Bitcoin miners are becoming AI data center operators. The conversion is no longer theoretical — it is showing up in quarterly filings, hashrate data, and stock prices. Network hashrate fell approximately 4% in Q1 2026, the first first-quarter decline since 2020, according to CoinDesk. The cause i...
"MARA is no longer simply a Bitcoin miner. We are already well down the path of building an energy-dominant digital infrastructure platform." — Fred Thiel, CEO, Marathon Digital Holdings
Bitcoin miners are becoming AI data center operators. The conversion is no longer theoretical — it is showing up in quarterly filings, hashrate data, and stock prices. Network hashrate fell approximately 4% in Q1 2026, the first first-quarter decline since 2020, according to CoinDesk. The cause is structural: publicly listed miners are redirecting power capacity, selling bitcoin reserves, and signing multi-billion-dollar leases with AI hyperscalers.
The economics are straightforward. According to CoinShares' Q1 2026 Mining Report, the weighted average cash cost to produce one bitcoin among listed miners has climbed to roughly $80,000–$90,000, while spot BTC trades near $67,000. That gap — a loss of up to $19,000 per coin mined — makes the pivot rational. AI cloud infrastructure generates margins of approximately 85%, according to industry estimates, compared to negative margins for a growing share of the mining fleet. Over $70 billion in AI and HPC contracts have been announced across the listed-miner peer group as of June 2026.
The market has noticed. A basket of publicly traded mining stocks is up 56% year-to-date, according to CryptoTimes, while bitcoin itself has fallen 17% over the same period. The divergence is the clearest signal yet that Wall Street now values these companies for their power assets and AI tenants, not their hash rate.
Bitcoin's network hashrate peaked near 1,045 EH/s in October 2025, according to CoinWarz. By early February 2026, it had bottomed near 850 EH/s — a decline of roughly 10%. A partial recovery followed, but Q1 still closed down approximately 4%, marking the first first-quarter drop in six years, per CoinDesk reporting on March 30, 2026.
The current hashrate sits at approximately 777 EH/s as of early June 2026, according to CoinWarz, well below the October peak. The next difficulty adjustment, estimated for June 13–14, 2026, is projected to drop difficulty from 138.96 T to approximately 124–126 T — a decline of roughly 9%, according to CoinWarz and BitRef estimators.
A 9% downward adjustment would be one of the largest single drops logged in 2026. It confirms that hashrate has meaningfully departed the network, and the departure is not temporary. Miners are not merely shutting off machines during a price dip — they are physically converting facilities.
The April 2024 halving cut block rewards from 6.25 BTC to 3.125 BTC. The math since then has been unforgiving.
CoinShares' Q1 2026 Bitcoin Mining Report provides the clearest picture of the margin crisis:
The energy efficiency frontier has improved dramatically — from 98 J/TH in 2018 to sub-15 J/TH for current-generation ASICs — but it has not kept pace with the halving's impact on unit revenue. The result: miners with large power positions are worth more as landlords than as hashers.
Core Scientific has executed the most aggressive conversion. In Q1 FY2026 (reported May 6, 2026), colocation revenue for AI data centers reached $77.5 million, surpassing crypto mining revenue of $30.1 million. AI is now the majority of the business.
The company's CoreWeave contract spans 590 MW across five sites (Denton TX, Dalton GA, Muskogee OK, Marble NC, Austin TX), with projected revenue of $10.2 billion over 12-year terms. An additional 300 MW at its Pecos, Texas campus — currently used for bitcoin mining — is being converted to AI infrastructure, with a total pipeline exceeding 3 GW.
Core Scientific sold $208 million of bitcoin in Q1 2026 to fund the transition, per CoinDesk.
Hut 8 signed a 15-year, 245 MW AI data center lease at its River Bend campus in Louisiana with Fluidstack and Anthropic (Google-backed), valued at $7.0 billion. The deal was announced December 2025, with the initial data hall scheduled for Q2 2027 commissioning.
In May 2026, Hut 8 announced a second major deal, doubling contracted capacity and bringing total contract value to $9.8 billion. CEO Asher Genoot stated the company has "signed about $17 billion of deals in the last six months" with an additional 8 GW development pipeline. Hut 8 is also selling $3.25 billion in bonds to finance the buildout.
Stock up approximately 67% YTD.
MARA sold over $1.5 billion in bitcoin as of May 2026 to fund its infrastructure pivot, per CoinDesk. The company has partnered with Starwood for a dual-purpose facility model capable of toggling between bitcoin mining and AI compute workloads based on demand and pricing.
CEO Fred Thiel has stated MARA prefers a real estate and leasing model over GPU-as-a-service, citing better economics and lower capital burden. The company's near-term priorities are signing tenant leases for existing portfolio sites and closing its Long Ridge transaction.
Stock up approximately 44% in 2026.
Riot reported Q1 2026 revenue of $167.2 million, including $33.2 million from its data center business. AMD exercised an option for an additional 25 MW, bringing total contracted capacity at Riot's facilities to 50 MW.
The company has positioned itself as a pure-play power and land holder, leasing capacity to tenants rather than operating GPU clusters directly. Stock up approximately 46% YTD.
