Bitcoin Miners' AI Revenue Set to Equal Mining by Year-End
CoinShares research puts AI/HPC revenue on pace to match Bitcoin mining revenue across the listed miner cohort by year-end 2026. IREN crossed parity first. The math on why miners won't easily switch back.
CoinShares data shows AI/HPC revenue closing fast on mining revenue across the listed miner cohort, with one major miner already past parity.
Key takeaways
- AI/HPC revenue could represent up to 70% of listed Bitcoin miners' total revenue by end of 2026, up from roughly 30% at the start of the year, per CoinShares' Q1 and Q2 2026 mining reports.
- IREN became the first major listed miner where AI cloud revenue ($70.5M) overtook mining revenue ($66.7M) in a single quarter, making Q2 2026 parity a fact, not a projection.
- More than $100 billion in disclosed AI/HPC backlog currently supports only ~$1.1 billion in annualized billing revenue, with roughly 550 MW actually live against more than 4 GW under contract. The ramp, not the contract, is the story.
Bitcoin miners are on track to generate as much revenue from AI and high-performance computing as from mining by year-end 2026, according to CoinShares research published September 15. One miner already got there. The post-halving margin squeeze made the pivot economically mandatory; the energy infrastructure miners built made it possible.
The Numbers Behind the Pivot
The economics are stark. AI/HPC workloads earn approximately $1.5 million per megawatt per year. Bitcoin mining earns roughly $500,000 per megawatt. That 3x spread, cited consistently across CoinShares' Q1 and Q2 2026 mining reports, is the engine driving reallocation.
CoinShares ETF portfolio manager Bill Cannon, speaking to VettaFi/ETFtrends, put the production cost context plainly: at the end of Q2, producing one bitcoin cost about $75,000 on average, citing CoinShares research. With hash price near post-halving lows, the margin on mining alone is thin. AI co-location does not require the same bet on BTC price.
IREN delivered the clearest proof point. In Q2 2026, its AI cloud revenue hit $70.5 million against $66.7 million in mining revenue, the first time a major listed miner's AI business outgrossed its Bitcoin business in a single quarter. That data point, sourced directly from the CoinShares Q2 2026 report, makes the "year-end parity" call look conservative for the broader cohort.
Miners with HPC contracts now trade at 12.9x EV/NTM sales. Pure-play miners trade at 3.7x. The market is not treating these as the same business.
The Gap Between Backlog and Billing
The headline number demands scrutiny. More than $100 billion in disclosed AI/HPC contracts have been announced across the public mining sector. Annualized billing revenue is approximately $1.1 billion. Roughly 550 MW is actually live against more than 4 GW under contract.
That gap between contracted and billing is the real risk. CoinShares equity index manager Alex Schmidt told VettaFi/ETFtrends that roughly 70 contracts ramping later this year should add about $400 million per quarter in AI revenue. That figure is Schmidt's projection from the VettaFi webcast and has not been reconciled with the CoinShares Q2 2026 report, which disclosed approximately 20 co-location contracts. If those ramps execute, the year-end math works. If anchor tenants delay, defer, or face financing stress, the $100 billion backlog stays on paper.
The conversion economics also lock miners in. Outfitting a site for AI tenants costs $10 to $15 million per megawatt according to Schmidt, with the CoinShares Q1 2026 report citing a broader range of $8 to $15 million per megawatt, versus $1 to $1.5 million for mining infrastructure. A fully leased AI data center recently transacted at roughly $27 million per megawatt, compared to below $3 million per megawatt for energized-but-unleased miner capacity.
Cannon said Bitcoin would need to be "much higher than it is now" to justify switching back from AI at those capital commitments. Long-duration leases cement the irreversibility further.
What This Means for the Bitcoin Network
Every megawatt converted to AI co-location is a megawatt not securing the Bitcoin network. That is not a hypothetical concern. As the largest listed operators redirect their best-positioned, most energy-stable sites toward AI tenants, mining competition gets pushed toward more intermittent and cheaper power sources.
The AI capex committed at $10 to $15 million per megawatt does not unwind quickly. If BTC price spikes sharply, hashrate won't surge back on a short timeline. The irreversibility of AI capex makes this a long-run network security budget question that no earnings call is pricing.
The energy monetization thesis that Bitcoin miners pioneered is now being applied to AI workloads. Miners built infrastructure in places where energy was stranded, curtailed, or underpriced. That same positioning is exactly what AI compute buyers need. CoinShares investment strategist Matthew Kimmell called it "the great monetization of electrons." The power scarcity that constrains every other data center operator is the moat miners didn't know they were building.
The falsifiable version of the bull thesis: if Bitcoin sustains above a price level that collapses the $1.5M versus $0.5M per-MW spread (CoinShares suggests "much higher than it is now"), miners recommit to hashing and the AI pivot stalls. Separately, if major AI anchor tenants face financing distress, contract cancellations could crater the projected revenue ramp before it materializes. Both risks are real. The backlog-to-billing gap is the canary.
What to Watch
The Q3 2026 earnings cycle across the listed miner cohort will be the first real test of whether the Schmidt projection ($400 million per quarter in incremental AI revenue from ramping contracts) is tracking. IREN has already crossed parity. The question is whether Core Scientific, TeraWulf, and the broader cohort follow on a similar trajectory, or whether the $100 billion in backlog sits another quarter without converting to cash. Watch MW billing, not contract announcements.
Sources
Frequently Asked Questions
Will Bitcoin miners go back to mining if BTC price surges?
Most converted sites won't flip back easily. AI infrastructure costs $10 to $15 million per megawatt to build out, versus $1 to $1.5 million to revert to mining configuration. Long-duration co-location leases lock in the commitment further. CoinShares notes that Bitcoin would need to be significantly higher than current levels to make reconversion economically rational.
Does the AI pivot threaten Bitcoin network security?
Megawatts redirected to AI co-location are not hashing. As the industry's largest operators convert their most energy-stable sites, mining activity migrates toward more intermittent and lower-cost power sources. The irreversibility of AI capex makes hashrate recovery a long-run network security budget question that no earnings call is pricing.
How reliable is the $100 billion AI/HPC backlog figure?
The CoinShares Q2 2026 report shows more than $100 billion in disclosed AI/HPC contracts against only approximately $1.1 billion in annualized billing revenue, with 550 MW live against more than 4 GW contracted. Contracts are announcements. Revenue requires operational ramp, tenant financing, and on-time buildout. The backlog is the ceiling; current billing is the floor. The distance between them is the execution risk.
