Timechain Wiki

Mining pool centralization and the AI infrastructure pivot

12 min read

Two mining-infrastructure controversies have crystallized in 2025-2026. Pool concentration: approximately six pools collectively control 90-95%+ of block production, with Foundry USA persistently near 25-30%, raising operational censorship-capability concerns. The AI infrastructure pivot: miners face intense competition from AI hosting for electricity and data-center capacity, and operators like Core Scientific, Hive, and Hut 8 are shifting hardware and energy contracts toward high-margin AI workloads. The dual dynamics make Bitcoin's hashrate trajectory uncertain — continued concentration compounds censorship concerns, while the AI pivot may either consolidate mining further or decentralize it as larger operators rotate out. Distinct from Mining centralization concerns (the structural-analytical critique), this note is the event-level engagement with specific 2025-2026 dynamics and their 2026-2030 implications. The dispute remains live.

Why this note matters

The contemporary mining-infrastructure landscape is in active flux. The note matters because:

  • It surfaces the specific 2025-2026 dynamics at concrete-event level (Foundry’s persistence; AI pivot specifics; energy-competition empirics)
  • It distinguishes the event-level controversy from the analytical structural concern (Mining centralization concerns)
  • It articulates the dual dynamics (concentration + AI pivot) and their potentially-opposing trajectories
  • It engages the uncertainty about Bitcoin’s future hashrate trajectory that the dual dynamics create
  • It surfaces the named operators and specific events that constitute the contemporary controversy

The defensible position: the contemporary mining landscape is in genuine flux; the trajectory is unpredictable; both bull and bear scenarios for Bitcoin’s hashrate and decentralization are plausible. Tracking the dynamics is necessary for understanding Bitcoin’s mining-infrastructure evolution.

What happened

A condensed event-level chronicle of the contemporary mining dynamics.

2021-2024 — Post-China-migration pool consolidation. Following the May-September 2021 China mining ban, hashrate redistributed substantially. Foundry USA (Digital Currency Group subsidiary) emerged as the post-migration dominant pool, with persistent ~25-30% market share. AntPool, F2Pool, ViaBTC, Binance Pool, MARA Pool, and others occupy substantial but smaller positions. The top six pools have consistently controlled 90-95%+ of hashrate across 2022-2026.

2022-2023 — Pool filtering experiments. Several pools experiment with transaction filtering — OFAC-sanctioned-address filtering by some US-regulated pools; specific Ordinals/Inscriptions filtering experiments (variable across pools). Community pushback leads most pools to maintain non-filtering or minimal-filtering policies, but the operational capability is now empirically demonstrated.

2023 — Generative AI compute demand surges. OpenAI’s ChatGPT, Anthropic’s Claude, Google’s various AI products, and adjacent AI infrastructure create explosive demand for GPU-based compute. AI data centers require substantial electricity and specialized cooling. Energy prices in some regions begin reflecting AI-driven demand pressure.

2024 — Core Scientific pivot. Core Scientific — a major Bitcoin mining company (NASDAQ: CORZ) — announces and executes substantial pivot to AI hosting. June 2024: 12-year contract with CoreWeave worth ~$3.5B for ~200 MW initially, expanding. The pivot is structurally significant — Core Scientific had been one of the largest Bitcoin miners; the AI pivot demonstrates the economic case for hardware-and-energy-infrastructure repurposing.

2024-2025 — Adjacent miner pivots. Hive Digital, Hut 8, IREN, various others announce AI-hosting expansions. Many maintain dual Bitcoin-and-AI operations; some shift substantially toward AI. The economics are clear: AI hosting margins (currently 0.04-0.08 per kWh marginal) by substantial multiples.

2025-2026 — Energy contract competition. In specific regions (Texas ERCOT; Pacific Northwest; some Canadian regions), AI data centers begin outbidding Bitcoin miners for energy contracts. Some mining operations are unable to renew contracts at historical rates; others are forced to relocate or scale back. The competition is geographic-specific; not all regions show the same dynamics.

