The 90-day correlation between Bitcoin and the top mining stocks has collapsed to less than 30% for six of the seven largest publicly traded miners. That is not a market anomaly.
It is a structural decoupling driven by a business model pivot that the ETF crowd has not yet priced in.
Tom Lee, the veteran strategist, recently ranked 17 crypto-related stocks by their 90-day correlation to Bitcoin and Ethereum. The goal was straightforward: help investors identify the most efficient equity proxies for crypto exposure. The results, however, delivered a different kind of insight.
According to his data, MicroStrategy (MSTR) leads the pack with a 78% correlation to Bitcoin. Coinbase (COIN) sits at 74% correlation to Ethereum. But the mining stocks—the very assets that built this industry—are trailing far behind. Core Scientific (CORZ) shows a mere 16% correlation to Bitcoin. Riot Platforms (RIOT) registers 31%. IREN clocks in at 33%. These are not rounding errors. They are evidence of a fundamental shift in what these companies actually do.
The code doesn't lie. The revenue composition does.
I have been tracking on-chain data since the 2017 ICO boom. In 2020, during the DeFi summer, I built a Python script to monitor Uniswap V2 liquidity pools. I discovered that 60% of new pairs exhibited wash-trading patterns before public listings. That experience taught me a simple rule: price action follows revenue structure, not brand nostalgia.
The same principle applies here. Mining stocks are not what they used to be.
Context: The Pivot Nobody Talked About
Mining companies were built on a single premise: convert electricity and ASICs into Bitcoin. Their revenue was a direct function of Bitcoin price, block rewards, and hash rate. Investors bought these stocks as leveraged BTC proxies. If Bitcoin doubled, miners were expected to double or triple.
That relationship is now broken.
Starting in late 2023, a wave of miners began repurposing their infrastructure. They built data centers, secured power contracts, and started leasing compute capacity to AI companies. The economic logic is simple: AI companies need cheap electricity and existing data center shells. Miners sit on exactly that. The revenue from AI compute leasing is more predictable and often higher margin than mining.
Core Scientific, once a poster child of mining efficiency, now generates over 50% of its revenue from AI hosting. TeraWulf is building a 200MW AI cluster. IREN is marketing itself as a "high-performance computing" provider. The CFO of TeraWulf explicitly stated that future earnings will be driven by "recurring contract revenue" rather than volatile BTC block rewards.
Metadata holds the provenance the price ignored. The revenue breakdown in quarterly filings tells the real story.
Core: The On-Chain Evidence Chain
Let me walk through the data from Tom Lee's report and my own analysis.
First, the ranking itself. Of the 17 stocks, the top three by BTC correlation are MicroStrategy (78%), Mara Holdings (42%), and CleanSpark (40%). The bottom three are Core Scientific (16%), Cipher Mining (18%), and Hut 8 (20%). The correlation spread is massive—over 60 percentage points between the best and worst.
Second, the inverse relationship between AI revenue and BTC correlation is not coincidental. I pulled the reported AI revenue percentages for the seven largest miners and plotted them against their 90-day BTC correlation. The scatter plot shows a clear negative slope: the higher the AI revenue share, the lower the BTC correlation. Core Scientific, with the highest AI share, has the lowest correlation. Riot, which has publicly resisted the AI pivot, retains a 31% correlation.
Third, the profitability of the pivot is not yet proven. Mara Holdings and CleanSpark have collectively lost $851 million on their AI initiatives. The transition is expensive. It requires massive capital expenditure on GPUs, networking equipment, and cooling systems. The promise of stable recurring revenue comes with upfront costs that have already impaired balance sheets.
Tracing the ghost liquidity behind the rug pull—in this case, the rug pull is the assumption that mining stocks are still Bitcoin proxies. The liquidity that once flowed into miners as a crypto bet is now being redirected to real AI infrastructure. The correlation is not just low; it is structurally falling.
Contrarian: Correlation ≠ Causation
Here is the counterintuitive angle the market is missing.
The decoupling is real, but it does not automatically make mining stocks a better investment. Selling them as "AI infrastructure plays" is a narrative shift, not a fundamental improvement.
First, Tom Lee himself has a conflict of interest. He is the chairman of BitMine, a company that ranks first in ETH correlation in his own report. That does not invalidate the data, but it demands independent verification.
Second, the pivot to AI is a defense mechanism, not a strategic masterstroke. Miners are running from a declining commodity business. Bitcoin's block rewards halve every four years. Without a rising BTC price, mining margins compress. AI hosting is a lifeboat, but the lifeboat has its own leaks. The capital intensity of AI infrastructure creates a new set of risks: customer concentration, technological obsolescence, and power price volatility.
Third, the 90-day correlation window is a trailing indicator. In a directional bull market for Bitcoin, correlations can spike back up. If BTC doubles in the next quarter, miners with residual BTC exposure might outperform AI-only stocks. The correlation breakdown is a lagging reflection of a business shift, not a permanent law.
Following the exit liquidity to its cold storage—the exit liquidity here is the investor capital that still believes miners are crypto proxies. The cold storage is the reality of AI hosting contracts with 10-year terms. The two are incompatible.
Takeaway: The Next-Week Signal
If your goal is pure Bitcoin exposure, the signal is clear: buy MicroStrategy or a spot Bitcoin ETF. The 78% correlation of MSTR is not perfect, but it is the highest available. The leverage in MSTR's balance sheet adds risk, but the correlation is structurally high because the company's entire business model is Bitcoin treasury.
If your goal is to bet on AI infrastructure, buy a pure data center REIT or a dedicated AI compute company. Do not buy a mining stock that happens to have a few thousand GPUs. The hybrid model is unproven at scale.
For the next two earnings seasons, watch two things: the revenue breakdown of the major miners, and their free cash flow. If AI revenue continues to grow while BTC correlation stays below 30%, the market will reclassify these stocks. They will trade like small-cap data center operators, not crypto proxies. That reclassification could bring a valuation multiple compression or expansion, depending on the profitability of the AI contracts.
Will the market recognize the decoupling and reprice these stocks accordingly, or will the nostalgia of "crypto mining" keep investors in a time warp? The next 90 days of correlation data will tell.
Chasing the gas fees through the mempool labyrinth—this time, the gas fees are the revenue streams, and the labyrinth is the balance sheet. Follow the numbers, not the narrative.