TeraWulf leads the peer group in stock performance, up approximately 85% YTD according to CryptoTimes. The company has leaned heavily into its nuclear-adjacent power position at its Nautilus facility in Pennsylvania as a differentiator for AI workload customers seeking carbon-free compute.
IREN reported Q1 FY26 revenue of $240.3 million and projects $3.4 billion in annual recurring revenue by end of 2026, anchored by a five-year AI cloud deal with Microsoft. Stock up approximately 18.5%.
The aggregate value of AI and HPC contracts announced by publicly listed bitcoin miners exceeds $70 billion as of June 2026. The scale of this pipeline dwarfs anything in bitcoin mining history. For context, total bitcoin network block rewards plus fees currently generate approximately $14 billion annualized.
The contract structure is also fundamentally different from mining economics. Mining revenue is variable — tied to BTC price, difficulty, and fees. AI leases are typically fixed-rate, multi-year, triple-net structures with investment-grade counterparties (Google, Microsoft, AMD, CoreWeave). This transforms miners from commodity producers into infrastructure REITs.
CoinShares' Q1 2026 report projects that some operators will derive up to 70% of their 2026 revenue from non-mining workloads, up from approximately 30% at the start of the year.
The divergence between mining stock performance and bitcoin price performance in 2026 is historically unusual:
| Metric | YTD Performance | |--------|----------------| | Bitcoin (BTC) | -17% | | Mining stock basket | +56% | | TeraWulf (WULF) | +85% | | Hut 8 (HUT) | +67% | | Riot Platforms (RIOT) | +46% | | MARA Holdings (MARA) | +44% | | Core Scientific (CORZ) | +40% | | Applied Digital (APLD) | +37% |
Source: CryptoTimes, May 3, 2026
Historically, mining stocks have traded as leveraged proxies for BTC. The 2026 decorrelation suggests that equity investors now price these companies primarily on their AI infrastructure value, with bitcoin mining as a secondary — or even legacy — business line.
Canaccord nearly doubled its price target on Hut 8 to $130 in May 2026, citing the AI data center pivot. This is a fundamental rerating of what these companies are.
The miner exodus raises a legitimate question about Bitcoin network security. With hashrate down from 1,045 EH/s to approximately 777 EH/s — a 26% decline from peak — the cost of attacking the network has decreased proportionally.
However, several mitigating factors apply:
Absolute hashrate remains enormous. At 777 EH/s, Bitcoin's network is still orders of magnitude more secure than any other proof-of-work chain and far exceeds the hashrate levels of 2020–2023 when no serious attacks occurred.
Geographic distribution may improve. U.S.-listed miners currently control over 40% of global hash power, per CoinDesk. As these operators redirect capacity to AI, the remaining hashrate may become more geographically distributed, which strengthens decentralization.
Difficulty adjustment is the mechanism. The ~9% downward adjustment expected June 13–14 is the protocol working as designed. Lower difficulty makes mining marginally more profitable for remaining operators, stabilizing the network at a new equilibrium.
Private miners absorb share. The publicly listed miners generating headlines are not the entire network. Private operators, sovereign miners, and stranded-energy miners continue to operate, and their relative share grows as public miners exit.
The risk is not imminent. But if BTC price remains below production cost for an extended period and the AI pull continues, hashrate could test lower levels that invite more serious security discussions.
The projected June 13–14 difficulty adjustment — from 138.96 T to approximately 124–126 T — is a lagging confirmation of what the market already knows. Hash rate has left, and it is not coming back to bitcoin.
For miners that remain, lower difficulty is a tailwind: each unit of hashrate captures a larger share of block rewards. This tends to create a floor — when difficulty drops enough, marginal miners become profitable again and hashrate stabilizes.
But the current cycle is different because the hashrate is not merely being turned off. It is being permanently reallocated to higher-value use cases. The machines being powered down are ASICs with no alternative function. The power and physical infrastructure they occupied is being filled with GPUs serving AI workloads on 5–15 year leases.
This is a one-way door. Facilities converted to AI data centers will not revert to bitcoin mining.
The bitcoin mining industry is undergoing a structural transformation that goes beyond a typical cycle adjustment. This is not miners temporarily powering down during a price dip and restarting when profitability returns. This is permanent infrastructure conversion — ASICs out, GPUs in, on multi-year leases with hyperscaler counterparties.
The economic logic is unambiguous. Mining bitcoin at $80,000–$90,000 cost while selling at $67,000 is a losing proposition. Leasing the same power and physical footprint to AI customers at 85% margins on 5–15 year contracts is not. The market has priced this accordingly: mining stocks have diverged from bitcoin for the first time in the sector's history.
For Bitcoin as a network, the departure of industrial-scale miners is a test of the protocol's incentive design. The difficulty adjustment mechanism — due to drop approximately 9% this week — is performing its intended function, recalibrating to maintain block production with fewer participants. Whether the remaining hashrate is sufficient for long-term security depends on how far the exodus extends and whether new entrants replace departing operators.
The data from Q1 2026 earnings, the CoinShares report, and the $70 billion contract pipeline all point in the same direction: the companies that once defined themselves by their hash rate now define themselves by their megawatt capacity. Bitcoin mining, for many of the industry's largest operators, has become the legacy business.