HIVE Digital Technologies

Phase A — Understand the business

Lens 1 · Company Overview

HIVE makes money two ways, and the mix is the whole story:

1. Bitcoin mining (94% of FY26 revenue). HIVE operates ~25 EH/s of ASIC hashrate across green-powered data centers, sells that hashrate to mining pools, and is paid in BTC. FY26 mining revenue $278.3M on 2,885 BTC mined (avg realized price $98,040).
2. High-performance computing / "BUZZ HPC" (6% of FY26 revenue, the growth narrative). ~5,000 NVIDIA GPUs (A5000/A6000/A40/H100/H200, plus new B200 Blackwell) rented on GPU marketplaces and fixed-term contracts for AI/LLM training, inference and rendering. FY26 HPC revenue $19.5M; contracted HPC ARR reached ~$35M subsequent to year-end.

Customers / contract structure. Mining: third-party mining pools (commodity-like, no customer concentration — HIVE is a price-taker on BTC and network difficulty). HPC: a mix of on-demand marketplace aggregators (variable, higher $/hr — H200 on-demand ~$3.99/hr) and fixed-term contracts (6 H200 contracts, ~360 GPUs at $1.45–1.85/hr; B200s signed to a 2-year contract at $2.90/hr). The HPC book is shifting from spot to contracted ARR — the more durable revenue line.

Suppliers. Bitmain (Antminer ASICs — S21+ Hydro, S21 XP), Canaan (Avalon), NVIDIA (GPUs via Supermicro/Dell servers), and the power utilities (ANDE in Paraguay, Swedish grid). Custody: Fireblocks (~89% of digital assets) + Bank Frick (~11%). Competitors: the listed BTC-miner cohort (MARA, Riot, CleanSpark, Cipher, IREN, Bitfarms) — increasingly also HPC/AI-cloud rivals.

Name change tell: "HIVE Blockchain Technologies" → "HIVE Digital Technologies" (Jul 2023) signposted the AI/HPC pivot well before the financials reflected it.

Lens 2 · Supply Chain

Map: NVIDIA / Bitmain / Canaan (silicon) → Supermicro / Dell (servers) → HIVE data centers (powered by ANDE Paraguay hydro + Swedish hydro + Canadian grid) → mining pools (BTC buyers) and GPU-marketplace / enterprise AI customers (compute buyers).

Named stakeholders & chokepoints:

• ASIC supply — Bitmain (single-source-ish). HIVE's hashrate growth is gated by Bitmain delivery. Notable: HIVE paid Bitmain in BTC — transferring 2,139 BTC as equipment deposits in FY26 with buy-back options (Black-Scholes-valued). This couples the equipment supply chain directly to BTC price.
• Power — ANDE (Paraguay state utility) is the critical dependency. ~307 MW of the ~385 MW utilized footprint sits in Paraguay (Yguazú 200 MW + Valenzuela 107 MW), ~78% of mining hashrate. Single-country, single-utility concentration is the dominant operational chokepoint. Sweden adds ~40 MW hydro (~9% of capacity); New Brunswick 70 MW + Quebec round out the rest.
• GPU supply — NVIDIA allocation gates the AI pivot. The B200 Blackwell units (504, in Dell XE9680 servers) are HIVE's first liquid-cooled deployment, via the Bell Canada "AI Fabric" partnership.
• Custody — Fireblocks + Bank Frick. Concentrated but treasury is now tiny (~$10.6M / 150 BTC), so custody risk is immaterial post-liquidation.

This lens is real, not generic: the company's fate is wired to two single points — ANDE power and Bitmain/NVIDIA hardware allocation.

Lens 3 · Competitive Advantages (moats)

The honest answer: the mining business has essentially no moat; the emerging moats are power siting and the energy-arbitrage option.

