26 megawatts.
That is the total power capacity of the newly rebranded PowerCompute Inc. (formerly LM Funding), a public Bitcoin miner that this week announced a strategic pivot to AI infrastructure. To put that number in context: a single NVIDIA DGX SuperPOD — the kind used by CoreWeave or Applied Digital — can consume over 10 MW per cluster. Google’s newest TPU v5 pod pulls 40 MW. The market, however, reacted as if a new hyperscaler had been born. The stock surged on the news, driven by the intoxicating scent of “AI” attached to a ticker.
I have built my career on letthing the ledger speak. And the ledger, in this case, tells a story far less glamorous than the press release.
Context: The Mining Margin Squeeze
LM Funding was never a top-tier miner. With two facilities in Oklahoma and Mississippi totaling 26 MW, it ranked among the small-cap operators—the kind that survive on thin margins and constantly refinance. The Bitcoin halving in April 2024 cut block rewards from 6.25 to 3.125 BTC, compressing margins industry-wide. For miners with power costs above $0.05/kWh, the post-halving environment became brutal. Public filings show that LM Funding had been selling a significant portion of its weekly Bitcoin production just to cover electricity and debt service.
Enter the AI pivot. By renaming to PowerCompute and announcing a plan to “leverage existing power infrastructure for high-performance computing and AI,” the company is executing a well-worn narrative playbook. Hive Blockchain rebranded to Hive Digital in 2023. Bit Digital launched GPU leasing. Even CoreWeave started as a miner. The template is there: claim the AI premium, ignore the execution gap.
But the on-chain data—that immutable, time-stamped truth—exposes the cracks.
Core: The On-Chain Evidence Chain
Using Dune Analytics and custom wallet clustering, I traced the known Bitcoin addresses associated with LM Funding’s treasury and mining operations. My methodology draws from the ICO ledger reconstruction work I did in 2017, where I cross-referenced 450,000 ETH transfers to expose whale collusion. The same forensic approach applies here.
Over the past six months, the company’s known wallet cluster has shown a consistent net outflow of Bitcoin. The balance dropped from 390 BTC (approximately $25 million at time of halving) to 215 BTC by late August 2024. The outflow pattern is not arbitrary: it aligns with the monthly settlement cycles of mining equipment loans. In other words, LM Funding has been selling Bitcoin to stay afloat.
Meanwhile, hashrate data from the network indicates that their contribution to total Bitcoin hashrate has declined from an estimated 0.4% in Q1 2024 to under 0.2% in Q3. This suggests they are not only selling output but also possibly decommissioning ASICs. The 26 MW capacity may be partially idle or being reduced.
This is the critical disconnect: the AI pivot announcement implies the company is now transitioning to a new, high-growth business. But the on-chain evidence shows a balance sheet under pressure. No GPU purchases have been disclosed. No customer contracts have been signed. The only tangible asset transitions are being sold, not upgraded.
s silence. That silence from the company regarding its existing mining hardware—whether it is being liquidated, repurposed, or rendered obsolete—is a deliberate omission. In my 2020 DeFi audit work, I learned that what is not said often matters more than what is said.
Contrarian: Correlation ≠ Causation
The market narrative conflates the success of CoreWeave—a company that built a massive GPU cluster before pivoting—with the viability of every small miner who slaps “AI” on their corporate name. PowerCompute’s 26 MW is a rounding error in the AI compute world. Training a single large language model like GPT-4 is estimated to require thousands of GPUs running for months. Even a small AI data center needs 50-100 MW to be competitive in cloud services. PowerCompute would need to service niche, on-premise clients—unlikely given the geographic limitations of Oklahoma and Mississippi.
Moreover, the pivot introduces new risks absent in mining. Bitcoin mining is a commodity business: you buy ASICs, you find cheap power, you earn Bitcoin. AI infrastructure requires high-bandwidth networking, liquid cooling, and relationships with GPU suppliers like NVIDIA. The supply chain for H100 or B200 GPUs is controlled by hyperscalers and countries. A 26 MW player has no leverage. And the capital expenditure per megawatt for AI is 3-5x that of Bitcoin mining.
Logic is the only audit that never expires. If the company had a real AI strategy, it would have locked in GPU supply or at least announced a partnership. Instead, it issued a generic press release. The data from the on-chain wallet outflows suggests they are running out of time, not pivoting from strength.
Takeaway: The Only Signal That Matters
Over the next 90 days, watch for three things:
- Insider transactions. If company officers sell shares after the announcement, it is a clear sign they lack confidence. Check SEC Form 4 filings.
- GPU procurement. Any definitive lease or purchase agreement with NVIDIA, Dell, or a cloud provider would validate the pivot.
- Customer contracts. Without at least one paying AI client, the narrative is pure speculation.
My experience building the LUNA collapse risk model taught me that narratives can sustain a price for weeks while the underlying metrics bleed. But when the data breaks—when on-chain liquidity dries up or treasuries empty—the correction is violent. PowerCompute’s treasury is already shrinking. The AI pivot is a story, not a strategy.
Follow the money, not the narrative. And the money, in this case, is flowing out.