2025-2026 — Foundry persistence. Foundry USA maintains ~25-30% market share despite various competitive pressures. Specific Foundry policies on transaction inclusion (including periodic Ordinals-related filtering experiments) continue to attract scrutiny. The Foundry-as-dominant-pool dynamic remains a substantial concern.

Ongoing as of 2026-05-15. Six mining pools control 90-95%+ of hashrate; Foundry persistently dominant; AI pivot accelerating; hashrate trajectory uncertain.

The contested matters

Layer 1: How concentrated is Bitcoin mining, really?

The “concentration is severe” position:

  • Six pools controlling 90-95%+ of hashrate is empirically established
  • Foundry’s persistent ~25-30% share is structurally concerning — a single operator with meaningful market dominance
  • Pool-level filtering experiments demonstrate operational censorship-capability
  • The trajectory has not improved over 2021-2026 despite various decentralization efforts

The “effective concentration is less than apparent” position (per Mining centralization concerns):

  • Pool-level concentration ≠ effective censorship capability
  • Stratum V2 adoption (with template-negotiation) allows individual miners to defect at the template level
  • Miners can switch pools easily; defection from a censoring pool is operationally trivial
  • The economic incentives against coordinated censorship are substantial

The “we don’t yet know” position:

  • Stratum V2 adoption is partial; effective decentralization depends on deployment
  • The 2024-2026 period has shown both concentration persistence and partial decentralization initiatives
  • The trajectory is uncertain; both improvement and worsening are plausible

Layer 2: Is the AI pivot good or bad for Bitcoin?

The “AI pivot is bullish for Bitcoin decentralization” position:

  • Largest miners shifting focus to AI may decentralize hashrate among remaining Bitcoin-focused operators
  • Smaller miners may gain market share as larger operators de-emphasize Bitcoin
  • Energy-contract competition may push Bitcoin mining toward truly stranded-energy operations (where AI cannot compete)
  • The AI pivot validates Bitcoin mining as an electricity-buyer-of-last-resort while AI takes on grid-utility-scale operations

The “AI pivot is bearish for Bitcoin decentralization” position:

  • AI pivot accelerates infrastructure-and-capital consolidation; only the largest operators survive
  • Mining operations that can’t compete on energy costs exit; survivors are heavily-capitalized
  • Smaller, decentralization-favorable operations are squeezed out
  • The institutional-investor pressure on public mining companies pushes toward AI-margin-driven decisions that don’t preserve Bitcoin focus

The “depends on specific dynamics” position:

  • Specific regional dynamics determine outcomes; ERCOT Texas has different dynamics than Pacific Northwest or Bhutan or Paraguay
  • Mining operations co-located with stranded energy (flare-gas; specific renewable) may persist while grid-supplied operations are pressured
  • The trajectory is heterogeneous across operations and geographies

Layer 3: What’s the hashrate trajectory?

The contested empirical question. Multiple scenarios:

Scenario A — Continued hashrate growth: Bitcoin price growth justifies continued mining investment; some operators maintain Bitcoin focus; new entrants emerge in cheap-energy regions; hashrate continues growing despite AI competition.

Scenario B — Hashrate stagnation: AI pivot absorbs marginal compute-and-energy that would have grown Bitcoin hashrate; existing operations persist but growth slows substantially.

Scenario C — Hashrate decline: Substantial AI pivot leads to net hashrate reduction; surviving operators benefit (lower difficulty; higher margins); reduced absolute mining decentralization.

Scenario D — Geographic-specific divergence: Some regions (low-energy-cost; stranded-energy) maintain or grow Bitcoin mining; other regions (grid-energy-competitive) see Bitcoin mining decline; net effect varies.

The empirical evidence as of 2026-05-15 is mixed; specific operators show different trajectories; geographic dynamics vary substantially.

Layer 4: What about the censorship-capability concern?