• No moat in mining itself. Hashrate is a commodity; revenue is BTC price × share of network hashrate, minus power. HIVE is a price-taker on both BTC and difficulty (difficulty rose to avg 135.8T in FY26 from 95.7T). Any miner with cheaper power and newer ASICs competes directly.
• Real edge #1 — low-cost green power at scale. Paraguay hydro via ANDE + Swedish/Canadian hydro give HIVE a structurally low, renewable power cost and an ESG-marketable footprint. Fleet efficiency ~16 J/TH (24.5 EH/s optimized) is a competitive, modern fleet. Power cost is the only durable cost moat in mining.
• Real edge #2 — energy-arbitrage optionality. In Sweden, HIVE holds fixed-price forward power contracts and can sell energy back to the grid at spot when that beats mining margin. A genuine, if small, optionality the pure-play miners lack.
• Emerging edge #3 — site control for AI conversion. Owning/controlling 460+ MW of permitted, powered, green sites is the scarce input for the AI-compute land grab. The bull thesis is that these sites convert to high-margin HPC/AI colocation — but conversion is capex-heavy and unproven at HIVE's scale (see Lens 11/12).

Bargaining power: weak over Bitmain/NVIDIA (allocation-constrained sellers' market); moderate over ANDE (HIVE is an anchor industrial load, but Paraguay sets the terms). Over BTC buyers: none (commodity).

Lens 4 · Segments

Two reported segments — mining and HPC — by product; geography is Canada / Sweden / Paraguay.

Cause of the trend: mining revenue +164% YoY driven almost entirely by the 300 MW Paraguay expansion (hashrate 6.5 → 25.1 EH/s over FY26) plus a higher avg BTC price ($98,040 vs $75,881), partly offset by the April-2024 halving and rising difficulty. HPC nearly doubled on the H200 cluster.

The deceleration the table hides — quarterly gross operating margin %: Q1 35% → Q2 49% → Q3 35% → Q4 24%. As difficulty climbed and BTC fell into the Q4 print (avg realized $76,476 in Q4 vs $98,040 FY-avg), mining economics compressed sharply. Geographically, Paraguay is now the revenue center of gravity (~78% of hashrate).

---

Phase B — Measure performance

Lens 5 · Earnings Result (FY26 10-K, the latest print)

• Revenue $297.8M, +158% YoY (mining $278.3M, HPC $19.5M). A genuine top-line blowout driven by the Paraguay capacity quadrupling.
• GAAP gross margin −21%. Cost of sales of $360.3M exceeded revenue, sunk by $170.4M of depreciation (up from $64.5M) on the newly-installed ASIC fleet. Ex-depreciation gross operating margin was +$107.9M / 36% (non-GAAP).
• Net loss $(148.4)M vs $(3.0)M FY25; EPS $(0.66) vs $(0.02). The swing is not operating cash — it's three non-cash/below-the-line hits: depreciation $170.4M, a $22.7M derivatives loss (on the Bitmain BTC-buyback options, $23.1M loss), a $16.0M unrealized investment loss, and SBC $25.5M. Net realized+unrealized gains on digital currencies collapsed to $10.7M (from $33.7M / $81.8M) because they no longer hold meaningful BTC.
• Adjusted EBITDA $72.9M (FY26) vs $56.2M (FY25) — but this is the misleading headline. Q4 Adjusted EBITDA was −$9.0M, down from +$44.6M (Q1) → +$31.5M (Q2) → +$5.7M (Q3). The company exited the year burning cash on an operating basis.
• Guidance/tone: HIVE doesn't give formal numeric guidance; it guides on hashrate and ARR targets. Q4 was reported as an earnings miss and the stock fell on the print; Chairman Frank Holmes publicly called the share-price volatility "a non-event".
• Balance-sheet flags (the big ones): working capital collapsed from $175.8M → $5.4M YoY; cash flat at $23.1M; digital currencies fell from $181.1M → $10.8M (150 BTC). PP&E +$277.6M to $480.5M. They funded $222.9M of investing outflow with $196M of equity (ATM) + selling $269M of mined BTC. Operating cash flow was positive $62.3M — the cash engine works; it's just dwarfed by the build-out.
• What's unusual vs HIVE's own history: first year of net-loss-after-positive-Adj.-EBITDA driven by depreciation catching up to the capex; first year the treasury was effectively zeroed to fund growth; first material negative operating-EBITDA quarter (Q4).

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on the shelf; sourced from the FY26 Q4 call coverage.

