Miners Stop Mining Bitcoin and Sell Electricity to AI

By Cathy

Mining one bitcoin costs $87,000. Selling it fetches only $67,000 on the market.

That’s a $20,000 net loss per coin mined—not from transaction fees or fluctuating electricity prices, but a hard, direct $20,000 loss on every single bitcoin produced. This is the reality in March 2026. Data from both Glassnode and MacroMicro point to the same conclusion: Bitcoin mining is currently an unprofitable business at prevailing prices.

Yet miners aren’t sitting idle waiting for collapse. They’ve made a move no one in the market anticipated: they’ve stopped mining—and started selling electricity to AI instead.

More precisely, it’s not that they’ve “stopped mining” entirely—but rather liquidated their entire bitcoin treasuries and poured all available capital into AI data centers, relegating mining to a secondary activity.

Since Bitcoin peaked at its all-time high of $126,000 in October 2025 and began its downward reversal, publicly listed mining companies have collectively sold over 15,000 bitcoins. This isn’t sporadic profit-taking—it’s a coordinated, strategic retreat.

Miners’ Mass Exodus: Where Did 15,000 BTC Go?

Core Scientific was the earliest and most decisive mover.

In January 2026, it sold approximately 1,900 bitcoins—raising $175 million. It plans to liquidate the remainder during Q1. Once a bankrupt and restructured miner, Core Scientific is now converting its Texas mining facilities one by one into high-density AI colocation sites, aiming to allocate its full 1.3 GW power capacity to AI.

MARA went even further. Known for its long-standing “never sell coins” policy, MARA quietly revised its treasury policy in its March 2026 10-K annual report—authorizing the sale of all 53,822 bitcoins held. Valued at nearly $4 billion at the time, this massive reserve instantly shifted from “strategic holdings” to “deployable capital.” Shortly thereafter, MARA signed a joint venture agreement with Starwood Capital to deliver 1 GW of AI data center capacity.

Most surprising was Cango. Originally a Chinese auto-financing platform, it entered Bitcoin mining only at the end of 2024—yet in February 2026, it sold 4,451 bitcoins, representing 60% of its reserves, raising $305 million to repay debt and fund its AI pivot. It also hired Jack Jin, former Zoom executive, as CTO of its AI business, planning to deploy containerized GPU compute nodes across its global mining sites. A car-loan company transformed into a miner—and then, within two years, into an AI inference service provider. Only the crypto industry moves this fast.

Bitdeer’s strategy resembled a carefully calculated chess move. In February, it fully liquidated its self-held bitcoin inventory. Founder Wu Jihan candidly explained: zero holdings don’t mean zero forever—liquidity is needed right now to seize opportunities in acquiring power and land. Unlike other miners, Bitdeer cleared its balance sheet while simultaneously accelerating growth: its January bitcoin output surged 430% year-on-year, and its proprietary hashrate reached 63.2 EH/s—surpassing MARA to become the world’s largest publicly listed miner by proprietary hashrate. Liquidating on-chain coins enabled rapid expansion of hashrate and infrastructure. It combined the decisiveness of “cutting off a limb to save the body” with the ambition of “loading up ammunition.”

The Same Electricity Is Worth 10x More to AI

Why are miners selling so uniformly? Because the math leaves no room for doubt.

Mining is unprofitable—but miners hold something the world is scrambling for: land with guaranteed power supply.

After the 2024 halving, Bitcoin mining margins collapsed from over 90% at their peak down to breakeven levels. Meanwhile, AI’s demand for electricity and data centers exploded. MarketsandMarkets forecasts the global AI inference market will grow from ~$10.6 billion in 2025 to nearly $25.5 billion by 2030.

Morgan Stanley ran the numbers: shifting 1 megawatt of power from mining to AI colocation delivers over a 10x valuation premium.

This isn’t hyperbole. AI colocation contracts typically run 10–15 years, with investment-grade clients like Microsoft and Meta—delivering stable, predictable cash flow. By contrast, mining revenue depends entirely on cryptocurrency price—and you know how volatile that is.

