The number is too clean. $9.8 billion for 352 megawatts at Beacon Point. That is not a lease—it is a lifetime subscription to a bet. Hut 8, the publicly traded Bitcoin miner turned AI infrastructure landlord, just signed the largest power agreement I have seen since my 2022 Terra post-mortem. The smoke is gone, but the mirrors remain.
Icebergs are not warnings; they are delays.
Context: The Narrative Loop The market reads this as metamorphosis: miner evolves into AI hyperscaler. The total contracted capacity now sits at 949 megawatts—a 2.7x jump. The premise is simple: Bitcoin mining margins are compressing post-halving, and AI hosting commands 30-50% gross margins. So Hut 8 pivots. The blueprint is Core Scientific's 2024 turnaround. The difference? Core had bankruptcy fresh in its boardroom. Hut 8 has a $9.8 billion liability with zero revenue commitment attached.
Hut 8 is a listed entity (NYSE: HUT). But listed does not mean liquid. A 10-year lease at $980 million per year? Even with 704 MW in play, the calculation smells of 10-20 year financing, likely a triple-net lease with annual escalators. The annual rent of $490 million to $980 million—compared to Hut 8’s trailing twelve-month revenue of roughly $250 million—is a leverage ratio that screams 'execution or collapse.'
Core: The Financial Dissection Let me save you the spreadsheet. The market will cheer capacity expansion. The tech press will write 'supply side growth.' But I see two red flags from my 2021 audit of a 'power purchase agreement' that turned into an infinite liability loop.
First, the unit economics. At 352 MW for $9.8B over 10 years, the per-megawatt-per-year cost is ~$278,000. For a typical data center, power costs alone should be 30-40% of that figure. The rest is land, building, cooling, and interconnection. If Hut 8 paid full market rate, they are paying a premium for strategic location. That premium only works if GPU rental prices stay high. Given the upcoming H200/B200 wave and massive supply overhang, margins will compress. The math breaks if utilization drops below 70%.
Second, the off-balance-sheet risk. Public companies love to hide long-term leases. $9.8B is a material commitment. If this is structured as an operating lease, it would not appear as debt on the balance sheet—only as a footnote. The stock could rally 20% before the next 10-Q reveals the accelerated cash flows. I saw the same trick in 2020 with Compound’s ‘liquidation threshold’ that I corrected using Hardhat. Numbers do not lie, but structures can delay the reveal.
Contrarian: What the Bulls Got Right The AI hosting thesis is not wrong. Core Scientific went from $0.40 to $10+ on the back of GPUs. Hut 8’s management may have secured an anchor tenant before signing—likely an enterprise AI startup or a cloud provider. The 704 MW at Beacon Point is ideally located in a grid with low carbon intensity, which satisfies both ESG and compliance requirements. If Hut 8 lands a 300 MW GPU contract at $12,000 per megawatt per month, the annual revenue from that alone is $43.2 million per 100 MW—healthy.
But the tail risk is asymmetric. If the AI market slows (or if NVIDIA’s Blackwell yields lower GPU demand), the empty floor space becomes a liability that compounds faster than a DeFi vault’s yield curve.
Check the inputs, ignore the hype.
Takeaway The market will treat this as validation. I treat it as a signal to examine cash flow statements and debt covenants. The code of public markets is no different from smart contracts: the logic can compile, but external variables—interest rates, GPU pricing, bitcoin volatility—can deconstruct the system in a single block.
A flat line is more dangerous than a spike.
Do not confuse capacity with competence. Hut 8 just played a high-risk card. Whether it folds or flushes remains unprovable until the first earnings call post-construction. Watch the revenue guidance. Ignore the press release. The compiler does not care about narratives.
