HIVE expands AI data center capacity to 16.6 MW across Canada

HIVE increased liquid-cooled AI data center capacity in Canada from 4 MW to 16.6 MW (including a 5 MW BC facility with an option for +7.6 MW in 2027) and targets deployment of 4,000 GPUs and $200M in contracted annualized run-rate revenue by March 31, 2027. The company reported LTM revenue of $257.14M (+112.5%) and LTM EBITDA of $37.97M but remains unprofitable at a $0.62 loss per share and faces noted cash burn, with the stock down 45.1% over six months. Several broker price targets were lowered (Cantor Fitzgerald to $5.00, Rosenblatt to $4.50, H.C. Wainwright to $7.00) even as HIVE phases down ASIC mining in Sweden and transitions the Boden site to NVIDIA GPU infrastructure. Management granted 2,849,400 RSUs and is targeting ~75% EBITDA margins on new long-term GPU contracts.

Analysis

Winners are those that supply and enable dense GPU estates rather than the colo operator itself: server OEMs and liquid-cooling / power-ops vendors gain pricing power and order visibility ahead of colo operators that front capital-lite marketing. Second-order winners include secondary markets for used ASIC miners and legacy hosting providers who can tout rapid, lower-risk conversions of racks to AI workloads; conversely, spot GPU cloud marketplaces and hyperscalers (who can undercut long-term colo pricing) are indirect competitive threats.

Key risks cluster around execution and unit economics rather than market demand: margin assumptions baked into small-cap GPU-hosting stories are fragile if contract duration, power pass-throughs, or pricing floors shift even a few percentage points. Supply-chain constraints (high-density PDUs, chillers, and rack-level networking) create timing risk — a multi-month delay in GPU deliveries or in commissioning high-density pods materially compresses throughput and defers contracted revenue recognition.

Near-term catalysts to watch (days–months) are: quarterly bookings cadence, disclosed long-term contract lengths and customer logos, and any NVDA supply cadence updates that change procurement lead times. Medium-term (6–18 months) inflection will come from demonstrated FCF generation per deployed GPU and any material margin drift once spot vs contracted yields normalize; legal/tax disputes in foreign jurisdictions introduce binary tail events that can delay conversions or raise capex.

The consensus is focused on headline capacity growth; it underweights the asymmetric sensitivity of unit economics to power, GPU pricing, and contract tenure. That makes the equity trade more binary — either the company proves sticky, margin-accretive contracts quickly, or valuation re-rates sharply if real cash generation lags bookings rhetoric.

Key Decisions for Investors

• Pair trade (6–12 month horizon): Short HIVE equity sized 3–5% portfolio weight vs long SMCI 2–3% (hardware exposure). Rationale: HIVE faces execution/cash-risk while SMCI captures durable OEM demand; target 40–60% relative return, stop-loss at 25% adverse move.
• Options hedge on NVDA (9–12 month): Buy NVDA 2026 Dec 1.5–2x call spread to play tight GPU supply and secular demand without full delta exposure. Cost-limited upside with 3–5x asymmetric payoff if supply tightness persists; cap loss to premium paid.
• Event-driven long HIVE call (6 months) as a tactical, limited-risk punt: Buy OTM calls (small notional, <1% portfolio) with strike ~20–30% above spot expiring after next booking cadence; exit if no material contracted revenue or margin disclosure. This captures upside if bookings convert while capping downside to premium.
• Risk management: set a portfolio-level stop where combined exposure to small-cap GPU hosts (equity + options) does not exceed 5% unhedged; reassess after next quarter’s customer-level disclosures and any NVDA supply updates.