← Back to Webthreepedia
WEBTHREEPEDIA RESEARCH

[DEEP DIVE] Bitcoin Miners Dump BTC, Sign $70B in AI Contracts

Zephyra|April 11, 2026|BPF
EXECUTIVE SUMMARY

Bitcoin's mining sector is undergoing its most severe structural realignment since China's 2021 ban. Network hashrate touched 1 zettahash per second (ZH/s) on March 28, 2026, only to fall back to approximately 870 EH/s within two weeks as operators redirected infrastructure toward artificial inte...

"We're not pivoting away from Bitcoin — we're pivoting toward where the economics work." — Asher Genoot, CEO, Hut 8

Executive Summary

Bitcoin's mining sector is undergoing its most severe structural realignment since China's 2021 ban. Network hashrate touched 1 zettahash per second (ZH/s) on March 28, 2026, only to fall back to approximately 870 EH/s within two weeks as operators redirected infrastructure toward artificial intelligence workloads. According to CoinShares' Q1 2026 mining report, hash price — the dollar revenue per petahash per second per day — collapsed to $29/PH/s/day, a five-year low, leaving 15–20% of legacy mining rigs globally cash-flow negative.

The weighted average cash cost to produce one bitcoin among publicly listed miners reached $79,995 in Q4 2025, per CoinShares data. With bitcoin trading near $72,800 as of April 11, 2026, the gap has narrowed, but all-in sustaining costs — including hardware depreciation, cooling, and overhead — remain near $88,000 per BTC for most operators. Publicly listed miners have collectively sold more than 15,000 BTC from treasury reserves since October 2025 to fund capital expenditures, primarily AI data center conversions. Over $70 billion in cumulative AI and high-performance computing (HPC) contracts have been announced across the sector.

Transaction fee revenue has collapsed from approximately 7% of miner income during the 2024 Ordinals boom to roughly 1% in Q1 2026, approximately $300,000 per day. Miners now depend almost entirely on the 3.125 BTC block subsidy, which halves again in April 2028. The sector's transformation raises a question that has moved from theoretical to urgent: whether a mining industry increasingly oriented toward AI can maintain the security budget Bitcoin requires.

Table of Contents

  1. The Hashrate Paradox: 1 ZH/s and Falling
  2. Production Economics: Mining at a Loss
  3. The AI Pivot: $70B in Contracts
  4. Fee Revenue Collapse
  5. Network Security Implications
  6. Energy and Infrastructure
  7. Key Takeaways
  8. Conclusion
  9. Sources & References

The Hashrate Paradox: 1 ZH/s and Falling

Bitcoin's network hashrate crossed the 1 zettahash milestone on March 28, 2026, reaching 1,022 EH/s according to CoinWarz data. The achievement was short-lived. By April 4, hashrate had declined to 961 EH/s, and CoinWarz reports it at 871 EH/s as of April 11 — a 15% drop from peak in under two weeks.

The decline followed three consecutive negative difficulty adjustments, a pattern not observed since July 2022. On March 21, mining difficulty fell 7.76%, the second-largest single-epoch drop of 2026, according to CryptoTimes. CoinDesk reported in February that the hashrate had at one point fallen to approximately 850 EH/s — a 27% decline from the October 2025 high — before partially recovering.

This marks the first quarter-over-quarter hashrate decline in six years, per CoinDesk data published March 30, 2026. The cause is not technical failure. Miners are rationally reallocating capital. One megawatt leased to an AI hyperscaler under a long-term infrastructure contract delivers predictable, fixed-rate income at margins that mining cannot match at current hash prices.

Production Economics: Mining at a Loss

According to CoinShares' Q1 2026 report, the weighted average cash cost to produce one bitcoin among publicly listed miners rose to approximately $79,995 in Q4 2025. All-in sustaining costs — incorporating hardware amortization, cooling, facility overhead, and SG&A — sit near $88,000 per BTC based on Checkonchain's difficulty regression model as of mid-March 2026.

Bitcoin's spot price as of April 11, 2026: $72,848 (CoinGecko). At these levels, the average publicly listed miner is operating at an approximate loss of $15,000–$19,000 per bitcoin mined. CoinDesk reported on March 22 that miners were "losing $19,000 on every BTC produced."

