The 90-day rolling correlation between Bitcoin and the largest publicly traded mining companies now sits between 16% and 33%. MicroStrategy, a company that does not mine a single block, posts 78%. This is not a statistical anomaly. This is a reclassification event.
Tom Lee's recent ranking of 17 crypto-related equities, published with a market cap threshold of $2 billion, was designed to answer one question: which stocks give investors the most efficient exposure to digital assets? The data answered a different question entirely. The mining sector has structurally detached from the asset it was built to produce.
I have spent the better part of a decade auditing protocol-level systems and analyzing how business structures map onto blockchain fundamentals. What the correlation data reveals is not a market inefficiency. It is a business model migration that has been underway since the 2022 bear market, accelerated by the AI infrastructure boom, and now visible in the price data.
The Hook: A Correlation Divergence That Demands Explanation
Core Scientific posts a 16% correlation with Bitcoin. Riot Platforms sits at 31%. IREN, the miner that remains closest to pure BTC exposure, manages only 33%. Meanwhile, MicroStrategy, a software company that converted its balance sheet into a Bitcoin treasury vehicle, delivers 78%.
These numbers are not noise. They are the output of a 90-day rolling window, which smooths short-term volatility and captures persistent structural relationships. When a mining company's stock price stops tracking the commodity it mines, something fundamental has changed in the revenue model.
The divergence is even more striking when compared to non-crypto benchmarks. Several mining equities now show lower correlation to Bitcoin than the Dow Jones Industrial Average. A stock that mines Bitcoin is less correlated to Bitcoin than a basket of 30 blue-chip industrial companies. That is not a statistical quirk. That is a business transformation.
The Context: What the Ranking Actually Measures
Tom Lee, the well-known market strategist, compiled a list of 17 crypto-related stocks and ranked them by their 90-day price correlation to BTC and ETH. The list spans the full spectrum of crypto equity exposure: treasury companies like MicroStrategy, exchanges like Coinbase, and a broad swath of mining operators including Core Scientific, TeraWulf, IREN, Riot Platforms, MARA, and CleanSpark.
The methodology is straightforward. Correlation measures whether two assets move in the same direction. A reading of 100% means perfect synchronization. A reading of 0% means no relationship whatsoever. The 90-day window ensures the calculation reflects recent market dynamics rather than historical averages that may no longer apply.
For ETH, the ranking produces different leaders. BitMine posts an 80% correlation to Ethereum, the highest on the list. Coinbase follows at 74%. These numbers make intuitive sense: Coinbase's revenue derives from trading volume, staking services, and institutional custody, all of which scale with Ethereum network activity.
But there is a complication embedded in the ETH data. Tom Lee serves as chairman of BitMine, the very company that ranks first in his ETH correlation analysis. This is not an automatic disqualification of the data. It is, however, a governance flag that any serious investor must acknowledge before acting on the ranking.
The Core: Business Structure Migration and the Death of the Mining Proxy
The correlation collapse is not a market inefficiency waiting to be arbitraged. It is the direct consequence of a deliberate strategic pivot by mining companies, executed over the past 24 months, and now reflected in their financial statements.
The mechanics are simple. Mining companies possess three assets that AI companies desperately need: cheap electricity, physical warehouse space, and high-capacity power infrastructure. These assets, originally acquired to run ASIC miners for Bitcoin, are now being repurposed to host GPU clusters for AI training and inference workloads.
The revenue implications are profound. Renting compute to AI companies generates more revenue per megawatt than mining Bitcoin. It also generates more predictable revenue. AI contracts are typically structured as recurring agreements with defined terms, whereas mining income fluctuates with Bitcoin price, network difficulty, and halving events.
Core Scientific, which emerged from Chapter 11 bankruptcy in early 2024, now derives a significant portion of its revenue from AI hosting agreements. TeraWulf's CFO has stated publicly that the business will increasingly be driven by recurring contract income rather than spot mining revenue. IREN, while maintaining the highest BTC correlation among miners, has still diversified into AI compute services.
The financial data confirms the shift. MARA and CleanSpark, two miners that pursued AI transition aggressively, have recorded combined losses of $851 million. These are not operational failures. They are capital allocation decisions. The companies are spending heavily on infrastructure that will generate AI revenue in future quarters, accepting current losses in exchange for a more stable, higher-multiple revenue stream.
