Key Takeaways
- 14 publicly traded Bitcoin mining companies are developing ~30 GW of power capacity dedicated to AI infrastructure—nearly 3x the ~11 GW currently operational for mining
- Total HPC/AI contracts signed by miners in 2025 exceeded $65B, with mining revenue projected to drop below 20% of total miner revenue by late 2026
- Bitcoin mining is structurally unprofitable at current prices: Riot Platforms' all-in cost is $89K/BTC vs. $60K-$70K market price
- Ethereum L2 TVL is projected to exceed L1 DeFi TVL by Q3 2026 ($150B vs. $130B), shifting economic activity away from the security-bearing layer
- Bitcoin's security becomes exogenously dependent on AI compute market demand rather than endogenously determined by Bitcoin economics
The AI Subsidy: Bitcoin Security as an AI Company Sidecar
Fourteen publicly traded Bitcoin mining companies are developing approximately 30 gigawatts of power capacity dedicated to AI infrastructure—nearly three times the ~11 GW currently operational for mining. The financial logic is overwhelming: AI data center profit margins run 80-90%, versus volatile-to-negative margins for Bitcoin mining at current prices.
Riot Platforms' all-in mining cost reached $89,000 per BTC against a $60,000-$70,000 market price—structurally unprofitable. The deals already signed confirm this is not aspiration but execution. IREN struck a $9.7 billion Microsoft GPU cloud services deal. Core Scientific signed large AI hosting contracts. HIVE Digital reported Q4 2025 revenue of $93.1 million (+219% YoY), driven by AI/HPC. Starboard Value's activist letter to Riot Platforms quantified the opportunity: 1.7 GW of power capacity at Corsicana and Rockdale, Texas could generate $1.6 billion in annual AI colocation profit. Total HPC/AI contracts signed by the mining sector in 2025 exceeded $65 billion.
Industry consensus projects mining will constitute less than 20% of revenue for pivoted miners by late 2026. This fundamentally changes the economic model that secures Bitcoin: hash rate becomes a side product of AI infrastructure companies rather than the primary output of dedicated mining operations.
The Dual Security Budget Squeeze: Key Metrics
Bitcoin and Ethereum face simultaneous but structurally different pressures on their security economics
Source: MinerWeekly, The Block, CCN, CryptoPolitan
The Security Implication: Exogenous Dependence
The security implication is subtle but profound. In the traditional model, miners invest in hash rate because hash rate generates revenue (block rewards + transaction fees). Mining economics are endogenous to Bitcoin—they respond to Bitcoin price, difficulty adjustments, and halving cycles.
In the AI-subsidy model, miners maintain hash rate as a hedge or brand commitment while their primary revenue comes from AI compute. Mining economics become partially exogenous—influenced by AI compute demand, hyperscaler contract pricing, and GPU availability.
This creates a new systemic dependency: if AI compute demand contracts (recession, AI winter, hyperscaler capex reduction), miners who maintain hash rate as an AI-subsidized side product may cut mining operations first because they are the marginal revenue line. Bitcoin's security budget would then be exposed to shocks in a completely unrelated industry.
Bitcoin was designed with an assumption that security is economically self-sustaining: miners are incentivized to secure the network because security generates block rewards and transaction fees. The AI subsidy introduces a new assumption: security is subsidized by an exogenous industry with different economic fundamentals and risk drivers.
The L2 Value Migration: Ethereum's Settlement Layer Paradox
On the Ethereum side, L2 collective throughput has approached 100,000 TPS (a 6,500x improvement over mainnet's 15 TPS). More significantly, L2 TVL is projected to exceed Ethereum L1 DeFi TVL by Q3 2026—$150B versus $130B. This is the first time a scaling layer surpasses its base layer in locked value.
EIP-4844 (proto-danksharding, activated March 2024) cut L2 data availability costs by 50%+. ZK-rollups achieve 15,000+ TPS at $0.0001 per transaction. The cost reduction is extraordinary—but it comes at the expense of L1 fee revenue. Each transaction that migrates from L1 to L2 reduces the fees paid to Ethereum validators, while the L2 captures the execution fee revenue.
Ethereum's security is currently sustained by a combination of staking rewards (consensus layer) and transaction fees (execution layer). As L2s capture an increasing share of execution activity, the execution-layer fee component declines. This does not immediately threaten security—staking rewards provide a base yield—but it changes the long-term value proposition for validators and the economic sustainability of Ethereum's security budget.
The enterprise rollup phenomenon intensifies this dynamic. Robinhood's tokenized equities, Sony's Soneium, Kraken's INK, and Uniswap's UniChain are each deploying dedicated L2 infrastructure. These enterprise rollups capture fees within their own ecosystems while paying minimal data availability costs to L1. The value accrues to the L2 operator, not the L1 security provider.
