Core Scientific Plans Near-Total Bitcoin Liquidation to Fund AI/HPC Shift, Converts Pecos Site to 430MW Colocation

Core Scientific will sell nearly its entire Bitcoin treasury to accelerate an AI and HPC buildout, pivoting away from self-mining as Pecos, Texas shifts to 430MW colocation.

Bitcoin
Cryptocurrency
Regulations
Economy
Because Bitcoin
Because Bitcoin

Because Bitcoin

March 3, 2026

Core Scientific just made a clear capital allocation choice: prioritize power, racks, and revenue visibility over Bitcoin on the balance sheet. The company said it intends to sell nearly its entire Bitcoin treasury this year to finance an aggressive buildout of AI and high‑performance computing (HPC) data centers—an explicit step away from self‑mining and toward long‑duration colocation contracts.

Here’s the fulcrum: replacing volatile, hashprice‑driven cash flows with contracted AI/HPC demand. In the current cycle, that trade can look rational. Compute buyers continue to bid up scarce high‑density capacity, and the cost of capital for data infrastructure has stayed elevated. Holding BTC becomes an opportunity cost when each megawatt you deploy into colocation can be leased at attractive rates for years.

Key moves and numbers - Treasury strategy: The firm disclosed in an SEC filing that it expects to sell the vast majority of its Bitcoin in 2026, with most disposals front‑loaded into Q1, subject to market conditions. Management said it currently holds less than 1,000 BTC. - January sales: It sold 1,900 BTC for $175 million in January at prices “materially” above current market levels—evidence of tactical timing rather than forced liquidation. - The pivot: Bitcoin went unmentioned in the latest earnings presentation. Core Scientific is now positioning itself as a digital infrastructure leader for high‑density colocation serving AI/HPC workloads. - Pecos conversion: The Pecos, Texas site is being converted from Bitcoin mining to colocation and can support up to 430 megawatts of gross power capacity. - Timeline: Management guided that every megawatt in the portfolio will be dedicated to colocation within three years—an eventual end to self‑mining if they stay the course.

Financial context - Q4 revenue mix: Self‑mining still led the quarter with $41.1 million, versus $31.3 million from colocation and $6.5 million from hosting for customers. That mix underscores the execution gap they must close as self‑mining winds down. - Profitability: Q4 net income came in at $216 million, a swing from a $291 million loss a year ago, while total revenue declined to $70 million from $94.9 million as self‑mining contracted. - Market reaction: Shares fell 6.4% Tuesday to $15.43, though they’re up 52% over the past year after peaking around $23.63 in November.

Sector read‑through Core Scientific isn’t alone. Cango recently sold 4,451 BTC to fund its AI effort. Bitfarms rebranded to Keel Infrastructure and said it is “no longer a Bitcoin company,” leaning fully into HPC. Industry players that once optimized for hash rate are now optimizing for power density, uptime SLAs, and long‑term lease yields.

My take on the decision quality - Technological reality: High‑density AI clusters need power, cooling, and interconnects that mining sites already approximate. Converting at scale, especially in power‑rich locations like Pecos, is a defensible moat if they execute on thermal and network constraints. - Balance sheet cognition: Selling BTC lowers treasury beta and funds capex without incremental dilution or expensive debt. For operators, that reduces reflexivity—revenue and equity no longer swing directly with hash economics. - Competitive psychology: Announcing an all‑MW colocation target within three years sets a commitment device internally and a signal externally. It raises the bar for execution but also anchors customers and partners around a predictable roadmap. - System trade‑offs: As miners divert capacity, Bitcoin’s network may see marginal shifts in hash distribution, but the broader market seems comfortable sacrificing some miner‑held inventory for an acceleration in compute capacity where demand is insatiable.

This story is less about “leaving Bitcoin” and more about cash flow engineering. If Core Scientific signs durable, high‑quality AI/HPC contracts against that 430MW runway and beyond, selling coins now could be the cleanest route to compounding infrastructure value while reducing exposure to crypto cyclicality.