Core Scientific Bitcoin Mining Exit: Why It’s Pivoting
Core Scientific plans to exit Bitcoin mining by 2026. See what its revenue shift to AI data centers means, and the operational risks the pivot brings.

Core Scientific Is Leaving Bitcoin Mining. What Comes Next?
Core Scientific says it expects to stop Bitcoin mining by the end of 2026 and move into 2027 by focusing on high-density data center colocation. That's a bigger shift than its earlier messaging around AI hosting as something added alongside mining. Its latest quarterly numbers make the change hard to miss: mining made up just 7.5% of reported third-quarter revenue, while colocation services accounted for the majority.
For a company built around mining infrastructure, the move looks like a response to economics. Mining revenue has weakened, equipment values have fallen, and the company has already committed to hosting compute workloads for AI and high-performance computing customers. Still, trading mining for data center hosting doesn't erase risk - it just swaps what kind of risk the business faces, and where demand depends.
In this article, we'll look at what Core Scientific has announced, what its revenue mix says about the transition, and what could determine whether the plan holds up. The key point is that the company isn't simply "replacing Bitcoin with AI." It's moving toward a different infrastructure business, with different constraints and incentives.
Why is Core Scientific exiting Bitcoin mining?
Core Scientific expects to complete its mining wind-down during the fourth quarter of 2026. On its third-quarter earnings call, CFO Jim Nygaard said the company aims to enter 2027 as a "pure-play, high-density colocation provider." In other words, the transition has been underway for a while - but the timeline is moving faster.
The company previously disclosed that hosted mining would end by December 31 as it reduced its own mining operations. And in December 2025, reporting by The Wall Street Journal said Core Scientific was targeting a mining wind-down by 2028. Compared with that, the current plan - end of 2026 - is a clear acceleration.
The pressure is economic. In the first quarter of 2026, Core Scientific recorded $266.5 million in mining-related asset impairments. The company pointed to weaker Bitcoin prices, historically low hashprice, and declining secondary-market values for mining equipment. Hashprice measures the expected revenue miners earn relative to their computing power. When it drops, the same machines earn less - while electricity and other operating costs don't go away.
There's also the cost of changing course. During the second quarter, Core Scientific terminated a mining equipment agreement with Block and Proto and recorded a $41.9 million loss tied to the transition. Shutting down or repurposing mining operations isn't just a paperwork exercise - contracts, machinery, and facility plans may need to be adjusted before a new line of business can use them.
The business case is changing, not disappearing
Bitcoin mining depends on Bitcoin's price, network competition, energy costs, and machine efficiency. Those factors can shift quickly, and a miner with high fixed costs can look profitable in one stretch and pressured in the next - even if the facilities are still physically useful.
Data center hosting works differently. Instead of earning Bitcoin by contributing computing power to the network, a provider rents space and infrastructure to customers running their own workloads. In the best case, that can mean steadier commercial visibility, depending on customer contracts and who actually signs. But it also forces the operator to hit demanding requirements around power, cooling, reliability, and day-to-day service.
So this isn't "volatile mining versus guaranteed AI revenue." It's one set of operating and market risks traded for another. Core Scientific's decision suggests management believes its infrastructure is better used for customer compute than for continuing to mine under current economics.
What do Core Scientific's revenue figures show?
Core Scientific reported $232.9 million in third-quarter revenue. Of that, $215.4 million came from colocation services. Bitcoin mining contributed $17.5 million, which is 7.5% of the total.
You can see the strategic shift reflected directly in the revenue mix. By the third quarter, hosting was the dominant part of the company's reported business.
Mining revenue also broke down between self-mining and hosting other miners. Self-mining revenue fell to $12.3 million from $57.4 million a year earlier. Revenue from hosting other miners dropped to $5.2 million from $8.7 million. Put together, those comparisons suggest the pressure stretches across more than one slice of the mining operation.
The company also said it had nearly 50% fewer miners online at the end of September than at the end of 2025. That lines up with a planned wind-down, but it also means the operating base that produces mining revenue is shrinking in real time. The transition shows up both in machine counts and in what mining contributes to revenue.
Revenue mix is useful, but it isn't the whole story
A revenue percentage doesn't tell you whether the business is profitable, how the cash flow looks, or how strong any specific customer contract is. Colocation being larger than mining indicates the mix has shifted - but it doesn't prove the new model will deliver better returns.
