A Bitcoin Miner Switches to AI for a $4.7 Billion Lease

Chamath Palihapitiya · Chamath Palihapitiya · August 09, 2026 at 14:12 · ⏱ 9 min read  | Read on Substack ↗
Summary
The article argues that the most lucrative layers of the AI stack are currently physical infrastructure (Land, Power, Shell) and software 'harnesses', while silicon has become too capital-intensive and competitive. This shift is evidenced by Bitcoin miners pivoting their powered sites to highly lucrative AI data center leases, new thermodynamic chip architectures challenging GPU efficiency, and Nvidia commoditizing self-driving software to drive its hardware sales.
  • The author and his partner have assembled nearly 6GW of power capacity, highlighting the premium on energized land with signed interconnections.
  • Bitdeer signed a 16-year, $4.7B lease to convert its Norwegian Bitcoin mining site into an AI data center for Volta, with Dell supplying the hardware.
  • CoinShares projects listed Bitcoin miners will derive up to 70% of their revenue from AI by the end of 2026, up from 30% in March.
  • Extropic unveiled Z1, a Thermodynamic Sampling Unit (TSU) with 269,568 pbits that claims to be up to 10,000x more energy-efficient than GPUs for specific sampling workloads.
  • Nvidia released its Alpamayo 2 Super autonomous driving model for free, which placed first among nearly 40 systems on LingoQA, ahead of Gemini 2.5 Pro and GPT-4o.
Read time 9 min
Length 9,154 chars
Category finance
Ideas
Chamath Palihapitiya CEO, Social Capital
Bitdeer signed a massive 16-year lease worth $4.7B (potentially $8B over 24 years) to convert its Tydal mining site into an AI data center, demonstrating the immense monetization potential of existing
Bitdeer signed a massive 16-year lease worth $4.7B (potentially $8B over 24 years) to convert its Tydal mining site into an AI data center, demonstrating the immense monetization potential of existing powered sites with grid connections. Risk: Execution risk in converting the facility from crypto mining to high-performance AI computing, and counterparty risk with the tenant.
Chamath Palihapitiya CEO, Social Capital
Nvidia is commoditizing the self-driving software layer by open-sourcing Alpamayo 2 Super, which forces automakers to buy NVIDIA clusters for fine-tuning and DRIVE AGX hardware for in-car inference.
Nvidia is commoditizing the self-driving software layer by open-sourcing Alpamayo 2 Super, which forces automakers to buy NVIDIA clusters for fine-tuning and DRIVE AGX hardware for in-car inference. Risk: Alternative chip architectures like Extropic's TSUs could eventually challenge GPU dominance in specific sampling and generative workloads.
Chamath Palihapitiya CEO, Social Capital
Nvidia's open-sourcing of a top-tier autonomous driving model gives competitors (Lucid, Uber, Lyft, Mercedes) a 'shortcut to developing self-driving, by skipping the step of funding a model from scrat
Nvidia's open-sourcing of a top-tier autonomous driving model gives competitors (Lucid, Uber, Lyft, Mercedes) a 'shortcut to developing self-driving, by skipping the step of funding a model from scratch,' eroding Tesla's proprietary software moat. Risk: Tesla's massive real-world driving data advantage and vertical integration may still outweigh the baseline capabilities of open-source models.
Chamath Palihapitiya CEO, Social Capital
Dell is explicitly named as the hardware supplier for the massive Bitdeer/Volta AI data center project in Norway, indicating strong enterprise AI server demand.
Dell is explicitly named as the hardware supplier for the massive Bitdeer/Volta AI data center project in Norway, indicating strong enterprise AI server demand. Risk: Hardware margins could be squeezed by frontier labs or data center operators like Volta.
More from Chamath Palihapitiya

This newsletter, published August 09, 2026, features Chamath Palihapitiya discussing BTDR, NVDA, TSLA, DELL. 4 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Chamath Palihapitiya  · Tickers: BTDR, NVDA, TSLA, DELL