The “concentration enables effective censorship” position:

  • 90-95% hashrate concentration in six pools means coordinated filtering by ~3-4 pools could meaningfully affect network operations
  • Foundry’s specific dominance and US-regulatory positioning compound concerns
  • Pool filtering experiments demonstrate operational capability; coordination at scale is plausible
  • The structural setup is genuinely concerning at the empirical level

The “individual miner defection prevents effective censorship” position:

  • Pool concentration does not equal coordinated-pool action
  • Miners hashing for a censoring pool would defect to non-censoring pools
  • Stratum V2 with template negotiation allows defection at the template level without leaving the pool
  • The empirical record (no successful coordinated censorship across 14+ years) is informative

The “Stratum V2 adoption is the variable to track” position:

  • Effective decentralization depends on Stratum V2 deployment
  • Current adoption is partial; 2026-2028 trajectory will be informative
  • If adoption proceeds, pool concentration becomes less operationally consequential
  • If adoption stalls, the structural concern remains

Layer 5: The institutional-mining-company dynamic

Public-equity-financed mining companies (Marathon, Riot, Core Scientific, Hut 8, Hive, CleanSpark, IREN, others) face distinctive dynamics:

  • Shareholder fiduciary duty pushes toward margin-maximizing decisions (AI hosting where economic)
  • ESG investor pressure affects energy-mix and operational decisions
  • Regulatory pressure affects operational positioning
  • Industry-association coordination (Bitcoin Mining Council; various trade groups) coordinates lobbying and standards

The institutional-mining-company sector is more responsive to financial-pressure than the private-mining or sovereign-mining sectors. The AI pivot specifically reflects shareholder pressure on margin optimization.

Where the dispute stands (as of 2026-05-15)

  • Pool concentration: empirically established (90-95% in 6 pools; Foundry persistently dominant); trajectory uncertain
  • AI pivot: empirically substantial (Core Scientific most prominent; many others); trajectory accelerating
  • Energy-contract competition: geographically variable; substantial in some regions
  • Hashrate trajectory: uncertain; multiple scenarios plausible
  • Censorship-capability: structurally enabled by concentration; empirically not demonstrated at scale
  • Stratum V2 adoption: partial; principal variable to track
  • Likely 2026-2030 trajectory: continued mining-and-AI dynamics; specific outcomes depend on AI growth, energy economics, and Stratum V2 deployment

Counter-arguments and tensions (criticisms of how this note frames the controversy)

“The ‘AI pivot’ framing may exaggerate Bitcoin-specific impact”

The framing concern: AI compute demand affects all energy-intensive industries (data centers; industrial processes; some manufacturing). Treating Bitcoin’s experience as distinctive may overstate the Bitcoin-specific dynamic versus the broader energy-economy trend.

Response: Partial. The AI pivot affects all electricity-intensive industries; Bitcoin’s specific exposure is distinctive because mining operations have hardware-and-data-center infrastructure that’s directly repurposable to AI hosting. The framing reflects Bitcoin’s specific competitive position; readers should recognize the broader context.

”Pool concentration ≠ effective censorship may be overstated”

The framing concern: The “individual miner defection prevents effective censorship” argument assumes miners will actually defect. In practice, defection takes time; coordinated short-window censorship could be effective even if not sustainable. The “no successful coordinated censorship” empirical record is not the same as “no possibility of coordinated censorship.”

Response: Valid concern. The note treats both positions; readers should engage the trade-off. The structural enablement is real even if empirical execution has not occurred.

”The Foundry-as-villain framing may be unfair”

The framing concern: Foundry USA has explicit policies; its dominance reflects market efficiency and Digital Currency Group’s investment, not malicious intent. Treating Foundry as a structural concern may unfairly characterize its actual operational behavior.

Response: Real. The note describes Foundry’s market position and specific experiments; it doesn’t allege malicious intent. The structural-concentration concern is about position, not behavior. Readers should distinguish positional concerns from intent allegations.

”The hashrate-decline scenario may be unduly pessimistic”

The framing concern: Bitcoin’s price growth has historically supported continued hashrate growth even through challenging conditions. Treating “hashrate decline” as a plausible scenario may understate Bitcoin’s price-driven hashrate dynamics.

Response: Partial. The note presents multiple scenarios; the decline scenario is one possibility. The empirical evidence is mixed; some operators are declining while others grow. Readers should weight scenarios according to their own assessments.