Profitability is now a function of electricity costs:

| Electricity Rate | Estimated All-In Cost per BTC | Margin at $72,800 | |---|---|---| | $0.03–0.05/kWh (hydro) | $38,000–$55,000 | +$17,800 to +$34,800 | | $0.06/kWh (industrial) | ~$65,000 | +$7,800 | | $0.10/kWh (commercial) | ~$92,000 | -$19,200 | | $0.13/kWh (US residential) | ~$121,000 | -$48,200 |

Only operators with sub-$0.06/kWh power contracts — typically in Paraguay, Iceland, Oman, or select Texas grid zones — retain positive margins. CoinShares estimates that any ASIC less efficient than the Antminer S19 XP running above $0.06/kWh is cash-flow negative.

Treasury liquidation has accelerated. Public miners collectively reduced BTC holdings by over 15,000 BTC from peak levels. Key dispositions reported by CoinDesk:

  • Core Scientific: Sold ~1,900 BTC (~$175M) in January 2026; plans to liquidate substantially all remaining holdings by end of Q1
  • Bitdeer: Reduced treasury to zero in February 2026
  • Riot Platforms: Sold 1,818 BTC (~$162M) in December 2025

The AI Pivot: $70B in Contracts

The economic logic is straightforward. A 1 MW data center rack generating AI inference revenue returns 3–5x the margin of the same rack mining bitcoin at current hash prices. The sector has responded accordingly.

Announced AI/HPC contracts across public miners exceed $70 billion in cumulative value, according to insights4vc analysis. Major deals include:

  • Core Scientific–CoreWeave: $10.2 billion over 12 years. Core Scientific is converting 1.2 GW of capacity toward AI data center operations.
  • TeraWulf: $12.8 billion in contracted HPC revenue, partnered with Fluidstack for hundreds of megawatts of compute in New York.
  • Hut 8: $7 billion, 15-year lease at its River Bend campus, backed by Google.
  • Riot Platforms: Publicly described as "transforming from Bitcoin miner to AI infrastructure giant" by financial media.

CoinShares projects that mining revenue could fall from 85% of total sector revenue to under 20% by late 2026 as miners complete their AI buildouts. Some operators could derive up to 70% of revenue from AI by year-end.

JPMorgan's January 2026 mining sector reset reflected the bifurcation. The bank upgraded Cipher Mining and CleanSpark — both maintaining pure-mining strategies with low-cost power — while trimming targets for Marathon Digital (MARA) and Riot, whose capital expenditure plans are weighted toward AI infrastructure.

CleanSpark reported 658 BTC mined in March 2026, with operational hashrate reaching 50.0 EH/s and treasury holdings of 13,561 BTC. Marathon mined 2,358 BTC in its most recent quarter with revenue of $238 million, up 64% year-over-year. Both companies represent the shrinking cohort still prioritizing mining economics.

Fee Revenue Collapse

Transaction fees have declined approximately 80% since the April 2024 halving, per The Block. Fee revenue now contributes roughly $300,000 per day — less than 1% of total miner income, down from approximately 7% during the 2024 Ordinals and Runes activity spike.

The Ordinals inscription market, BRC-20 token minting, and Runes protocol activity briefly generated sustained fee pressure in H1 2024. That demand has largely dissipated. The Block reported in March 2026 that bitcoin miner fees fell to a 12-month low, "underscoring long-term reliance on block subsidies."

For context: Bitcoin miners earned approximately $17.2 billion in total revenue in the 12 months following the April 2024 halving, per The Block. Of that, block subsidies at 3.125 BTC per block (approximately $227,000 per block at current prices) constitute over 99% of miner income. Transaction fees are effectively negligible.

This dynamic presents a long-term structural problem. The next halving in April 2028 will cut the subsidy to 1.5625 BTC per block. At current fee levels, the post-2028 security budget will be roughly half of today's — unless either bitcoin's price doubles or fee revenue increases by an order of magnitude.

Network Security Implications

The hashrate decline, while modest in absolute terms, has prompted renewed analysis of Bitcoin's security model. CryptoSlate published in March 2026 that "mining revenue hits historic low as infrastructure is sold to AI giants, permanently altering the network's security architecture." Techi.com reported that the 7.76% difficulty drop, combined with the AI capital reallocation, "exposes Bitcoin to potential 51% attack risks if hashrate doesn't rebound swiftly."

The concern is structural, not immediate. At 870+ EH/s, Bitcoin remains prohibitively expensive to attack through raw hashrate acquisition. However, the trend direction matters. If AI economics continue to outperform mining economics, rational capital allocation will continue to flow away from Bitcoin security.