This is where the analysis must be precise. The correlation decline does not mean mining companies are performing poorly. It means their stock prices are now driven by different factors. When an AI company signs a large hosting contract with Core Scientific, the stock moves on that news, not on Bitcoin's price action. When electricity prices rise in Texas, TeraWulf's margins tighten regardless of what BTC does.
The valuation framework is shifting accordingly. Mining stocks are increasingly being priced like data center REITs or AI infrastructure plays, not like commodity producers. The market is assigning recurring revenue multiples to the AI portion of their business and traditional mining multiples to the residual crypto exposure. This is a rational repricing, not a mispricing.
The Contrarian Angle: Blind Spots in the Reclassification Narrative
Three blind spots emerge from this analysis, and each carries material risk for investors who accept the reclassification narrative without scrutiny.
First, the conflict of interest embedded in the ranking itself. Tom Lee's dual role as ranking publisher and BitMine chairman creates an inherent incentive structure that cannot be ignored. The data may be perfectly accurate. The methodology may be sound. But when the person publishing the analysis has a direct financial interest in one of the ranked companies, the burden of proof shifts. Independent verification is not optional. It is mandatory.
Second, the correlation decline is not evidence of fundamental improvement. A mining company that pivots to AI hosting may achieve more stable revenue, but it also assumes new risks: customer concentration, contract quality, capital expenditure overruns, and competition from dedicated data center operators. The $851 million in combined losses at MARA and CleanSpark is a warning. AI transition is expensive, and the payoff is not guaranteed.
Third, the 90-day correlation window is a trailing indicator. It reflects what has happened, not what will happen. If Bitcoin enters a sustained bull market, mining stocks may regain correlation as mining revenue becomes more profitable relative to AI hosting. Conversely, if the AI narrative cools, miners could lose both the AI premium and the BTC linkage simultaneously, creating a double-negative scenario.
There is also a deeper structural question. If mining companies continue to shift toward AI infrastructure, who will secure the Bitcoin network? The hash rate that secures Bitcoin is increasingly concentrated in fewer hands, and if publicly traded miners abandon pure mining for AI compute, the remaining hash power will migrate to private operators and overseas facilities. This is not a near-term threat to network security, but it is a long-term centralization risk that the market is not pricing.
The Takeaway: Asset Reclassification and the Investor's Dilemma
Execution is final; intention is merely metadata. The intention behind buying mining stocks was to gain Bitcoin exposure. The execution has produced something entirely different: a hybrid asset with partial crypto beta and partial AI infrastructure beta.
For investors whose goal is Bitcoin exposure, the data is unambiguous. MicroStrategy provides a 78% correlation, more than double the best-performing mining stock. Bitcoin spot ETFs provide near-perfect correlation. Mining stocks provide neither. The strategy of using miners as a crypto proxy is no longer efficient, and continuing to do so is a cognitive error that will produce persistent underperformance.
For investors who believe in the AI infrastructure thesis, the mining sector offers an interesting entry point. These companies possess power assets, physical facilities, and operational expertise that are genuinely scarce. But the investment thesis must be framed correctly. This is not a Bitcoin trade. It is a bet on AI compute demand, electricity markets, and the execution capability of management teams that have historically been volatile.
The market is in the process of reclassifying mining stocks from crypto assets to AI infrastructure assets. This reclassification is rational, data-supported, and likely to continue. But reclassification does not eliminate risk. It transforms it. The investor who understands this transformation can position accordingly. The investor who does not will discover, eventually, that the asset they thought they owned was never the asset they held.
Inheritance is a feature until it becomes a trap. The mining sector inherited the crypto equity mantle, and now that inheritance is being repurposed. The question is not whether the transition will continue. The data says it will. The question is whether investors will update their mental models fast enough to avoid being trapped in a position that no longer represents what they intended to own.
The next two earnings cycles will provide the definitive evidence. If AI revenue continues to grow as a percentage of mining company income, the correlation to Bitcoin will continue to decline, and the reclassification will be complete. If AI revenue disappoints, the miners will face a valuation reckoning. Either outcome demands that investors stop treating mining stocks as Bitcoin proxies and start evaluating them for what they are: hybrid infrastructure companies operating at the intersection of two volatile industries.
I have audited enough systems to know that when the underlying structure changes, the surface metrics eventually follow. The correlation data is the surface metric. The business structure migration is the underlying change. The market has already begun to price this shift. The question is whether you have adjusted your portfolio accordingly.