Bitcoin miners spent $5.11 billion on capital assets in H1 2026 while generating $341.2 million in AI and high-performance computing (HPC) revenue — a 15-to-1 capex-to-revenue gap. Across 15 publicly listed miners and AI data center companies, total capital expenditure in the latest 2026 reportin...
"AI companies pay much more per electron compared to mining." — Fred Thiel, CEO, MARA Holdings
Bitcoin miners spent $5.11 billion on capital assets in H1 2026 while generating $341.2 million in AI and high-performance computing (HPC) revenue — a 15-to-1 capex-to-revenue gap. Across 15 publicly listed miners and AI data center companies, total capital expenditure in the latest 2026 reporting periods reached $30.7 billion, already 42.6% above the $21.53 billion deployed across all of 2025, according to BlocksBridge Consulting.
The numbers describe an industry in the middle of a structural transformation. Bitcoin mining difficulty has declined 19.9% from its November 2025 peak of 156 trillion to approximately 126 trillion, the third-deepest ASIC-era drawdown on record. Network hashrate dropped roughly 12% from above 1 zettahash per second to approximately 868 EH/s by late July 2026. The cause is not a Bitcoin crash. It is a reallocation of power capacity. Public miners controlling over 40% of global hashrate are actively redirecting megawatts from ASIC racks to GPU clusters. Miners have signed more than $160 billion in AI colocation contracts to date, with line of sight to over 14 gigawatts of total power capacity. CoinShares projects that AI and HPC could account for 70% of listed miners' revenue by end-2026, up from approximately 30% in Q1.
The transition follows a clear economic logic: HIVE Digital Technologies estimates that 10 MW of Nvidia H100 GPUs generates revenue comparable to 100 MW of Bitcoin mining. But the industry faces a $50 billion near-term funding gap, according to VanEck, with long-term capital needs approaching $221 billion. Only about 25% of contracted AI capacity has been energized. The question is no longer whether miners will pivot. It is whether they can build fast enough — and finance the build — before the window closes.
The math driving the transition is straightforward. Publicly listed miners lost approximately $19,000 per coin produced in recent quarters at current difficulty and BTC price levels, according to CoinDesk. Electricity — the single largest cost — commands a premium when directed toward AI workloads that mining economics cannot match.
At the H.C. Wainwright 28th Annual Global Investment Conference on September 14, 2026, executives from CleanSpark, MARA Holdings, Soluna Holdings, Bitdeer, Core Scientific, and others presented a unified message: the sector's competitive advantage is not hash power. It is energized infrastructure. Bitcoin miners own large-scale power contracts, physical data center space, and cooling systems — the same assets AI hyperscalers need but cannot build fast enough.
The incremental cost of repurposing existing mining facilities is substantially lower than greenfield data center construction. AI data centers require power-dense environments with direct-to-chip liquid cooling (now the industry standard following the Nvidia Blackwell GB200 platform rollout in early 2026), but the underlying power and land assets are interchangeable.
Publicly traded miners sold more than 32,000 BTC in Q1 2026 alone — exceeding their combined sales for all of 2025 — to fund the transition. Core Scientific sold $175 million worth of Bitcoin (approximately 1,992 coins) in March 2026 for operational retooling. Collective Bitcoin treasury reductions across public miners exceeded 15,000 BTC from peak levels.
The aggregate pipeline is enormous. More than $160 billion in AI colocation contracts have been announced across the sector. The major deals:
| Company | Counterparty | Contract Value | Term | Capacity | |---------|-------------|---------------|------|----------| | TeraWulf | Anthropic | $19B ($33B with extensions) | 20 years | 401 MW | | IREN | Microsoft | $9.7B | 5 years | Nvidia GB300 GPUs | | Hut 8 | Undisclosed | $9.8B | 15 years | Beacon Point campus | | TeraWulf | Fluidstack | $3.7B ($8.7B with extensions) | 10 years | 200+ MW | | CleanSpark | Undisclosed | $6.6B | 20 years | Triple-net lease | | Core Scientific | AMD/Neocloud | $24B+ potential | 15 years | 529-530 MW (options to 2.5 GW) | | Core Scientific | CoreWeave | Multiple contracts | Multi-year | ~590 MW | | Cipher Mining | AWS | $5.5B | Multi-year | Not disclosed | | IREN | AI developers | $2.8B | Multi-year | Multiple sites | | Riot Platforms | AMD | $636M | 10 years | Rockdale facility |
TeraWulf alone has contracted $12.8 billion in long-term, credit-enhanced customer agreements signed during 2025, with the Anthropic lease adding up to $33 billion in potential lifetime revenue through extensions.
Hut 8's total contracted AI portfolio reached $26.6 billion with expected annual net operating income above $1.75 billion. IREN's revised 2026 annualized run-rate revenue target now exceeds $4 billion, with approximately 85% of that target already under contract.
The sector-wide H1 2026 AI/HPC revenue of $341.2 million, while dwarfed by capex, showed acceleration. Q2 AI/HPC revenue hit $205.8 million, up 52% from Q1. Company-by-company:
Core Scientific reported Q2 2026 total revenue of $164.2 million, up from $78.6 million a year earlier. Colocation revenue reached $136.7 million, up from $10.6 million in Q2 2025. The company was billing 437 MW of capacity by mid-July, corresponding to annualized colocation revenue of approximately $635 million. Colocation gross margin: 59%. Total leased customer power: 1.1 GW representing over $24 billion in potential contracted revenue. Liquidity: $1.8 billion.