The Dual Squeeze: Simultaneous Pressure on Both Networks
Bitcoin and Ethereum face structurally different but temporally coincident security budget pressures:
Bitcoin: Hash rate security increasingly subsidized by AI compute revenue. Security budget becomes contingent on exogenous AI demand rather than endogenous Bitcoin economics. Risk: AI demand shock → mining cutbacks → hash rate decline → reduced security.
Ethereum: Transaction fee revenue migrating from L1 to L2. Security budget sustained by staking rewards but execution-layer fee contribution declining. Risk: L2 dominance → L1 fee atrophy → reduced economic incentive for validators → validator centralization among entities that stake for strategic reasons (exchanges, custodians) rather than economic returns.
The coincidence matters because Bitcoin and Ethereum collectively represent the security backbone of the broader crypto ecosystem. If both networks face security budget compression from different vectors simultaneously, the systemic risk is higher than either individual risk suggests.
The Cross-Network Interconnection: How Both Risks Amplify
L2 consolidation and Bitcoin mining pivot intersect at the institutional capital layer. Institutional RWA issuers (BlackRock BUIDL, Franklin Templeton FOBXX) deploy on L2 infrastructure, using stablecoins backed by Treasuries as settlement currency. Bitcoin serves as collateral and store-of-value in institutional portfolios. If either network's security degrades, the institutional confidence that supports the other also weakens.
Consider the chain: Bitcoin miner pivot reduces hash rate → network security perception declines → institutional confidence in Bitcoin as collateral weakens → institutional capital rotation toward tokenized Treasuries on L2 increases → L2 TVL grows further → L1 fee revenue declines further → validator economics compress.
The two security budget pressures can amplify each other through the institutional confidence channel. A perception shock to either network's security could trigger a cascade of capital rotation that simultaneously pressures both networks.
The Whale-Mining Linkage: Convergent Bets on Price Recovery
While Goldman sold BTC ETF shares, on-chain whales accumulated 70,000 BTC ($4.6 billion) in early February—the largest whale accumulation episode since 2022. This is not contradictory positioning—both whales and miners are betting on long-term Bitcoin price recovery to justify the near-term economics.
Miners are structurally unprofitable at current BTC prices while whales accumulate aggressively. If whales are right about long-term BTC price recovery (Bernstein $150K target), miners survive via AI subsidy until BTC price recovers. If BTC remains below mining profitability, AI-subsidized mining is the only mechanism sustaining network security.
The convergence of whale conviction and mining economics through the AI subsidy bridge creates an interesting dependency: network security is now tied to both (a) whale beliefs about long-term BTC price recovery and (b) AI compute industry health. Neither of these dependencies existed in Bitcoin's original design.
What Could Make This Analysis Wrong
Four counterarguments deserve serious consideration:
1. Bitcoin's difficulty adjustment mechanism is specifically designed to maintain security: as miners exit, difficulty drops, remaining miners become more profitable, and hash rate stabilizes at a new equilibrium. This self-correcting mechanism has worked through every previous mining stress event.
2. The AI pivot may actually strengthen Bitcoin security by providing stable cross-subsidy revenue that funds hash rate maintenance through BTC price downturns—smoothing the mining revenue volatility that historically caused hash rate drops. Starboard Value explicitly argued this case.
3. Ethereum's EIP-4844 blob fees represent a new revenue mechanism specifically designed to capture L2 data availability demand. As blob usage scales, L1 revenue may recover through this channel even as execution fees migrate.
4. The projected L2 TVL > L1 TVL crossover (Q3 2026) is based on continuation of current growth rates, which may slow as the initial L2 migration wave matures. Ethereum L1 DeFi may stabilize at a larger absolute size than current projections suggest.
What This Means
Bitcoin and Ethereum are not facing security crises in the immediate term. Hash rate remains robust, and Ethereum continues to onboard new capital despite L2 migration. But both networks are undergoing structural transformations to their security economics that weaken the endogenous incentives that have historically sustained network security.
Bitcoin's pivot to AI subsidy is rational from a miner perspective—it improves profitability in the near term. But it introduces a new risk: security becomes contingent on an exogenous industry (AI compute) rather than self-sustaining through Bitcoin economics. A severe downturn in AI demand could expose Bitcoin to mining pressure that would require multiple years of difficulty adjustments to resolve.
Ethereum's L2 migration is rational from a user perspective—it provides massive throughput and cost improvements. But it weakens the economic incentives for Ethereum L1 validators by capturing execution fees at the L2 level. Over a 5-10 year horizon, this could lead to validator consolidation and reduced economic participation, even as absolute network security remains robust.
For institutional investors, both trends signal that crypto's base-layer security architecture is gradually becoming more fragile and more dependent on external factors. For miners and validators, both trends signal that long-term profitability requires either (a) betting on exponential price appreciation, or (b) building adjacent infrastructure that captures more of the value chain. The era of security-as-a-pure-commodity is ending.