26 megawatts. That number matters.
LM Funding, a small-cap Bitcoin miner, just rebranded to PowerCompute Inc. New ticker: PWCM. Their play? Pivot from mining SHA-256 to hosting AI compute.
This article is the data behind that headline.
I've been in the analytics trenches for 12 years. I've seen pivots that changed industries—and pivots that changed nothing. PowerCompute falls into the latter category until proven otherwise.
Context: Why now?
The company was LM Funding, a publicly traded Bitcoin miner. Two sites: Oklahoma and Mississippi. Total power capacity: 26MW. On March 2024, they announced a name change to PowerCompute Inc. (PWCM) and a strategic shift.
"We plan to leverage our existing power infrastructure to enter the high-performance computing and AI infrastructure business," the CEO stated. "We will continue to hold our Bitcoin assets."
Sounds strategic. But strategic pivots require more than a press release.
Bitcoin mining farms use ASIC chips—single-purpose, air-cooled, power-intensive. AI data centers use GPUs—general-purpose, liquid-cooled, interconnect-intensive. The technology stack is completely different. The skills required are different. The customer base is different.
PowerCompute is essentially announcing they want to become a real estate company for GPUs. They provide the building, power, cooling. The tenant brings the GPUs? Or they buy GPUs? Unclear from the release. This ambiguity is a red flag.
Core: The data behind the hype.
Let's dig into the numbers.
Size: 26MW. A typical AI-driven data center built today is 100-200MW. CoreWeave's new facility in Texas is 200MW. Applied Digital's campus in North Dakota is 400MW. PowerCompute's 26MW is tiny. It can host maybe 2,500-3,000 H100s. That's less than 1% of the H100s NVIDIA ships in a single quarter.
GPU access: The AI GPU market is tight. Lead times are 12-24 months for large orders. NVIDIA prioritizes customers with proven demand. PowerCompute has zero track record. They might get GPUs through a broker or leasing arrangement, but that raises costs. Their 26MW could be fully utilized with used GPUs, but used GPUs are less efficient. The economics become marginal.
Facility conversion: Bitcoin miners use fans and filters. AI GPUs need precision cooling—liquid or high-density. Retrofitting 26MW could cost $5-10 million. That's significant for a company with a $20M market cap and $12M in Bitcoin. If they sell Bitcoin to fund conversion, they lose the "safety net." If they issue equity, dilution hurts shareholders. Debt services would consume cash flow.
Customer acquisition: AI compute clients are selective. They want low-latency interconnect (InfiniBand or RoCE), high reliability, and SLAs. Small facilities struggle. Large clients like Meta or OpenAI won't touch a 26MW site. PowerCompute might target smaller AI startups, but those startups are price-sensitive and may not survive. The churn risk is high.
Let's compare with successful pivots. Hive Digital (formerly Hive Blockchain) began transitioning in 2021. They bought GPUs, built a team, and now have legitimate AI cloud revenue. Their stock trades at a premium to mining peers. But Hive's facility is over 100MW. They started small but scaled. PowerCompute is starting small with no clear scaling plan.
I've lived through these transitions. In 2020, I ran manual arbitrage on Uniswap V2. I learned that execution beats narrative every time. A pivot announcement is just noise. The real signal is when you see GPUs arriving at a facility. Or when a contract is signed.
Arbitrage opportunities don't last. The gap between announcement and delivery will close fast.
Another data point: PowerCompute's Bitcoin holdings. They own about 200 BTC. At current prices, that's ~$12M. If they convert that to GPUs, they might get 2,000 H100s. But they lose their Treasury. They become solely reliant on AI revenue. Risk increases.
Alternatively, they could use Bitcoin as collateral for a loan. But loan rates are high (12-20% in crypto lending). That eats margins.
Hype is a trap; data is the only map I trust.
Let's look at the valuation impact. Before the announcement, LM Funding was trading at roughly 1.5x book value. Post-announcement, the stock might double or triple as "AI" multiple expands. But that multiple is temporary. Without execution, it will revert.
Contrarian: The unreported blind spots.
The contrarian angle: This pivot might be value-destructive.
Why? Because it dilutes focus. Bitcoin mining at 26MW can generate steady, predictable revenue. The hashprice is down, but efficient miners still make 30-40% profit margins. AI hosting at small scale might generate similar revenue but with higher capex and operational risk.
I recently analyzed a similar pivot in the 2026 NeuroTrade AI agent fiasco. That project claimed AI-powered trading bots. I traced on-chain wallet clusters and found synthetic volume—AI agents looping trades. The liquidity vacuum collapsed the protocol. The lesson: AI narratives without substance are dangerous.
PowerCompute's AI pivot is not a scam, but it's a narrative play. The CEO knows that "AI" gets a higher multiple than "mining." So they rebrand. The real question: Are they building a business or selling a story?
Another contrarian insight: The 26MW facility might be better suited for Bitcoin mining than AI. Let's compute: 26MW at $0.04/kWh (industrial rates in Oklahoma) costs ~$9M per year in electricity. Mining efficiently with S19XP (95TH, 27.5W/TH) yields ~10-12 PH/s per MW? Actually 1 PH/s consumes ~1 MW at 30W/TH. Rough estimate: 26MW can support ~26 PH/s. At current network hash rate of ~600 EH/s, that's 0.004% of network. Revenue per PH/s per day is about $55. So 26 PH/s gives ~$1,430/day or $43k/month. That's not great. Mining profitability is low.
AI hosting at $2/GPU-hour for 2,500 GPUs at 85% utilization yields ~$42,500/day or $1.3M/month. So AI revenue potential is 30x higher. If PowerCompute can achieve that, the pivot makes financial sense. But the risk is they won't secure GPUs or customers.
The contrarian trade? Wait for pullback. If the stock doubles on hype, short it. The narrative will fade.
Takeaway: The next watch.
My judgment: PowerCompute's pivot is a high-risk, high-variance event. The market will likely spike, but the real test is execution.
Three signals to track:
1) GPU acquisition: Any filing (8-K) about purchase or lease of H100/B200 GPUs. 2) Client contract: A named client with binding revenue commitment. 3) Insider buying: Form 4 filings showing management buying stock.
If none of these emerge within 90 days, the stock will retrace. The sustainable play is to wait for proof.
I don't trade narratives; I trade execution.
Data over drama. Always.