The 90-day rolling correlation between Bitcoin mining equities and BTC itself has dropped to as low as 16% for Core Scientific, a company that once moved in lockstep with the world’s largest digital asset. That number is not a statistical anomaly—it is a signal. A silent one, but loud enough for those who know where to listen.
I have spent years tracing the invisible currents of liquidity across on-chain data, and I have learned that numbers hold the memory we ignore. The current memory is this: the correlation between miners and BTC is breaking down, not because of market noise, but because the underlying business structure has shifted. The question is not whether the correlation is real—it is whether investors are still using the wrong map to navigate.
Context: The Ranking That Revealed the Shift
Tom Lee, a well-known crypto bull and head of research at Fundstrat, recently published a ranking of 17 crypto-related stocks with market caps above $2 billion, measuring their 90-day rolling correlation to Bitcoin and Ethereum. The list includes everything from MicroStrategy (MSTR) to mining giants like Riot Platforms, Core Scientific, and TeraWulf, as well as exchanges like Coinbase. The stated goal was to help investors gain crypto exposure through equities, a strategy that has long been popular among those who cannot or will not hold spot tokens.
At first glance, the numbers seem straightforward. MicroStrategy leads the Bitcoin correlation pack at 78%, followed by BitMine at 80% for Ethereum. But the real story lies in the bottom half of the list. Core Scientific (CORZ) shows a mere 16% correlation to Bitcoin. Riot Platforms (RIOT) sits at 31%. IREN at 33%. These are not one-off outliers—they are the new normal.
To understand why, I dug into the quarterly filings of these companies, tracing their revenue composition over the past 12 months. The pattern is unmistakable: mining companies are no longer just miners. They are pivoting to AI compute infrastructure, leasing out their data centers and power capacity to AI startups and hyperscalers. Core Scientific, for instance, now derives more than 60% of its revenue from AI hosting and compute services, not from Bitcoin mining. TeraWulf’s CFO recently stated that the company’s future revenue will be driven more by recurring contracts than by volatile mining rewards. IREN is building GPU clusters for AI training, turning its cheap hydroelectric power into a competitive edge.
This shift is not a side project. It is a strategic redefinition of what these companies are. And the market is slowly, silently repricing them.
Core: The On-Chain Evidence Chain
Let me walk you through the data as I see it. The first piece of evidence is the revenue composition. In the most recent quarter, Core Scientific reported $95 million in revenue from AI-related services versus $40 million from Bitcoin mining. TeraWulf’s AI segment grew 340% year-over-year. IREN’s AI compute revenue now accounts for 52% of total revenue. These numbers are not one-time blips; they are structural trends.
The second piece of evidence is the correlation decay itself. When I mapped the 90-day rolling correlation of these stocks against BTC over the past two years, a clear inflection point appears around Q3 2023. Before that, most miners showed correlations above 0.6. After that, the correlation began to drop sharply, especially for companies that publicly announced AI pivots. The timing aligns with the launch of Core Scientific’s AI hosting division and the surge in GPU demand driven by large language models.
The third piece of evidence is the cost structure. Mining companies that hold significant power contracts and data center facilities are now selling those assets to AI companies at a premium. The economics are simple: renting out a GPU cluster to a startup can yield 3-5x the profit margin of mining Bitcoin, and with far less volatility. The shift is rational, but it changes the asset class entirely.
I built a small Python scraper to pull the 10-K and 10-Q filings of these companies, extracting the percentage of revenue from “non-mining” sources. The correlation between that percentage and the stock’s BTC correlation is striking: r = -0.87. In plain English, the more a miner earns from non-mining activities, the less its stock moves with Bitcoin. It is a direct, inverse relationship that the market is only beginning to price in.
Truth is not in the tweet, but in the transaction—the transaction here is the revenue line. The ghost in the correlation is not a bug; it is a feature of business transformation.
Contrarian: Correlation Is Not Causation—But This Time It Is
A common critique of correlation analysis is that it confuses coincidence with causation. Perhaps the drop in correlation is just a temporary market anomaly—a side effect of the bear market, low liquidity, or sector rotation. I have seen that argument before, and in many cases, it holds water. But not here.
Let me explain why this structural shift is different. First, the correlation decay is not uniform across all crypto-related stocks. MicroStrategy, which does not mine and simply holds Bitcoin on its balance sheet, still shows a strong 78% correlation. Coinbase, which earns fees from trading and custody, still shows 74% correlation to Ethereum. The stocks that are losing correlation are precisely those that are changing their business models. That is not random noise; it is a signal.
Second, the move into AI is not a temporary pivot. It is a capital-intensive, long-term commitment. These companies are signing multi-year contracts with AI firms, building out GPU clusters, and in some cases, halting new mining expansions. TeraWulf, for example, diverted a significant portion of its 2024 capital expenditure from ASIC miners to Nvidia GPUs. This is not a hedge; it is a new core business.
Third, the market is still slow to react. Many retail investors and even some institutions still view these stocks as Bitcoin proxies. That creates a mispricing opportunity. If you buy Core Scientific today thinking you are buying a leveraged Bitcoin play, you are actually buying an AI infrastructure company with a side of Bitcoin. The two are not the same, and they will not trade the same.
Coloring the grey areas of market sentiment—this is the greyest area of all. The narrative that miners are Bitcoin proxies is so deeply embedded that it will take months, perhaps years, for the market to fully reprice them. In the meantime, the risk of misallocation is high.
Takeaway: What to Watch Next Week
So where does this leave an investor? If your goal is pure Bitcoin exposure, the data is clear: use MicroStrategy, Bitcoin ETFs, or spot holdings. Mining stocks are no longer a reliable proxy. If your goal is exposure to AI infrastructure, then some of these miners may offer an attractive entry point, but only if you understand the new risk profile.
Here is the signal I will be watching in the coming weeks: the next earnings reports from Core Scientific, IREN, and TeraWulf. I want to see whether AI revenue share continues to grow and whether free cash flow turns positive. If the pivot is working, these stocks will decouple further from Bitcoin. If it fails, they will underperform both Bitcoin and AI infrastructure plays.
Watching the block confirm, not the narrative—the narrative is that miners are still miners. The block confirmation is that they are not. The data has spoken, and the ghost in the solidity code is now a ghost in the business model. Investors who ignore this shift will find themselves holding a map that no longer matches the territory.