• Management focus has visibly shifted from "BTC miner" to "AI infrastructure / sovereign compute." The FY26 results PR headline itself is "...BUZZ HPC Positioned for Growth," and the call centered on the path to $660M ARR by 2028 anchored by the GTA Gigafactory — not on BTC production.
• Recurring phrases: "ARR," "sovereign AI," "green energy," "Gigafactory," "HPC pivot." Stopped saying: "HODL"/treasury accumulation — consistent with having sold the stack.
• Tone: promotional and forward-leaning (a Holmes hallmark) even as the current-quarter print missed and EBITDA turned negative. Calling a double-digit-% earnings-day drop "a non-event" is a tell about how management frames the gap between the narrative and the Q4 tape.

Lens 7 · Comps

Peer set = listed BTC miners pivoting to HPC/AI.

Read: the market is re-rating the cohort on contracted HPC/AI capacity, not mining. IREN (Microsoft anchor tenant) is the template HIVE is chasing; JPM's IREN "Underweight despite $39 PT" is the cautionary read-across — the Street will pay up for AI optionality but flags when undeveloped sites are already in the price. HIVE's consensus PT (~$6.76, ~65% above the ~$4.03 spot) shows the sell-side is crediting the Gigafactory story.

Lens 8 · Stock-Price Catalysts

What actually moves HIVE >5%:

• The AI pivot announcements (the dominant 2026 catalyst). The stock roughly doubled from ~$2.19 (Nasdaq ref price, 2026-04-16) to ~$4.0–4.5 (Jun 2026). The leg-up tracks the GTA Gigafactory announcement (2026-05-18) and the cascade of BUZZ HPC ARR / Bell Canada expansion headlines.
• Earnings prints / Adj.-EBITDA misses move it down — Q4 FY26 results triggered a drop.
• BTC price + network difficulty — the classic miner beta; mining is 94% of revenue, so HIVE remains a high-beta BTC proxy underneath the AI story.
• Capital-markets events — ATM raises (dilution) and the April-2026 $115M convertible-note issue.
• Analyst actions — B. Riley / Canaccord / HC Wainwright PT changes.

Pattern: in 2026 the marginal price-setter switched from BTC/hashrate to AI-capacity narrative. That's the re-rate — and the risk, because the narrative is running ahead of funded, online capacity.

---

Phase C — Judge people & books

Lens 9 · Management

• Aydin Kilic — President & CEO (since Jul 2022). Operating leader of the global build-out; came up through HIVE operations. Track record: delivered the headline operational feat — hashrate 6.5 → 25.1 EH/s and the 300 MW Paraguay expansion inside FY26, plus standing up BUZZ HPC to ~$35M ARR. Execution on capacity has been real.
• Frank Holmes — Executive Chairman. Also CEO/CIO of U.S. Global Investors (asset manager) — a promotional, capital-markets-savvy operator who is the public face of the equity story. Co-founder/director Marcus New also signs the board attestation.
• Tenure & skin in the game: founder-adjacent, long-tenured chair; precise insider ownership is in the proxy (incorporated by reference, not in the 10-K) — n/a on exact %. No insider-transactions.csv on the shelf.
• Capital-allocation history — the defining judgment call: management chose to liquidate essentially the entire BTC treasury (2,201 → 150 BTC; $477.6M of disposals) to fund the Paraguay capex and equipment deposits, rather than dilute even more or take on more debt. Defensible if BTC stagnates and the assets compound; a major opportunity cost if BTC rallies (they sold the call on their own thesis). They also fund growth through relentless ATM dilution — weighted shares 90M (FY24) → 128M (FY25) → 225M (FY26), and still issuing post-YE — which is the most shareholder-unfriendly pattern here. The April-2026 $115M exchangeable note (with capped calls) is a sensible step toward non-dilutive project funding.
• Red flags: related-party dealings are minor and disclosed — ~$0.21M/yr paid to a director-controlled company for marketing services; ~$0.1M expense reimbursements to officers. Promotional tone (the "non-event" comment) is the main soft flag. No accounting red flags in the audit (Sarbanes-Oxley 404(b) attestation present; large accelerated filer; PCAOB-audited).
• Archetype: founder/promoter chair (Holmes) + operator CEO (Kilic). Implies aggressive growth, narrative-forward IR, and a high tolerance for dilution — appropriate for a land-grab phase, dangerous if the capital markets window closes before the Gigafactory is funded.

