Skip to main content

Bitcoin's First Q1 Hashrate Drop in Six Years: How the AI Pivot Is Rewriting Mining

· 12 min read
Dora Noda
Dora Noda
Software Engineer

For the first time since 2020, Bitcoin's hashrate ended a first quarter lower than it began. The world's most powerful computer network shrank by roughly 4% in Q1 2026, breaking five straight years of double-digit growth. The cause is not a regulatory crackdown or a hardware crisis. It is a more fundamental shift: the people who once raced to deploy ASICs are now racing to deploy GPUs, and they are paying for the transition by selling the very Bitcoin they used to hoard.

This is not a cyclical wobble. It is the moment that Bitcoin mining stopped being a single-purpose industry. According to the CoinShares Q1 2026 Mining Report, the weighted average cash production cost for publicly listed miners has climbed to nearly $90,000 per BTC, while spot prices hover closer to $67,000. With margins this deep underwater, "HODL" became a luxury, and AI hosting became an exit ramp. Over $70 billion in AI and HPC contracts have already been announced across the listed-miner peer group, and analysts now project that some operators will derive up to 70% of their 2026 revenue from non-mining workloads.

The Numbers Behind the First Q1 Drop in Six Years​

The headline figure obscures how violent the underlying move was. Network hashrate fell roughly 10% from its October 2025 peak of about 1,045 EH/s, bottomed near 850 EH/s by early February, and then partially recovered to leave the quarter down ~4%. Three consecutive negative difficulty adjustments accompanied the decline, the first such streak since the post-LUNA shake-out of July 2022.

The economic stress test is even sharper at the hashprice level. Hashprice — the daily USD revenue earned per petahash of capacity — touched roughly $28/PH/s/day in late February before stabilizing in the $30–$35 range. CoinShares estimates that 15% to 20% of miners are now operating at a cash loss, with mid-generation Antminer S19 fleets needing sub-5¢/kWh electricity simply to break even. The breakeven electricity price for an S19 XP collapsed from about $0.12/kWh in December 2024 to roughly $0.077/kWh by December 2025 — a near-implosion of the legacy mining cost curve in a single year.

Three forces converge to explain the squeeze. First, the April 2024 halving cut block subsidies in half just as the network's energy-intensive race for marginal hashrate kept pushing aggregate consumption higher. Second, winter 2025–2026 brought elevated power costs and ERCOT curtailment events that compressed Texas margins. Third, renewed Chinese enforcement actions in Xinjiang accelerated capacity flight just as the West's most expensive fleets were already underwater. The result was a margin compression unmatched since the 2018 bear market — but with one decisive difference.

The 2018 Comparison That Misses the Point​

Historical analogs only go so far. The 2018 hashrate retracement (down ~30% from peak) was driven by ASIC obsolescence colliding with China's first major crackdown — miners had no alternative buyer for their power and physical footprint. The 2022 post-Terra plateau (down ~10% over six months) was a financial-stress event in which hashrate corrected as overleveraged operators capitulated.

Q1 2026 is different in kind. The marginal megawatt no longer competes only against other miners. It competes against AI training clusters willing to pay 3–5x more per megawatt-hour, with multi-year contracted revenue that mining's stochastic block rewards cannot match. For the first time in Bitcoin's history, miners face a credible non-mining alternative for their core asset — energy, transformer capacity, and substation interconnects. That changes how shutdown decisions are made, and it changes how new capital is allocated. A mining rig that goes dark in 2018 was a fixed cost waiting to be re-deployed when prices recovered. A megawatt that gets contracted to Anthropic in 2026 is gone from the mining network for fifteen years.

Where the AI Money Is Actually Going​

The dollar figures clarify just how decisive this pivot has become. Visible Alpha consensus projections, reported across S&P Global and CoinDesk coverage, now expect a stunning composition shift in listed-miner revenue by year-end 2026:

  • IREN: 71% of revenue from HPC, up from 3% in 2024
  • Core Scientific: 71% of revenue from HPC, up from 5%
  • TeraWulf: 70% of revenue from HPC, anchored by $12.8 billion in contracted AI revenue and a 73.6% YTD stock gain
  • Cipher Mining: 34% of revenue from HPC, up from negligible
  • HIVE: 15% of revenue from HPC
  • Riot Platforms: 13% of revenue from HPC

The deal-by-deal flow tells the same story. Hut 8 anchored a $7 billion, 15-year lease at its River Bend campus with Anthropic and Fluidstack, while building an 8.5 GW pipeline. Galaxy Digital signed a 15-year, 800 MW commitment with CoreWeave at its Helios data center in West Texas — a contract Galaxy expects to generate roughly $4.5 billion in revenue over its term, and which the company financed with a $1.4 billion expansion package covering 80% of construction. MARA Holdings struck a strategic agreement with Starwood and acquired a controlling stake in French operator Exaion, signaling that even one of the most ideologically HODL-aligned miners has accepted that diversified compute is the baseline survival strategy. CleanSpark's CEO Matt Schultz has framed the company's identity bluntly: "evolving into a comprehensive compute platform that is prepared to optimize value from both AI and bitcoin workloads."

