HIVE · HIVE Digital Technologies Ltd. · Stock research

https://orbyd.app — research note

HIVE — HIVE Digital Technologies Ltd.

Current Thesis

The leg on offer since June was the bitcoin-miner-to-AI-compute re-rating with a named counterparty behind it: BUZZ HPC's USD $220M, three-year sovereign-AI contract with Bell AI Fabric for Cohere, closed 2026-06-18, 2,304 NVIDIA Grace Blackwell GPUs, roughly $70M of incremental ARR once energized. That leg has been repriced out. The last completed daily close was $2.84 on 2026-08-07, 59.2% below the $6.96 52-week high, with the three-month return at +0.7% — the entire June spike has round-tripped to where the shelf sat in early May. What separates this from an ordinary miner drawdown is the backdrop: per Hashrate Index's 2026-08-03 roundup, bitcoin was near $63,160 and up about 1.6% over the prior 30 days, with USD hashprice around $32.10/PH/s/day after a rebound to a four-month high, and July network mining revenue at $875.35M, $38.93M above June. INFERRED, and stated as inference: with the commodity input stable-to-better through July, the decline from $3.79 (2026-06-25) to $2.84 is idiosyncratic — funding structure and cash burn, not BTC beta. The fiscal Q1 print for the quarter ended 2026-06-30 is the next event that can put a number on whether contracted ARR is actually compounding.

Bullish and bearish views on HIVE Digital Technologies Ltd.

The model's bull view on HIVE Digital Technologies Ltd. (HIVE), in brief: Contract is signed, not projected (2026-06-18): the $220M Bell AI Fabric / Cohere agreement is a three-year commitment with 2,304 Grace Blackwell GPUs at the Merritt, BC site, adding roughly $70M ARR on energization in late-2026/early-2027, on top of the $35M contracted HPC ARR… The bear view: The June structure failed outright: $3.79 on 2026-06-25 to $2.84 on 2026-08-07 is a full retrace of the Bell/Cohere breakout, and the drawdown occurred while bitcoin held near $63,160 (2026-08-03) and hashprice recovered to a four-month high. Both cases follow in full.

Bull Case

• Contract is signed, not projected (2026-06-18): the $220M Bell AI Fabric / Cohere agreement is a three-year commitment with 2,304 Grace Blackwell GPUs at the Merritt, BC site, adding roughly $70M ARR on energization in late-2026/early-2027, on top of the $35M contracted HPC ARR disclosed with the 2026-06-02 fiscal-year report.
• A second anchor tenant (2026-06-25): a long-term HPC colocation agreement with a Swedish investment-grade client at Boden contemplates 25 MW of critical IT load and up to 10,000 GB300 GPUs — a second named demand source rather than a repeat of the same contract.
• Scale actually delivered in FY2026 (2026-06-02): revenue $297.8M, +158% YoY; hashrate from 6.5 EH/s to 25.1 EH/s; 2,885 BTC mined vs 1,414 in FY2025 (+104%); gross operating margin $107.9M (+329%); adjusted EBITDA $72.9M.
• Funding was cheaper than the ATM fear implied (2026-06-30): the exchangeable note offering closed upsized at $130M, 0% coupon, due 2031, ~$124.5M net proceeds — no cash interest, with the exchange price at $4.83/share and a capped call lifting effective dilution to $8.5275.
• Independent hardware validation (2026-06-22): a Columbia University study run on HIVE's Paraguay A40 cluster and submitted to NeurIPS reported normalized performance in line with H100s — third-party evidence the fleet benchmarks as marketed.
• New sell-side sponsorship (2026-07-27): Chardan Capital initiated with a Buy and a $7.50 price target, against the $2.84 close on 2026-08-07.

Bear Case

• The June structure failed outright: $3.79 on 2026-06-25 to $2.84 on 2026-08-07 is a full retrace of the Bell/Cohere breakout, and the drawdown occurred while bitcoin held near $63,160 (2026-08-03) and hashprice recovered to a four-month high. A stock that cannot hold a gain on its best contract news, into a stabilizing commodity, is being sold for reasons the contract does not address.
• Issuance capacity sits above the tape: the equity distribution agreement was amended up to $300M on 2026-06-22 with roughly $214.7M remaining. That capacity is undrawn optionality for the company and overhead supply for the shareholder, and it grows more dilutive the lower the print.
• Two hard reference prices overhead: $4.83 (note exchange price, 2026-06-30) and the June shelf near $3.20. Both sit above the 2026-08-07 close, and convert holders are structurally short-delta into rallies.
• Margin of the core business is thin: at roughly $32/PH/s/day hashprice, industry commentary in early August put many operators at or below breakeven depending on machine fleet and power cost. Mining cash flow is what funds the AI capex, and it is being generated at the bottom of the cycle.
• The last print missed on the headline (2026-06-02): fiscal Q4 revenue $71.8M against $78.9M consensus and GAAP EPS of $(0.30); the adjusted $0.02 beat did not carry the tape.
• Energization risk is the whole ARR bridge: the ~$70M Bell/Cohere ARR and the $660M-by-2028 target both depend on power and buildout timelines (Yguazú's 100 MW substation targeted ~Sept 2026, GTA Gigafactory at 320 MW) that have not yet been demonstrated on schedule.