Lens 10 · Forensic Red Flags

Acting as a forensic analyst over the FY26 10-K:

• Revenue recognition (mining): BTC mining revenue is measured at the fair value of BTC at receipt, daily. Standard for the sector; low manipulation risk, but it injects BTC-price volatility straight into the top line.
• Depreciation / impairment is the number to watch. $170.4M of depreciation drove the GAAP gross loss; ASICs depreciate fast and HIVE upgrades aggressively. Long-lived-asset impairment is tested at the asset-group level — a sustained BTC drop or a stranded Paraguay cohort could force a writedown of the $480.5M PP&E. No impairment taken in FY26, but it's the obvious future risk.
• Cash-flow vs earnings divergence — benign direction. Net loss $(148.4)M but operating cash flow +$62.3M; the gap is non-cash depreciation + derivatives + SBC. This is the good kind of divergence (earnings worse than cash), not the fraud-pattern (earnings better than cash). Worth noting the mining cash conversion: $278.3M mining revenue recognized vs $269.1M cash proceeds from BTC sales — they sell roughly what they mine, so "earnings" are close to cash, modulo the HODL swing.
• Derivatives / BTC-buyback structure. The 2,139-BTC equipment-deposit-with-repurchase-option scheme created a $22.7M FY26 derivatives loss and ongoing Black-Scholes remeasurement noise. It's disclosed and economically rational (preserves BTC upside while paying suppliers), but it's a non-obvious, model-dependent line that obscures clean reads of the P&L.
• SBC flatters non-GAAP. $25.5M SBC (8.6% of revenue) is added back in Adjusted EBITDA — material, and the dilution is real.
• Investments mark-to-market. $9.7M of equity investments produced a $16.0M unrealized loss in FY26 — a non-core source of P&L volatility for a compute company.

Regulatory findings (required):

• SEC: No Litigation Releases and no AAERs name HIVE Digital (or HIVE Blockchain) in 2021-06-18→2026-06-18 — verified via SEC EDGAR EFTS (LR + AAER). Web search for SEC investigation/fraud/enforcement returned only routine HIVE filings, no enforcement.
• Item 3 / Note 17 (Legal Proceedings & Contingencies) — the material item, and it's a tax fight, not litigation:
• Sweden — contingent VAT liability of SEK 765.6M (~$80.5M). The Swedish Tax Agency seeks to reject all input-VAT recovery for Bikupa / Bikupa 2 (Dec 2020–Dec 2025) and claw back amounts already received + interest. Unprovisioned — HIVE believes it is not probable (backed by EU guidelines, a Swedish VAT professor's opinion, and an IT-forensic opinion), but the cases have lost at the Administrative Court and the Court of Appeal and are now appealed to the Supreme Administrative Court. At ~$80.5M this is ~13% of total assets and the single largest balance-sheet risk in the file.
• Canada — ~C$48.6M of withheld consumption-tax credits (9376-9974 Quebec Inc. C$8.2M + HIVE Atlantic C$40.4M) withheld by the CRA pending the Feb-2022 measure restricting hashrate-service credits; $4.5M provision taken in FY24.
• Routine litigation otherwise; management asserts adequate provisions.
• Net: clean on securities/accounting enforcement; carrying a large, contested, unprovisioned Swedish VAT exposure that an adverse Supreme Court ruling would crystallize.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection

HIVE is structurally loss-making on a GAAP basis (depreciation > gross profit) and management guides on ARR and hashrate, not EPS. Building an EPS line three years out for a company whose P&L is dominated by BTC price, difficulty, depreciation timing, and ongoing dilution would be false precision — so this lens frames the value drivers and the funding gap, with explicit `` arithmetic, rather than a fabricated EPS path. No Brier forecast logged (watchlist/unattended mode; EPS not the right scoreable metric here, and I will not fabricate a consensus number — n/a).

The bull bridge (management's own framing):

• HPC/BUZZ ARR: ~$35M today → $660M by year-end 2028, anchored by:
• GTA AI Gigafactory — 320 MW, 25-acre Toronto-Waterloo site (acquired CAD $58M), >100,000 NVIDIA GPUs, online H2 2027, targeting $360M ARR at $150/kW/month on 15-year terms; total dev cost ~CAD $3.5B.
• Plus New Brunswick 70 MW → Tier-III hyperscaler colo (mgmt-estimated ~$85M ARR) and BC/Manitoba liquid-cooled expansion (16.6 MW).