IREN reported Q2 FY26 revenue of $184.7 million. AI cloud revenue surged 137% quarter-over-quarter to $17.3 million, though total revenue fell 23% sequentially as Bitcoin mining revenue declined 28% on lower hashrate from the ASIC-to-GPU transition. The $9.7 billion Microsoft contract is expected to contribute approximately $1.94 billion in annualized run-rate revenue once fully commissioned.
TeraWulf generated Q2 2026 revenue of $44.8 million, including $31.9 million in HPC lease revenue — 71% of total revenue. This marked the first quarter where HPC revenue exceeded Bitcoin mining revenue. Visible Alpha expected HPC to account for 70% of TeraWulf's 2026 revenue.
Riot Platforms reported Q2 2026 total revenue of $174 million, up 14% year-over-year. Bitcoin mining revenue: $113.7 million. Data center segment revenue: $23 million, including $4.9 million in operating lease revenue. Direct mining cost: $49,912 per Bitcoin. Riot targets portfolio NOI of $1.6 billion to $2.1 billion on full development of its AI leasing pipeline.
VanEck's June 2026 analysis quantified the sector's central problem: a near-term funding gap of approximately $50 billion, with long-term capital expenditure needs approaching $221 billion. Only about 25% of the AI and HPC capacity contracted to customers has been energized and delivered.
IREN completed a $3 billion convertible notes offering — upsized multiple times due to strong investor demand — making it one of the largest capital raises by a public Bitcoin miner. The proceeds fund the transition from mining to AI infrastructure.
The 15-to-1 capex-to-revenue ratio across nine comparable miners in H1 2026 reflects the inherent lag between infrastructure buildout and revenue generation. Mining facilities require significant retrofitting: air-cooled ASIC environments must be converted to direct-to-chip liquid cooling systems for Nvidia Blackwell-class GPUs. Lead times for high-voltage transformers and switchgear remain the primary execution risk.
The power reallocation has measurable effects on Bitcoin's network security. Mining difficulty fell from its November 2025 all-time high of approximately 156 trillion to 126 trillion by late July 2026 — a 19.9% decline. Network hashrate dropped from above 1 ZH/s to approximately 868 EH/s.
The drawdown has since partially recovered. By September 6, 2026, difficulty recalculated to 127.45 trillion (up 1.31%) and hashrate readings fluctuated near 934 EH/s, with daily snapshots briefly touching 1,001 EH/s on September 4.
Mining economics improved during the recovery. Hashprice — revenue per unit of hashing power — gained 22.24% over the period, driven primarily by BTC price appreciation rather than hashrate growth. Bitcoin traded at $78,113 as of September 14, 2026.
The data suggest a new equilibrium is forming: a smaller but more efficient mining fleet coexists with a growing AI hosting business within the same corporate structures. Unprofitable smaller operators that cannot cover electricity costs are shutting down machines entirely, concentrating hashrate among larger, better-capitalized miners who can arbitrage between mining and AI hosting based on relative economics.
The sector's primary constraint, repeatedly cited by management teams at the H.C. Wainwright conference, is not demand. Global AI compute demand is running at approximately 15 GW per year. Hyperscalers are expected to spend over $1 trillion on AI infrastructure in the near term.
The constraint is physical buildout. Transformer and high-voltage switchgear lead times stretch months. The shift from air cooling to direct-to-chip liquid cooling (required for Nvidia Blackwell GB200 systems) adds engineering complexity. Political resistance in some jurisdictions and lender caution around the still-nascent AI hosting sector compound the timeline risk.
A January 2026 lawsuit against CoreWeave alleged the company overstated its ability to meet demand and concealed construction delays, underscoring the execution risks inherent in rapid infrastructure scaling.
The Nvidia Blackwell GB200 platform has rendered air cooling obsolete for frontier AI training workloads. As of February 2026, the industry standard shifted to direct-to-chip liquid cooling, adding both capital cost and technical complexity to facility conversions.
The market is differentiating sharply between companies based on execution stage. According to VanEck:
The valuation gap reflects the market's shifting focus from contract announcements to delivered megawatts. Companies that have energized capacity and are billing customers command a premium. Those with large pipelines but unbuilt infrastructure trade at a discount.
Core Scientific's stock reflects this dynamic: Q2 colocation revenue of $136.7 million represented a 12.9x year-over-year increase, yet the company reported a $1.2 billion net loss driven by non-cash accounting charges tied to its rapid expansion.
The Bitcoin mining industry's AI pivot is the largest capital reallocation in crypto-adjacent infrastructure history. The sector is transitioning from a single-product business — converting electricity to Bitcoin at thin or negative margins — to a dual-purpose infrastructure play that monetizes the same power assets at substantially higher per-megawatt returns.
The economic logic is sound: AI hosting generates 10x the revenue per megawatt compared to Bitcoin mining, according to HIVE Digital Technologies. The contracts are signed. The demand from hyperscalers is real and growing.
But the execution gap is equally real. A $50 billion funding shortfall, 75% of capacity still unbuilt, construction bottlenecks, and the technical demands of converting air-cooled mining facilities to liquid-cooled GPU environments create substantial delivery risk. The sector has layered a credit cycle (long-term lease obligations, convertible debt, construction financing) on top of a commodity cycle (Bitcoin price and difficulty).
The outcome will separate the industry into two tiers: companies that deliver energized AI capacity on schedule and collect multi-decade lease revenue, and those that burn through capital on unfinished buildouts. For Bitcoin's network security, the reallocation of mining power to AI workloads has already produced the third-deepest difficulty decline in ASIC history. The network is adjusting, but the direction is clear — the miners are going where the margins are.