The single anchor tenant on the demand side deserves attention. CoreWeave — a former crypto mining firm itself — completed a $1.5 billion IPO in March 2025 and signed a five-year, $12 billion infrastructure contract with OpenAI in the same month. That single counterparty effectively underwrote the financing case for converting Bitcoin mining capacity into AI hosting capacity across the public miner peer group. The pivot is not speculative; it is contracted, financed, and under construction.

The Geographic Reshuffle​

Capital flight has a geography. As US-listed miners convert footprint into AI workloads, traditional Bitcoin mining capacity is migrating toward jurisdictions where stranded energy and supportive policy still make $90K-cost production economically viable. Paraguay, Ethiopia, and Oman have all entered the global top 10 in 2026:

  • Paraguay holds about 4.3% of network hashrate, anchored by HIVE Digital's 300 MW operation, and benefits from abundant hydroelectric power.
  • Ethiopia climbed to roughly 2.5% of global hashrate (#8 worldwide) despite a mid-2025 permit freeze, with Bitdeer's 60 MW site in phased energization driven by SEALMINER deliveries.
  • Oman has emerged rapidly thanks to state-supported energy initiatives and a national interest in monetizing low-cost power.

Meanwhile, US capacity, which previously accounted for over 40% of global hashrate among publicly listed miners, is flattening as those same miners convert their megawatts to AI. The decentralization optimist's read is that this rebalances the network. The pessimist's read is that mining is being chased to jurisdictions with weaker rule-of-law frameworks because the rich-world economics no longer pencil. Both can be true simultaneously.

What This Means for Bitcoin's Security Budget​

A lower hashrate is not, by itself, an existential threat. Bitcoin has weathered hashrate retracements before, and the network's technical security properties remain unchanged regardless of who is mining where. The deeper question is whether the long-run security budget — block subsidy plus transaction fees — can survive a structural shift in miner identity.

The block subsidy declines mechanically every four years. Transaction fees were always supposed to fill the gap, but fee revenue has not grown proportionally to compensate. If the marginal miner's reservation price is now set by AI hosting margins rather than by competing miners, the cost of attacking the network — measured in opportunity cost — could become harder to model. Defenders of the long-term thesis point out that the very same companies converting capacity to AI still maintain meaningful Bitcoin mining operations, often using the AI cash flows to subsidize ASIC fleets through bear cycles. That is a more resilient business model than pure-play mining, even if it produces less aggregate hashrate.

The new hashrate trajectory, formerly forecast at 2.5+ ZH/s by Q1 2027, has been quietly revised down to roughly 1.8 ZH/s by end of 2026 and 2.0 ZH/s by Q1 2027. That re-rating cascades through the valuation models of every mining-treasury company that has been priced as "BTC beta plus operational leverage." Going forward, those same names will trade more like data center REITs with embedded BTC optionality — a structurally different security with different correlations and different multiples.

The Investor Re-Rating Already Underway​

The market is already pricing the transition. Mining equities have outperformed Bitcoin itself by roughly 70% YTD in 2026, with TeraWulf's 73.6% gain and Riot's stock jumping after Starboard pushed for a $1.6 billion AI data center expansion. Riot reported record annual revenue of $647 million for 2025, with its data center arm contributing $33 million in Q1 2026 alone. The "mining stock as a leveraged Bitcoin proxy" thesis that drove 2024 retail allocation is being replaced by a "mining stock as energy-arbitrage compute provider" thesis — and the multiples are following.

For an asset allocator, the implication is sharp. A long mining-equity position is increasingly a long AI-infrastructure position with a residual BTC exposure, not the other way around. Hedging strategies that paired long mining stocks against short BTC futures will need to be re-modeled. So will any portfolio thesis that assumed mining capacity would expand alongside Bitcoin's price recovery, because the next price rally may pull less hashrate back online than historical analogs would predict — those megawatts are already contracted to OpenAI and Anthropic for the next decade.

What Comes Next​

Three open questions will define the rest of 2026. First, will any publicly listed miner cross the 50% threshold of revenue from AI/HPC contracts? Galaxy, Applied Digital, and HIVE are closest, but TeraWulf's existing contracted backlog suggests it could be the first to cross structurally rather than just episodically. Second, will Bitcoin's price recovery to $80K+ pull capacity back to mining, or has the AI alternative already entrenched to the point where conversions are one-way? Early indicators suggest one-way: the fifteen-year lease structures dominating new contracts make reverse migration economically irrational. Third, can transaction fee growth begin to fill the security-budget gap that a permanently lower hashrate trajectory implies — and if not, what happens at the next halving in 2028 when subsidies again drop by half?

The structural answer is that Bitcoin mining is no longer a single-purpose industry. The companies that survive 2026 will be data center operators that mine Bitcoin when it is the highest-value workload, and host AI training when it is not. That is a more durable business model than what came before. Whether it is the same Bitcoin network — economically, ideologically, in terms of who controls the marginal hash — is the question that the next two years will answer.

BlockEden.xyz provides reliable Bitcoin RPC and indexing infrastructure for builders who need data continuity regardless of how the underlying miner mix evolves. Explore our API marketplace to build on infrastructure designed to outlast the cycle.

Sources​

Share on Twitter