Finvest
CORZ Digital Infrastructure · Bitcoin mining · AI infrastructure · Data centers · Thesis updated July 19, 2026

AI data centers can rescue a miner

01 Running thesis

A miner racing toward AI

Core Scientific owns and operates power-heavy sites. That used to mostly mean Bitcoin mining. Now the main story is high-performance computing, or HPC, which means data centers built for AI chips and other intense computing work.

The bull case is clear. CoreWeave has signed for about 590 MW of HPC capacity, with total contracted value above $10 billion. Core Scientific has also started delivering capacity at Denton, with 8 MW in Q2 2025 and another 40 MW scheduled by the end of that quarter. This makes the pivot feel more real than it did when the contracts were only future plans.

The bear case is also real. Most of the 590 MW still has to be built and delivered, with full delivery now expected in early 2027. CoreWeave is still the key customer. If Core Scientific misses deadlines, faces supply chain delays, or cannot sign another large customer, the market may keep valuing it like a risky crypto-linked company.

The next big proof point is customer diversity. Management says it has several non-hyperscale prospects in the 50 MW to 100 MW range. A signed deal with a large enterprise or hyperscale customer would help show that Core Scientific owns scarce power assets, not only a single-customer AI hosting contract.

May 2025▲Core Scientific began proving the Denton ramp, with 8 MW of billable capacity expected in Q2 2025 and another 40 MW scheduled by the end of that quarter. Management also pointed to several 50 MW to 100 MW enterprise prospects, which could reduce customer concentration sooner than expected.
Feb 2025→The CoreWeave relationship expanded to about 590 MW and more than $10 billion of contracted value. The same update raised execution and financing concerns because full delivery moved to early 2027 and the new 70 MW deal requires Core Scientific to fund $1.5 million per MW.
Nov 2024▲CoreWeave committed to all 500 MW then offered, and Core Scientific shifted another 100 MW from Bitcoin mining toward HPC. The company also added an Alabama site with 11 MW of critical IT load and possible expansion to 55 MW more.
Aug 2024▲The initial public thesis formed around the pivot from Bitcoin mining to HPC hosting. At that time, management cited 382 MW contracted for HPC hosting and about $6.7 billion of projected aggregate revenue over 12-year contracts.
02 Business model

Power sites become rent checks

Core Scientific makes money in three main ways. It mines Bitcoin for itself, hosts third-party Bitcoin miners, and hosts HPC customers that need dense power for AI hardware. The hosting model provides space, power, cooling, and site operations.

The strategic focus is HPC hosting. These contracts can last 12 years and usually pass power costs through to the customer. That can make revenue more predictable than Bitcoin mining, where revenue changes with Bitcoin price and network competition.

The capital model is changing. Earlier CoreWeave deals were customer-funded, but the newer 70 MW Denton expansion requires Core Scientific to fund $1.5 million per MW. In return, Core Scientific receives full rental payments during the first two years because there is no capex credit tied to that deal.

That shift could raise returns, but it also raises financing risk. The company has more than 1,300 MW of contracted power infrastructure, which is a scarce asset if AI demand stays high. But turning power into finished data centers takes money, equipment, permits, and time.

03 Product portfolio

What Core Scientific sells

Growth engine

HPC hosting

This is the main growth bet. Core Scientific builds and operates high-density data center capacity for customers such as CoreWeave.

Cash cow

Bitcoin self-mining

The company runs its own ASIC miners and earns Bitcoin. This can be profitable when Bitcoin is strong, but cash mining cost reached $51,035 per Bitcoin in Q4 2024.

Steady

Digital asset hosting

Core Scientific hosts third-party Bitcoin miners by providing power, space, and operations. This legacy segment is expected to shrink as contracts expire.

Steady

Contracted power portfolio

The company controls more than 1,300 MW of contracted power infrastructure. That power access is the base for both mining and AI data center expansion.

Option

New enterprise HPC deals

Management says the pipeline includes non-hyperscale customers looking for 50 MW to 100 MW deployments. Signing one would reduce reliance on CoreWeave.

Option

Next-generation mining systems

Core Scientific is developing mining solutions using new Block 3-nanometer chipsets. The goal is better mining efficiency, though HPC remains the larger strategic focus.

04 Business segments

Revenue is still mining-heavy

Digital Asset Self-Mining84%declining
Digital Asset Hosting7%declining
HPC Hosting9%growing fast

The mix is based on Q4 2024 revenue: $79.9 million from Digital Asset Self-Mining, $6.5 million from Digital Asset Hosting, and $8.5 million from HPC Hosting. The mix should shift if Core Scientific delivers the contracted CoreWeave capacity on time.

05 Risk factors

What could break the pivot

HPC build-out delays

High impact · Medium odds

The full 590 MW CoreWeave build-out is now expected in early 2027, later than the prior mid-2025 target. Large data center projects need power equipment, cooling, permits, and construction work to line up. More delays would push out revenue and weaken trust in management execution.

We watchTrack delivered HPC megawatts versus the target of about 250 MW by the end of 2025.

CoreWeave concentration

High impact · High odds

CoreWeave is the anchor tenant for the HPC pivot and represents most of the contracted HPC load. That gives Core Scientific huge visibility, but it also ties the story to one customer. Management wants CoreWeave to be less than 50% of critical IT load by the end of 2028, but no new major customer has been announced yet.

We watchWatch for a signed HPC contract with a new hyperscale or large enterprise customer.

Weak mining economics

Medium impact · Medium odds

The self-mining business is sensitive to Bitcoin price and the total computing power on the Bitcoin network. Core Scientific's direct cash cost to mine one Bitcoin rose to $51,035 in Q4 2024. If Bitcoin falls or network difficulty rises, mining can become a drag while HPC capacity is still ramping.

We watchTrack Bitcoin price, network hash rate, and Core Scientific's cash cost per Bitcoin mined.

More company-funded capex

High impact · Medium odds

The newest 70 MW Denton agreement requires Core Scientific to fund $1.5 million per MW. Future customers may also ask the company to contribute capital. That could improve rents, but it can pressure the balance sheet and increase financing needs.

We watchWatch capex per MW, debt or equity financing, and whether new contracts include customer capex support.

Contract economics disappoint

Medium impact · Medium odds

New enterprise customers may not match the economics of the CoreWeave deal. Rental rates, power pass-through terms, and capex credits will decide whether growth creates strong returns. A deal can look big in megawatts but still be less valuable if Core Scientific funds too much of the build.

We watchCompare rental rates, term length, and capex contribution in any new HPC contract.
06 Quick answers

In one breath

Is Core Scientific still a Bitcoin miner?

Yes. Digital Asset Self-Mining was still the largest segment in Q4 2024 revenue. But the company's main growth plan is now HPC hosting for AI and GPU-heavy computing.

Why does the CoreWeave contract matter?

CoreWeave has contracted about 590 MW of HPC capacity with Core Scientific, with total contracted value above $10 billion. That gives long-term revenue visibility if Core Scientific builds and delivers the capacity on schedule.

What is the biggest catalyst for CORZ?

The biggest catalyst is a major HPC contract with a new customer. It would prove that demand is broader than CoreWeave and help move the company toward its goal of reducing CoreWeave below 50% of critical IT load by the end of 2028.

Why is the stock still risky?

The company has to execute a large data center build-out, fund more capex than before, and manage a mining business with volatile economics. The market is already giving value to the AI pivot, so missed deadlines or weak new deal terms could hurt.