The news hit the terminal at 9:17 AM EST: LM Funding, a small-cap Bitcoin miner with two facilities totaling 26 megawatts in Oklahoma and Mississippi, was renaming itself to PowerCompute Inc., ticker PWCM. The press release spoke of a “strategic pivot” into high-performance computing and AI infrastructure, leveraging existing power assets. Within hours, retail traders on Reddit were already calling it “the next CoreWeave.” But as I sat down to parse the filing—having spent years auditing ICO whitepapers and DeFi protocols where narratives often outpaced reality—I felt a familiar chill. The hype was writing a check that the technical stack couldn’t cash.
Context: The Bitcoin Miner-to-AI Playbook
The playbook is well-worn by now. In 2020, Hive Blockchain rebranded to Hive Digital Technologies. In 2023, Applied Digital pivoted from mining to hosting AI workloads. CoreWeave, the poster child, transitioned from Ethereum mining to become a multi-billion-dollar AI cloud provider. The logic is seductive: Bitcoin miners already own land, power infrastructure, cooling, and electrical substations. Why not plug in GPUs instead of ASICs? The market rewards the narrative: AI infrastructure carries higher multiples than Bitcoin mining, which is volatile and commoditized.
But there is a vast gulf between having a 26MW power capacity and operating an AI-ready data center. Based on my own experience auditing mining operations during the 2021 bull run, I remember visiting a facility that boasted 50MW of “available power.” The reality: the transformers were sized for ASIC’s steady 65W per hashboard, not for the bursty, high-density demands of NVIDIA H100s that pull 700W each and require liquid cooling. The gap between “power capacity” and “usable power for AI” is as wide as the gap between a pickup truck and a Formula 1 car—both have engines, but you can’t race the truck at Monaco.
Core: The 26MW Reality Check
Let’s run the numbers. A 26MW facility, assuming a power usage effectiveness (PUE) of 1.3 (optimistic for a retrofitted mining site), leaves about 20MW for IT load. An NVIDIA H100 GPU draws around 700W under load. That means you can theoretically fit about 20,000 / 0.7 ≈ 28,570 H100s—but that’s before networking, storage, and cooling overhead. In practice, a more realistic density for a retrofitted site with air cooling is around 10-15 H100s per rack, at 15kW per rack. That gives you about 1,300 racks, each pulling ~15kW, for a total of ~19.5MW IT load. That’s not a supercomputer cluster; that’s a small-scale inference farm. For comparison, CoreWeave’s new facilities exceed 500MW. PowerCompute’s 26MW is a rounding error in the AI compute market.
But the bigger issue is hardware procurement. As of early 2026, NVIDIA’s B200 GPUs (the successor to H100) have lead times of 6-12 months for new customers. Tie that to a small-cap miner with no prior relationship with NVIDIA, no reseller agreement, and a market cap under $50 million—good luck getting on the BOM list. Companies like AWS, Google, and CoreWeave have pre-committed billions of dollars in GPU purchases. PowerCompute has… a press release. “We plan to offer infrastructure services to AI compute clients,” said the company in a boilerplate statement. There was no mention of GPU purchase agreements, no mention of a cloud partnership, no hint of a customer pipeline. The ledger remembers what the hype forgets: without silicon, you have nothing to sell.
Contrarian: The Hidden Risk of Double Dilution
What the market is ignoring is that PowerCompute is not just a pure-play AI pivot; it is a dual-business model that could dilute shareholder value from both sides. The company stated it will continue to hold its Bitcoin assets on the balance sheet. Say Bitcoin drops 30%—the mining side loses value, but the AI side is still unproven. Conversely, if the AI pivot succeeds, management will likely need to raise capital to buy GPUs, issuing new shares that dilute existing shareholders. In fact, the most likely path for a small miner is to sell Bitcoin holdings to fund GPU purchases, which would remove the crypto asset that originally attracted investors.
This is the classic “neither fish nor fowl” trap. The company will be judged by two sets of investors: Bitcoin maximalists will see it as a “cash grab” that dilutes their exposure, while AI infrastructure investors will see it as a second-tier player without scale. The stock could fall into a valuation no-man’s land. Based on my conversations with three sell-side analysts during the DeFi summer, I learned that the worst thing a company can do is confuse its narrative. PowerCompute is doing exactly that.
Takeaway: The Sprint Ends, But the Chain Remains
In the short term, the stock may pop on the AI narrative—that’s the nature of the market. But the fundamental equation remains: 26MW of power, no GPU commitments, no clients, and a management team that hasn’t publicly demonstrated AI infrastructure expertise. “Bridging the gap between code and community” is my mantra, but here the gap is between a press release and a functioning AI data center. The real question is not whether AI infrastructure has a future—it does. The question is whether this particular miner will survive the transition. The ledger remembers what the hype forgets: execution is everything, and 26 megawatts isn’t enough to build a moat. As for PowerCompute, I’ll be watching the insider trading filings, not the press releases. If the C-suite starts selling before the next quarterly, you’ll know the game.