The security budget — defined as total annual miner revenue — currently runs at approximately $17 billion annualized. This figure is sustained almost entirely by the block subsidy and is directly tied to BTC price. A sustained price decline toward $50,000 would compress the security budget to approximately $12 billion, potentially triggering further miner exits and a negative feedback loop.

Energy and Infrastructure

Bitcoin mining consumes an estimated 128–175 TWh annually as of early 2026, according to Cambridge's CBECI index and Digiconomist. The range exceeds the annual electricity consumption of Sweden or Norway. As of March 2026, mining one bitcoin in the U.S. requires over 750 MWh at an electricity cost of approximately $106,135 per coin at average residential rates, per BestBrokers data.

The AI pivot does not reduce total energy consumption by these facilities. It redirects it. Core Scientific's 1.2 GW capacity, Hut 8's River Bend campus, and TeraWulf's New York operations will continue drawing comparable power — but for AI inference and training workloads rather than SHA-256 hashing. From an energy-grid perspective, the transition is neutral. From Bitcoin's perspective, it represents a permanent capacity loss.

Key Takeaways

  • Hashrate touched 1 ZH/s on March 28 and fell 15% within two weeks. The first quarterly hashrate decline in six years signals structural, not cyclical, change.
  • Average all-in cost: ~$88,000/BTC. Spot price: ~$72,800. Most publicly listed miners are operating at a loss of $15,000–$19,000 per BTC produced.
  • $70B+ in AI/HPC contracts signed by former mining companies. Core Scientific, TeraWulf, Hut 8, and Riot are converting gigawatts of capacity.
  • Transaction fees at ~1% of revenue. Down from 7% during the Ordinals boom. Miners are almost entirely dependent on block subsidies.
  • 15,000+ BTC sold from public miner treasuries since October 2025 to fund AI infrastructure transitions.
  • CoinShares projects mining could fall to <20% of sector revenue by late 2026, making "bitcoin mining companies" a misnomer for most listed operators.

Conclusion

Bitcoin's mining sector is undergoing a repricing of its own economic model. The post-halving subsidy cut, the collapse of fee revenue from Ordinals activity, and the emergence of AI as a higher-margin use for the same infrastructure have created conditions where mining bitcoin is, for most operators, a money-losing proposition at current prices.

The sector's response — selling BTC reserves, signing multi-billion-dollar AI hosting contracts, and converting hashrate capacity to GPU racks — is rational. It is also irreversible. Data center infrastructure repurposed for AI workloads under 12- to 15-year contracts will not return to mining absent a fundamental change in bitcoin economics.

For the Bitcoin network itself, the implications are measured but non-trivial. The security budget is contracting in real terms. Fee revenue provides no meaningful offset. The 2028 halving will cut the subsidy again. The question is not whether Bitcoin can survive an industry where its miners are becoming AI companies. It is whether the remaining subsidy-funded security budget, roughly $17 billion annually and falling, is sufficient for a network that holds over $1.4 trillion in value.

The data suggests the answer depends almost entirely on one variable: bitcoin's price. And that, as the mining industry has been forcefully reminded, is outside any miner's control.

Sources & References

  1. CoinShares Bitcoin Mining Report Q1 2026 — Hash price, production costs, miner profitability data
  2. CoinDesk: Bitcoin Hashrate Posts First Quarterly Drop in 6 Years (March 30, 2026) — Hashrate decline analysis
  3. CoinDesk: Bitcoin Miners Losing $19,000 per BTC (March 22, 2026) — Per-unit loss calculations
  4. CoinDesk: Miners Becoming AI Companies (March 27, 2026) — AI pivot, treasury liquidation data
  5. The Block: Bitcoin Miner Fees at 12-Month Low (March 2026) — Fee revenue collapse
  6. CryptoTimes: Mining Difficulty Drops 7.76% (March 21, 2026) — Difficulty adjustment data
  7. CoinDesk: Miner Capitulation, 60-Day Hashrate Drop (January 19, 2026) — Capitulation metrics
  8. Carbon Credits: Hut 8 $7B Google-Backed AI Deal — Hut 8 contract details
  9. insights4vc: Bitcoin Mining's AI Pivot 2026 — $70B+ contract aggregate, revenue projections
  10. CryptoSlate: Mining Revenue Historic Low (March 2026) — Network security analysis
  11. The Block: 2026 Bitcoin Mining Outlook — Sector-wide outlook and fee revenue data
  12. CoinWarz: Bitcoin Hashrate Chart — Real-time hashrate data