Wall Street has already voted with real money. Morgan Stanley extended Core Scientific a $500 million loan facility—with an option to increase it to $1 billion. This isn’t a loan to a “crypto company”; it’s a credit endorsement for a “digital infrastructure company.” TeraWulf and Cipher Mining earned Morgan Stanley’s “Overweight” rating thanks to their successful hybrid models, while MARA—once doggedly holding bitcoin—was downgraded due to excessive exposure to price risk.

The message from capital markets couldn’t be clearer: these companies’ value is no longer defined by how many bitcoins they hold—but by how much power they control.

On-Chain Metrics Suggest a Bottom May Be Near

Miners’ collective sell-off has triggered widespread market gloom. Yet on-chain data reveals a set of intriguing signals.

The Hash Ribbon indicator inverted in late November 2025 and remained inverted through February 2026—a full three months. This marks one of the longest miner capitulation periods on record. The last time a similar signal combination appeared was December 2022—when Bitcoin bottomed at $15,500. As of early March, the 30-day moving average is approaching the 60-day moving average from below—suggesting a recovery signal may soon trigger.

The MVRV Z-Score held between 0.43 and 0.49 in early March. This metric measures how far current market price deviates from “realized value.” Historically, Z-Score values between 0 and 1 have almost always coincided with strategic accumulation windows.

The Puell Multiple dropped to ~0.6—indicating miners’ daily revenues have been compressed to roughly 60% of their annual average. That’s approaching the 0.3 level seen at the 2022 bear market bottom. Miners’ profit margins are being squeezed to historic lows.

The most extreme signal came from sentiment. During February’s “Bitcoin Polar Vortex,” the Crypto Fear & Greed Index plunged to 5—the lowest since inception—and on February 5, single-day realized losses hit a record $3.2 billion.

Four independent indicators flashing red simultaneously hasn’t happened since Bitcoin was forming a bottom.

Are Miners’ Bitcoin Sales Actually Bullish?

This is the most counterintuitive part of the entire story.

Historically, miner sales have been viewed as bearish—these are Bitcoin’s “native sellers,” whose continuous selling creates persistent downward pressure. But the 2026 wave of sales is fundamentally different: after selling bitcoin, these firms are pivoting to earn USD revenue from AI.

Consider what that implies. Previously, Core Scientific had to sell hundreds of bitcoins monthly just to cover electricity bills and operating expenses. Now, backed by Microsoft’s long-term contract and Morgan Stanley’s credit line—even though it still plans to monetize most of its remaining bitcoin holdings (holding ~2,537 BTC by year-end, having sold the majority)—this is no longer passive “selling to survive.” It’s active treasury liquidation to concentrate capital on AI infrastructure. Once MARA and Starwood’s joint venture goes live, the USD cash flow generated by that 1 GW data center will fully cover all costs.

In other words, miners transitioning to AI are shifting from structural Bitcoin sellers to neutral—or even potential future buyers. The market’s largest cohort of “natural short sellers” is exiting permanently.

And Bitcoin mining itself hasn’t disappeared—it’s merely evolved. MARA’s hybrid model points the way forward: mine when electricity is cheap; switch to GPU compute during AI demand peaks. Bitcoin becomes a “flexible grid load” and “insurance mechanism,” while AI handles revenue generation and mining provides downside protection.

Summary

In 2025, Bitcoin’s network hashrate surpassed the 1 zettahash milestone. In the short term, some mines shifting to AI will indeed slow hashrate growth—for example, Cango has taken 31% of its hashrate offline for upgrades. But this represents healthy capacity rationalization: inefficient miners exit, leaving behind more efficient, focused participants—and network security improves, not declines.

This isn’t miner surrender—it’s the evolution of mining.

When mining becomes a side business and AI the core business, Bitcoin loses a cohort of forced sellers—but gains a healthier supply structure.

Miners have sold all their bitcoins—but the electricity remains.