The arithmetic that decides the thesis:

• Dev cost ~CAD $3.5B (~US $2.55B at ~0.73) against US $23.1M cash, $5.4M working capital, ~$10.6M BTC, and ~$110M from the new convertible. The funding gap to a 320 MW build is >US $2B.
• This is unbridgeable from operating cash flow ($62.3M/yr) or the BTC treasury. It requires project-level / infrastructure financing and, almost certainly, signed hyperscaler anchor tenants to underwrite the debt — none of which is yet disclosed as committed. The 15-year, $150/kW/mo pricing is described as "peer-comparable," i.e., not yet contracted at the Gigafactory.
• Base-case read: FY27 (ending Mar-2027) likely remains a mining-driven, GAAP-loss year with HPC ARR climbing from $35M toward ~$50–75M as B200/BC capacity comes online; the Gigafactory contributes ~zero revenue until H2-FY28 at the earliest and only if financed. Equity value over the next 12–18 months is therefore set far more by financing/anchor-tenant news flow and BTC price than by near-term EPS.

Lens 12 · Bull vs Bear

Bull case. HIVE controls 460+ MW of permitted, green-powered, low-cost sites — the single scarcest input in the AI-compute land grab — and is converting from commodity BTC mining (where it's a price-taker) into contracted, 15-year HPC/AI revenue (where it's a landlord). The operating team just proved it can build at speed (6.5→25 EH/s in a year). If even a fraction of the $660M ARR / GTA Gigafactory plan gets financed and signs a hyperscaler anchor (the IREN-Microsoft template), HIVE re-rates from a ~1x-sales BTC miner to an AI-infrastructure multiple — and the consensus ~$6.76 PT (Canaccord $10) starts to look conservative. Positive operating cash flow ($62.3M) and a fresh $115M convertible buy time; a BTC rally would re-load the treasury optionality they just spent.

Bear case (2–3 permanent-impairment risks):

1. The funding gap is fatal to the dream. A >$2B build on $23M of cash, financed by chronic ATM dilution, with no committed project finance and no announced anchor tenant. If the capital-markets window closes (BTC bear, rates, AI-capex digestion), the Gigafactory stalls and the entire re-rate unwinds. Shareholders are diluted whether it succeeds (more equity) or fails (down round / impairment).
2. Mining economics are deteriorating now. Q4 FY26 Adjusted EBITDA went negative (−$9.0M) as difficulty rose and BTC fell; 94% of revenue is still this commodity line. The "record $298M revenue" headline masks a business that exited the year burning operating cash.
3. The $80.5M unprovisioned Swedish VAT clawback — already lost twice in court — could crystallize and take a real bite out of a balance sheet with $5.4M of working capital.

Pre-mortem (18 months out, thesis broke): BTC chopped sideways-to-down through 2026–27, mining margins stayed thin, and HIVE couldn't secure project financing or a hyperscaler anchor for the GTA Gigafactory at acceptable terms. It kept the lights on via dilutive ATM raises (share count through 300M+), the Swedish VAT case went against it, and the stock round-tripped back toward the ~$2 April-2026 level as the market repriced unfunded AI optionality to zero. The "sovereign AI" narrative outran the capital structure.

Are multiples too high? On current mining cash flows, yes — the ~$1.1B cap prices substantial unbuilt AI capacity (JPM's exact IREN warning). On the option value of 460 MW of sites + a credible operator, it's a defensible call-option price. It is not an earnings stock; it's a financing-execution + BTC-beta call.

Contrarian view (what the market is refusing to see): Bulls are pricing the GTA Gigafactory as a near-certain $360M ARR annuity; the market is under-weighting that HIVE has not shown it can finance a multi-billion-dollar build without crushing dilution, and has zero announced anchor tenant. The skill is real; the balance sheet is not built for this. Conversely, the under-appreciated upside is the energy-arbitrage + site-control optionality if AI-power scarcity intensifies — but that's a smaller, slower story than the Gigafactory headline.

