A miner trying to become AI infrastructure
- The big idea is a pivot from Bitcoin mining to long-term AI and HPC data center leases.
- Keel says its infrastructure assets represent a 2.2 GW power capacity pipeline.
- Management reported about $533 million of liquidity in cash and Bitcoin as of May 8, 2026.
- Legacy mining is being cut back, with hashrate expected to fall from about 14 EH/s to about 5 EH/s by year-end 2026.
- Q1 2026 revenue was $37.0 million, but the company still posted a $145.4 million net loss.
The lease test
Keel is trying to turn old Bitcoin mining power sites into AI data centers. HPC means high-performance computing, the heavy computer work used for AI models and other large workloads. If Keel signs long-term leases with giant cloud companies or AI cloud firms, the business could start to look more like a power-backed infrastructure utility.
The bull case is clearer after the latest update. Management said zoning is complete at Panther Creek, Sharon, and Moses Lake. It also said about $533 million of cash and Bitcoin fully funds work needed to advance those sites through lease execution, start construction at Moses Lake, and cover general costs through 2028.
The bear case is still real. Keel has not yet proven it can build and run AI-grade data centers at scale. New GPU chips such as Vera Rubin can change power, cooling, and layout needs, so the engineering work may keep moving. The key test is simple: can Keel secure its 3 targeted hyperscaler or neocloud leases by year-end 2026?
From hashpower to rent
Today, Keel still makes most of its money from Bitcoin mining. It sells computing power to mining pools under Full Pay Per Share contracts, which pay miners based on a formula for expected Bitcoin mining rewards. That cash is meant to help fund the AI data center pivot.
The future model is different. Keel wants to develop and own power generation, grid connections, and data centers, then lease that capacity under long-term contracts. If it works, revenue should depend more on signed leases and less on the daily price of Bitcoin.
The break point is timing. Mining cash is shrinking on purpose, while AI data center revenue is not expected to start until 2027. That leaves a period where the company is spending heavily before the new business proves itself.
What Keel is building
Legacy Bitcoin Mining
This is the current revenue base, but it is being wound down. Management expects hashrate to fall from about 14 EH/s to about 5 EH/s by year-end 2026.
Power and data center pipeline
Keel says its infrastructure assets represent a 2.2 GW power capacity pipeline. The plan is to turn that power into leased AI and HPC data center capacity.
Panther Creek
Panther Creek is one of the first three sites aimed at near-term AI infrastructure leases. Zoning is complete, and management says permits are on track.
Sharon
Sharon is another near-term site in the lease pipeline. It matters because customer demand will likely judge Keel by how many sites can be made ready together.
Moses Lake
Moses Lake is the near-term site where management says liquidity covers the start of construction. It is a key proof point for the buildout plan.
Scrubgrass load study
Scrubgrass is an expansion option, with a detailed load study underway for an additional 750 MW. It is not the first test, but it could matter if early leases arrive.
Revenue is still North American mining
The mix below is Q1 2026 continuing revenue by geography, because Keel does not yet report HPC revenue. It excludes discontinued Argentina and Paraguay operations; the U.S. was 52% of revenue and Canada was 48%.
What could break the pivot
No anchor leases
High impact · Medium oddsKeel needs large cloud or AI customers to sign long-term leases before the story changes. Without those contracts, the 2.2 GW pipeline is only potential, not a cash-flowing business. Missing the 3 targeted leases by year-end 2026 would weaken the bull case.
Buildout delays
High impact · Medium oddsAI data centers need exact power, cooling, networking, and uptime standards. Keel is moving from mining sites to more complex digital infrastructure, and its own filing says executive management has limited or no prior experience in the space. A delay can push revenue further out while costs keep running.
Cash burn before revenue
High impact · Medium oddsLiquidity risk is lower after management reported about $533 million of cash and Bitcoin. Still, Q1 2026 showed a $145.4 million net loss and negative cash flow from operations. If spending rises before leases are signed, Keel may need more capital.
Bitcoin bridge weakens
Medium impact · Medium oddsThe legacy mining business is supposed to help fund the transition, but it is being cut back. Revenue still depends on Bitcoin price, network difficulty, and power costs. If mining profits fall faster than planned, the bridge to 2027 gets thinner.
Power cost and regulation
Medium impact · Medium oddsBoth mining and AI data centers use a lot of electricity. Keel says governments and regulators are paying more attention to energy use by data centers, and power costs can change with markets and policy. Higher power costs would make leases harder to price and mining less useful as a funding source.
In one breath
Is Keel still a Bitcoin miner?
Yes, but Bitcoin mining is now the legacy business. Keel is keeping mining mainly to generate cash while it shifts power sites toward AI and HPC data centers.
When could Keel start making money from AI data centers?
The internal view expects HPC infrastructure to be pre-revenue until 2027. The bigger near-term sign is not revenue yet, but signed long-term leases with large cloud or AI customers.
Why does liquidity matter so much for Keel?
Keel must spend money before the new data centers produce revenue. Management said about $533 million of liquidity funds the three near-term sites through lease execution and covers general costs through 2028.
What changed in the latest update?
Keel completed the Paso Pe sale, ending its Paraguay operations, and confirmed the Latin America exit. Management also said zoning is complete at all three near-term sites.