We audit the code, but who audits the business model? A few weeks ago, the WGMI ETF—a fund tracking Bitcoin miners pivoting to AI—was flying high, up over 100% from its lows. Then came the news: TeraWulf signed a $19 billion lease with Anthropic, CleanSpark locked in $6.6 billion, and Hut 8 was re-rated by Benchmark as a “power-first data center REIT.” The market cheered, then sold. WGMI dropped 34% from its peak. In crypto, we’re used to volatility, but this wasn’t a Bitcoin drawdown—it was a recalibration of faith. The question is not whether miners can pivot, but whether the narrative can survive the weight of its own assumptions.
Context This pivot is not a tech upgrade; it’s a resource arbitrage. Bitcoin miners built massive energy infrastructure to secure the network—gigawatt-level power draw, established grid connections, cooled facilities. After the fourth halving, miner revenue from block rewards and fees collapsed, making hashprice a brutal metric. Meanwhile, AI labs were desperate for compute. The match was natural: miners stopped selling hashpower and started leasing megawatts. Empery Digital, a hedge fund, sold its Bitcoin holdings to buy miner stocks, signaling a belief that infrastructure assets would be revalued. The logic is clean: rent your power to AI labs for 10–20 years at a fixed rate, and the market stops treating you as a commodity Bitcoin miner and starts pricing you like a real estate trust. But clean logic doesn’t mean safe reality.
Core: The Fragile Architecture of Assumptions I spent the first months of 2020 reverse-engineering the yield logic of Harvest Finance. I found that their “alpha” was built on unsustainable token emissions—growth at all costs with no floor. When the market woke up, the tokens collapsed. That experience taught me to look for the hidden assumptions in any business model. The miner-to-AI pivot rests on three assumptions, each more fragile than the last.
First, compute scarcity must persist. The entire lease structure is a bet that AI labs will continue to need massive, dedicated power for training. If open-source models like Llama or Qwen match or surpass closed models, the demand for proprietary compute could plummet. The market is already whispering this—articles from July highlighted how open-source progress is spooking investors. If compute becomes abundant, the long-term leases signed at today’s rates become liabilities.
Second, miners must execute like data center operators. I’ve stood in mining farms—rows of ASICs, simple cooling, basic networking. Running a GPU cluster for AI training is a different beast: high-density cooling, low-latency interconnects, 24/7 uptime requirements. The teams at TeraWulf and CleanSpark have strong mining backgrounds, but as I wrote in my 2022 newsletter “The Quiet Chain,” technical competence in one domain doesn’t transfer effortlessly. Hiring AI infrastructure talent is expensive and slow.
Third, the financial engineering must hold. A $19 billion lease over 20 years sounds enormous, but it’s a future stream of cash flows that gets discounted to present value. If Anthropic’s business falters, or if regulation cracks down on energy consumption, those leases could be restructured or broken. The market’s recent sell-off suggests that investors are starting to price in these risks. The 34% drop in WGMI isn’t random—it’s the sound of a narrative hitting reality.
Let me ground this in personal experience. During the bear market of 2022, I interviewed 50 women artists in the NFT space. They faced systemic bias, but also they saw NFTs as a direct path to income—no galleries, no gatekeepers. That human story is what blockchain should be about. But when I look at miners “pivoting,” I see an industry that is abandoning its core mission—securing a decentralized network—for a rent-seeking position in a centralized AI economy. The irony is thick. We audit the code, but we rarely audit the conscience of the business model.
Contrarian: The Centralization Trap Most analysts celebrate this pivot as a natural evolution. I see it as a centralization accelerant. In Bitcoin, miners compete globally; hashpower is distributed across thousands of operators. In AI, the customers are a handful of giant labs—Anthropic, OpenAI, Google DeepMind, Alibaba. By becoming landlords to these oligopsonists, miners are tying their fate to the very centralized power structures blockchain was meant to disrupt.
Furthermore, the pivot creates a dangerous dependency. If the AI bubble pops—and history suggests all hype cycles eventually correct—these miners will be left with power contracts designed for customers who can’t pay. Their Bitcoin mining rigs will have been mothballed or sold. The transition back to Bitcoin will be slow and costly. The contrarian truth is that the miner AI story is a leveraged bet on the continued explosion of AI capex, which itself is a bet on indefinite technological progress. That’s a fine hedge, but it’s not a bedrock.
Consider the open-source model threat. If a breakthrough algorithm reduces the compute needed for training by 90% (as some research suggests is possible), the decade-long leases become albatrosses. The market has already started differentiating: not all miners are bouncing back equally. Hut 8’s stock has held up better than some smaller players, precisely because it has a more credible plan and a stronger balance sheet. The differentiation is the beginning of the end for the broad-brush narrative.
Takeaway Build not for the peak, but for the plain. The best infrastructure is flexible, serving both Bitcoin mining and AI compute with equal efficiency. The miner companies that survive the next cycle will be those that don’t over-leverage on AI, that maintain a diversified energy portfolio, and that remember why they built power in the first place: to serve a decentralized future. The market is currently in the phase of “show me the revenue.” Those who can deliver real AI service income will thrive; those who only have a press release will be sorted out.
We audit the code, but who audits the conscience? As an evangelist, I believe in blockchain’s transformative power. But transformation must be rooted in integrity, not hype. The miner pivot is a fascinating experiment—but it is also a mirror reflecting our own willingness to believe in stories without checking the assumptions. I’ll be watching the next earnings calls not just for numbers, but for signs that these companies are building for the long haul, not just the peak.