Setup & Price Structure

Life-cycle label: DEAD — the narrative leg failed and the structure broke. What dates it: the 2026-06-18 contract marked the high-water mark of attention, price stalled at $3.79 by 2026-06-25, and by 2026-08-07 the close was $2.84, 59.2% under the $6.96 52-week high with the June gap fully filled. The operating story is still advancing; the price leg built on it is not. The label describes the trade, and it can be re-earned — a reclaim of the June shelf on volume, with a Q1 ARR figure above the $35M+$70M framework, would open a new leg rather than resume the old one.

Positioning and crowding observables, stated as observables:

• RSI(14) at 43.1 on 2026-08-07 — mid-range. There is no oversold extreme here and no evidence of forced liquidation; this is drift, and drift does not mark bottoms on its own.
• Three-month return +0.7% against a 59.2% drawdown from the high: the round-trip is complete, which means every buyer from the June headline window is now underwater at $2.84.
• Sell-side coverage arrived after the drawdown, not before it — Chardan's Buy/$7.50 initiation is dated 2026-07-27, five weeks past the contract that framed the story.
• ~$214.7M of remaining equity-distribution capacity as of 2026-06-22 is issuance-into-strength capacity that a reader should assume is live on any rally.
• An unconfirmed earnings date inside the next two weeks is the near-term binary; there is no basing pattern to lean on ahead of it.

Catalyst Calendar (next 30 days)

• ~2026-08-18 (est., unconfirmed): fiscal Q1 FY2027 results, quarter ended 2026-06-30. The number that matters is BUZZ HPC contracted ARR versus the $35M disclosed 2026-06-02 and the ~$70M forward from Bell/Cohere, plus any restatement of the ~$140M annualized BUZZ run-rate targeted for Q4 2026 at ~80% gross margin.
• ~2026-09-03 (est.): monthly bitcoin production and BUZZ HPC update. HIVE publishes these in the first week of the month; the series carries hashrate, BTC mined and, when management chooses, the ARR line.

Elapsed catalysts

• ~Sept 2026 (company guidance, 2026-06-02): Yguazú, Paraguay 100 MW substation energization — the first schedule checkpoint on the buildout the ARR bridge assumes. (passed 68d ago)
• Ongoing: any 6-K disclosing draws against the $300M equity distribution agreement (~$214.7M remaining as of 2026-06-22) would date the dilution the market is already pricing. (passed 48d ago)

What Would Change Our Mind

The structure that has to come back first is the June shelf near $3.20 — the level that failed after the Bell/Cohere close and the one a genuine re-rating would have to reclaim and hold on expanding volume, ideally with the Q1 print showing contracted ARR stepping up rather than restating $35M. On the downside the gradeable break is a weekly close below $2.70, which puts the stock beneath the early-May base it has now round-tripped to and confirms that the AI-compute story is not supporting a valuation floor even with bitcoin stable near $63,160 (2026-08-03) and hashprice at four-month highs. Two further conditions would each independently change the read: an ARR figure at the ~2026-08-18 (est.) print that is flat against the June framework, which removes the compounding argument entirely; or a disclosed ATM draw large enough to consume a material share of the ~$214.7M remaining, which converts the dilution risk from overhang to fact. Conversely, a Q1 print that puts BUZZ HPC revenue on a visible path toward the ~$140M annualized target, paired with a weekly close back above $3.20, would argue the leg is re-forming rather than finished.

Correlation Notes

• Bitcoin, with leverage: mining cash flow scales with BTC and hashprice. The July decoupling matters — BTC +1.6% over 30 days into 2026-08-03 while the stock fell — because it shows the equity discount is not currently a BTC problem, and it means a BTC rally alone may not repair it.
• Miner-pivot basket: IREN, CIFR, WULF, APLD and CORZ trade the same miner-to-AI-compute story. HIVE has been the laggard through this cycle; the basket's leaders tend to move first, so relative strength versus that group is the cleaner read on whether the theme itself is being bid.
• NVIDIA supply chain and sovereign-AI policy: the ARR bridge depends on Grace Blackwell and GB300 deliveries and on Canadian sovereign-AI demand routed through Bell and Cohere. A change in Canadian federal AI-compute procurement, or GPU allocation slippage, hits the contract pipeline directly.
• Power and datacenter build cycle: substation energization dates (Yguazú ~Sept 2026) and the Boden, Sweden 32 MW site put HIVE on the same timeline risk as any datacenter developer — capex is spent before revenue arrives.
• Currency and listing: dual-listed on Nasdaq in USD and the TSXV in CAD; US analyst targets are USD, and CAD/USD moves shift the reported spread between the two lines.