”Stratum V2 framing may overstate its decentralization significance”

The framing concern: Stratum V2 with template negotiation provides theoretical decentralization gains, but actual deployment is partial and adoption may not reach levels that produce effective decentralization. Treating it as a principal variable may overstate its decentralization potential.

Response: Valid concern. The note presents Stratum V2 as a variable to track; whether it produces effective decentralization depends on adoption. The framing reflects the technical potential and current trajectory; readers should engage the realistic adoption questions.

Verdict: Remains genuinely contested as of 2026-05-15; trajectory uncertain in multiple directions

The contemporary mining-infrastructure landscape is in genuine flux. Pool concentration and the AI pivot create competing dynamics with multiple plausible outcomes. The hashrate trajectory is uncertain; the censorship-capability concern is structurally real but empirically unexpressed.

A serious assessment:

  • Pool concentration: empirically established; trajectory unclear
  • Foundry dominance: persistent; specific dynamics continue to attract scrutiny
  • AI pivot: substantial; trajectory accelerating; effects on Bitcoin mixed
  • Energy-contract competition: regionally variable; substantial in some markets
  • Hashrate trajectory: uncertain; multiple scenarios plausible
  • Censorship-capability: structurally enabled; empirically not demonstrated at coordinated scale
  • Stratum V2: principal variable to track
  • Trajectory through 2030: continued flux; specific outcomes depend on AI economy, energy markets, deployment trajectories

This is a controversy worth tracking actively. The 2026-2028 period will produce substantial data — AI pivot outcomes; Stratum V2 adoption; specific mining-operator trajectories; hashrate metrics.

Open questions for further development

  • What’s the realistic AI-pivot trajectory among public-mining companies through 2028? Which operations maintain Bitcoin focus; which fully pivot; which dual-operate?
  • The Stratum V2 adoption variable is principal; what’s the realistic deployment path, and what would accelerate or stall it?
  • The hashrate-trajectory scenarios are mixed; what specific indicators would weight one scenario over others?
  • The Foundry-specific dominance is persistent; what conditions would shift its market share materially?
  • The intersection with Long-term security budget is substantial — hashrate trajectory affects security; AI pivot may produce more efficient mining (better hash-per-watt) but also potentially less mining absolute capacity. How do these dynamics evolve through the next-decade halvings?

Canonical sources for this note

Mining pool concentration data:

  • Mempool.space — real-time mining pool distribution
  • Hashrate Index — pool-concentration tracking and analysis
  • Foundry USA, AntPool, F2Pool, MARA Pool, etc. — pool-level publications and disclosures

AI pivot specifics:

  • Core Scientific CoreWeave contract disclosures (June 2024)
  • Hive Digital, Hut 8, IREN, CleanSpark public filings on AI-hosting expansion
  • Biggest Challenges Facing Bitcoin Miners (Yahoo Finance) — broader landscape coverage
  • AI killed Bitcoin debate erupts (Yahoo Finance) — adjacent contemporary coverage

Cambridge CCAF and adjacent academic:

  • Cambridge Bitcoin Electricity Consumption Index
  • Cambridge Digital Assets Programme reports
  • Various academic mining-economics papers

Industry analysis:

  • Hashrate Index quarterly reports
  • BitMEX Research mining-economics analyses
  • Nick Carter — various essays on mining-economics dynamics
  • Daniel Batten — BEEST methodology and adjacent mining-impact research

Within-Bitcoin engagement:

Adjacent contemporary coverage:

  • Bitcoin Magazine ongoing mining coverage
  • The Block, CoinDesk, Decrypt mining-industry coverage
  • AI infrastructure war analysis (panewslab, various)

As of 2026-05-15: pool concentration persistent; AI pivot accelerating; energy-contract competition ongoing; hashrate trajectory uncertain.

Paired Criticism note (cross-section):

  • Mining centralization concerns — the analytical structural-concentration critique; this controversy note treats the event-level contemporary dynamics (Foundry persistence; AI pivot; energy-competition) at concrete operational scale

Within the Controversies section:

Criticisms-section adjacency:

Mining section (cross-listed):

Adjacent thinker pages:

The sub-MOC home:

Graph View

Backlinks