revenue is concentrated: historically 96.6% (FY24) / 70.7% (FY25) in ONE customer that is now winding down; prospectively the bulk shifts to Nscale. This is not diversification — it's serial single-counterparty risk dressed as a backlog.
• Why the moat is weaker than bulls think: there is no technology moat. The "$8-10M/MW" cost edge is management's own estimate, unaudited, and replicable by any competent retrofit developer. CoreWeave/Nebius/IREN/APLD have 10-100x the capacity, scale procurement leverage with NVIDIA, and cheaper capital. WYFI is the price-taker.
• Most dangerous competitor bulls underestimate: not the mega-caps — it's Applied Digital's exact playbook (HPC-hosting-for-hyperscalers, then spin the cloud into ChronoScale). APLD is doing the same retrofit-colo model at 5-10x the scale and ~9x EV/Sales. WYFI has no differentiated answer.
• Worst capital-allocation / governance moves: CEO and CFO are shared with the 71.5% parent; a $100M related-party loan from that parent; a $120M zero-strike-call structure that cut net convert proceeds to ~$102.5M; $19M/yr SBC diluting minorities. Every related-party lever runs through two conflicted executives.
• Assumptions that must hold for $38: (a) Duke energizes NC-1 on schedule; (b) Nscale performs the full 10-year $865M; (c) FY27 revenue ~3x's; (d) the ~18x trailing multiple doesn't compress. All four must hold.
• If growth disappoints 20-30%: FY27 revenue of ~$200M instead of ~$280M, still loss-making, on a stock priced for the bull path → a plausible 40-60% drawdown as the multiple re-rates to the cohort's forward sales on a lower number.
• Single scenario that permanently impairs: Nscale renegotiates or defaults on NC-1 (a young private under its own funding strain) while the Initial Customer termination drags — leaving a half-built 1M-sq-ft NC-1 facility with leveraged financing and no anchor tenant. Plausibility: low-to-moderate, but non-trivial given both counterparties' youth.

Lens 14 · Management Questions (ordered by information value)

1. Nscale NC-1: What are the exact contractual conditions, milestone dates, and termination/penalty terms of the $865M order, and what is your remedy if Duke Energy fails to deliver 40 MW / 99 MW on the agreed dates?
2. Initial Customer: What is the current status and expected cash impact of the termination negotiation — the 40%-of-remaining-fees early-termination fee, the non-refundable prepayment, the service deposit, and the outstanding receivables?
3. Concentration: Pro-forma for NC-1 ramping, what % of FY2027 revenue is Nscale, and what is your hard plan to get the top customer below 40%?
4. Power: What is Duke Energy's current, confirmed delivery schedule at NC-1 versus the 24/40/99 MW capacity-agreement targets, and where are you in the interconnect queue?
5. Governance: Given the CEO and CFO both serve Bit Digital and WhiteFiber, how does the Audit Committee independently price related-party transactions (the $100M loan, the TSA), and is there a path to fully independent management?
6. Capital structure: Walk through the $120M zero-strike call — why that structure, what it cost in net proceeds, and how it settles at maturity / on conversion of the 2031 notes.
7. Liquidity runway: With $75.8M cash, $268M+/yr capex pace, $10.4M/yr convert interest, and the BTBT $100M facility — what is the fully-funded build plan, and when do you next need external capital?
8. Unit economics: What is the contracted unrleveraged yield-on-cost (stabilized EBITDA / build cost per MW) for NC-1 and MTL-3, and how does it compare to the $8-10M/MW build claim?
9. Receivables: AR is $91.7M against $21.9M quarterly revenue — what is the aging, how much is Nscale, and what is the collection plan given the DNA Fund experience?
10. Pipeline: Of the 1,500 MW "under review," how much has site control + power commitments, and what is the realistic FY2027 conversion?
11. Cloud strategy: With cloud revenue decelerating (+13% in Q1) and concentration risk, is the GPU-cloud fleet now a redeployment buffer for colo, or still a standalone growth line?
12. Margins: When does Adjusted EBITDA growth resume — it fell YoY in both FY2025 and Q1-2026 — and what is the steady-state colo vs cloud margin you underwrite?
13. Tariffs: Quantify the Canadian/Mexican tariff impact on NC-1 and Quebec build costs, and what is the mitigation.
14. Bit Digital separation: Is full independence from Bit Digital (buy-in, spin-off, or majority sell-down) on the table, and on what timeline — and how would a BTBT block sale affect WYFI's float?
15. Competition: Applied Digital is running the identical retrofit-colo-then-spin-the-cloud playbook at far greater scale — what is your durable differentiation beyond speed-of-build?

---