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

Dismantling the bull case:

• Where revenue is concentrated: 94% Bitcoin mining — a pure commodity with no pricing power, now margin-compressing (Q4 EBITDA negative). The "AI company" is 6% of revenue. You are short a BTC miner wearing an AI costume.
• Geographic concentration: ~78% of hashrate in Paraguay, dependent on one state utility (ANDE) and one country's political/regulatory/FX stability. A single ANDE tariff change or political shift impairs the core.
• Why the moat is weaker than bulls think: site control only becomes an AI moat after multi-billion-dollar conversion capex that HIVE can't self-fund. Until financed and tenanted, it's 460 MW of mining sites, not AI sites. Every other miner (IREN, Cipher, Riot, CoreWeave-adjacent) is making the identical pivot pitch.
• Most dangerous competitor bulls underestimate: IREN — already has the Microsoft anchor, ~250k GPUs modeled, and a funded path. HIVE is announcing what IREN has signed. CoreWeave/Nebius-class neoclouds raise the bar on what a credible AI-cloud actually requires (financing + tenants + Tier-III ops at scale).
• Worst capital-allocation moves: selling 2,000+ BTC near a treasury low to fund capex (selling their own thesis), and funding the company through serial ATM dilution (shares +150% in two years). The BTC-buyback-option scheme adds derivative P&L noise.
• Assumptions that must hold for ~$4: (1) Gigafactory gets financed at acceptable cost; (2) a hyperscaler signs ~$150/kW/mo, 15-yr terms; (3) BTC doesn't crater mining cash flow before the AI revenue arrives; (4) the Swedish VAT case is won; (5) dilution slows. Break any one and the re-rate reverses.
• Growth disappoints 20–30%: if HPC ARR stalls at ~$35–50M (no Gigafactory financing), the stock is a sub-$2 BTC miner again — i.e., roughly the April-2026 price before the AI narrative.
• Single scenario that permanently impairs: a sustained BTC bear + closed capital markets forces a halt on the Gigafactory and a PP&E impairment on stranded ASIC capacity, while the VAT clawback hits — a balance-sheet and narrative double-break. Plausibility: moderate — it requires BTC weakness, which is the base rate for this asset class over any 18-month window.

Lens 14 · Management Questions (ordered by information value)

1. The GTA Gigafactory needs ~CAD $3.5B. What is the committed financing structure today — how much project debt is term-sheeted, at what cost, and how much further equity dilution does full buildout imply at the current share price?
2. Have you signed (or do you have an LOI with) any hyperscaler/enterprise anchor tenant for the 320 MW Gigafactory? What % of the 100,000 GPUs is contracted vs speculative?
3. Q4 FY26 Adjusted EBITDA was negative. At what BTC price and network difficulty does the mining segment go cash-flow-negative, and what's the downclock/curtailment plan if BTC trades below that through 2027?
4. You sold ~2,000 BTC to fund capex. What is the go-forward treasury policy — do you hold mined BTC, or remain a flow-seller, and how do you think about that opportunity cost?
5. Walk through the path from ~$35M HPC ARR to $660M by 2028 quarter-by-quarter — what has to be financed, built, and signed at each step, and what's the critical-path risk?
6. On the Swedish VAT case (~$80.5M, now at the Supreme Administrative Court): what is the realistic timeline and the cash impact if you lose, and why is it unprovisioned given two adverse rulings?
7. What returns (unlevered IRR / yield-on-cost) do you underwrite for converting mining sites (New Brunswick 70 MW, GTA) to AI colocation, and how do they compare to just running ASICs?
8. How dependent is the thesis on Paraguay/ANDE? What's the contingency if ANDE raises tariffs or curtails power to the 307 MW there?
9. The $115M exchangeable notes mature/convert at a $4.92 cap. What's the refinancing/repayment plan, and what other non-dilutive capital is available before you return to the ATM?
10. NVIDIA GPU allocation gates the AI pivot — what is your committed B200/Blackwell (and successor) allocation for 2026–27, and through which channel (Dell/Supermicro/direct)?
11. What is current insider ownership, and have insiders been net buyers or sellers through the 2026 AI-narrative run-up?
12. How do you defend HPC pricing/margin against the neoclouds (IREN, CoreWeave-class) that have anchor tenants and cheaper capital than HIVE?
13. What is the realistic 2027 depreciation load, and at what point does the PP&E base face impairment-testing pressure under a flat-BTC scenario?
14. The "non-event" framing of the Q4 drop: what specifically in the quarter do you think the market mispriced, and what KPI should investors track to verify the AI pivot is on track?
15. What is the single milestone in the next 12 months that, if missed, would cause you to pause or rescope the Gigafactory?

---