Rothschild Redburn initiates Hut 8 stock with neutral rating

Rothschild Redburn initiated Hut 8 coverage at Neutral with a $96 price target, below the stock's $98.74 trading price, citing adequately priced execution and pipeline-conversion risks. Hut 8 posted Q2 2026 adjusted revenue of $105.18M, below the $111.29M consensus forecast, although revenue grew 81% year over year on compute operations and Bitcoin production. The company is expanding from Bitcoin mining into AI-oriented data-center infrastructure, supported by a conditional Texas grid classification and an Nvidia capacity agreement, but remains highly volatile with a 5.98 beta.

Analysis

HUT's valuation is increasingly dependent on converting power rights and development projects into contracted AI/HPC cash flows, not on its legacy mining earnings. That creates a material timing mismatch: the market is assigning some data-center optionality today, while financing costs, interconnection milestones, customer commitments, and construction spend occur before recurring EBITDA. A DCF using a 25% cost of equity implies that even modest delays in utilization or capital spending can disproportionately reduce equity value; the current price leaves limited margin for a neutral execution outcome.

The key non-obvious issue is customer-quality and contract structure. An association with NVDA is not equivalent to a take-or-pay lease, a GPU purchase commitment, or a guaranteed revenue stream; investors should require disclosure of contracted MW, term, pricing, counterparty credit, and capex responsibility before underwriting AI-infrastructure multiples. Until then, HUT should trade as a high-beta hybrid of BTC sensitivity and speculative data-center development, meaning a BTC drawdown or higher long-end rates can compress the multiple regardless of project progress over the next 1-3 months.

Consensus upside targets appear to capitalize a successful conversion before evidence of stabilized margins and funded buildout. The contrarian view is not that the AI pivot fails, but that capital intensity and grid-conversion timelines make the value realization 6-18 months out; a revenue miss weakens the case for paying forward for that outcome. Falsification of the cautious thesis would be a binding, long-duration AI lease with disclosed economics, fully funded construction, and guidance showing compute EBITDA scaling faster than incremental corporate and interest expense.

Key Decisions for Investors

• Initiate a small short HUT / long IREN pair over the next 1-3 months, sized beta-neutral rather than dollar-neutral. IREN offers similar BTC and AI-data-center exposure with less reliance on unverified project optionality; target 15-20% relative outperformance, with a stop if HUT announces a funded take-or-pay contract covering a material share of planned capacity.
• Do not use NVDA as a read-through beneficiary. Maintain any NVDA exposure on GPU demand and hyperscaler capex fundamentals, but treat the HUT relationship as immaterial unless NVDA discloses a revenue-bearing supply, hosting, or capacity commitment.
• For existing HUT longs, reduce into strength above the neutral valuation reference and retain only optionality-sized exposure until the next earnings release or project financing update. Rebuild only after management discloses contracted MW, expected in-service dates, utilization assumptions, and the funding source for remaining capex.
• Set a catalyst watch for BTC weakness and real-rate increases: a 15% BTC decline or a renewed rise in 10-year real yields would likely expose HUT's dual valuation sensitivity before operational milestones can offset it. Conversely, a credible grid approval plus contracted AI revenue is the trigger to cover the short leg rather than mechanically averaging.