Investors and customers would still need to evaluate capital spending, operating costs, contract duration, facility utilization, and the practical work required to complete the transition.
That distinction matters when you hear headlines about a company becoming an AI infrastructure provider. Data centers don't run on demand alone - they need real power and cooling, plus physical configuration and operational readiness for customer workloads. Sometimes converting an existing site can be faster and cheaper than building from scratch. But it still isn't automatic.
Even so, the figures do help explain why the exit is commercially plausible. If mining supplies a small share of revenue and the company already has a larger hosting business, keeping mining running may not be the best use of capital or capacity. Whether that's still true after conversion depends on what Core Scientific earns from hosting - and what it costs to deliver that capacity.
How is the company shifting to AI data centers?
Core Scientific's AI hosting strategy started forming in 2024. The company signed an initial hosting agreement with CoreWeave in March and later announced a larger deal in June. At the time, it described AI hosting as an expansion alongside mining and said it would preserve Bitcoin mining capacity. The current plan is different: it expects to end mining and concentrate on high-density colocation.
Colocation generally means providing customers with data center space plus the supporting infrastructure for their own equipment. In a high-density facility, the operator needs to handle substantial compute demand within a limited footprint. AI and high-performance computing can also demand more electrical capacity and more intensive cooling than facilities built for lighter workloads. Because requirements vary by customer and hardware, "data center capacity" isn't a single interchangeable product.
For a former miner, some assets may be a starting point. Mining operations already require access to power, buildings, electrical systems, and operational support for heavy equipment. That makes repurposing at least conceivable. But mining sites aren't automatically ready for AI hosting. Customers still need dependable service and infrastructure built around their workload, and the provider has to meet the standards set in the commercial agreements.
The transition depends on execution at the facility level
A practical way to judge the pivot is to break it into four questions:
- Capacity: Can the facilities deliver the power and cooling density customers require?
- Conversion: What equipment, construction, or upgrades are needed, and how long will they take?
- Utilization: How much of the available capacity is actually contracted and in service?
- Economics: Does customer revenue cover operating costs, investment, and the costs of winding down mining?
These questions matter because treating "AI" as a self-explanatory source of growth can blur the real issue: a site being ready in theory isn't the same as having projects come online on schedule. In infrastructure, the gap between a signed commercial plan and revenue-producing capacity can be large.
The shift also changes how the company should be measured. Mining performance tends to be evaluated using factors like hashprice, fleet efficiency, and power costs. A colocation business also has to show reliable delivery, customer retention, facility utilization, and disciplined capital spending. Core Scientific is moving from producing Bitcoin with its own machines to selling infrastructure services to customers. Even where the buildings and power connections overlap, it's a different operating job.
What risks could affect the transition?
The first risk is execution. Core Scientific has to finish winding down mining while preparing its facilities for a different use. If conversion takes longer or costs more than expected, the company could face a stretch where mining revenue keeps shrinking before replacement capacity becomes productive. The reported mining impairments and the contract termination loss already show that the transition has come with financial costs.
The second risk is customer and workload concentration. Hosting revenue depends on customers deploying and using capacity under commercial agreements. Core Scientific's source material points to CoreWeave as an early and significant counterparty, but it doesn't include enough detail to judge the full duration, economics, or concentration of customer commitments. Those details matter because a data center can be physically ready yet still sit underused if customer plans change.
A third risk is electricity cost and availability. Large data centers need substantial, reliable power. That requirement can spark friction with local concerns about electricity prices, grid capacity, and competing land and resource needs. Reuters reports that recent US midterm results added political uncertainty, with some opponents of data center development among winners of Democrats' gains. Local resistance can slow permitting and project timelines even if demand from technology companies stays strong.
The same site can carry different risks under a new model
Mining companies and data center operators both run energy-intensive facilities, but the commercial risk shifts. For miners, economics are tied to Bitcoin's market price and network competition. For hosts, economics rely more on customer demand, contract terms, utilization, and operational performance. Power still matters in both cases, but it sits inside a different revenue relationship.
For Core Scientific, the question isn't whether AI infrastructure is a larger market than Bitcoin mining. The question is whether this company can convert specific facilities into dependable capacity and earn an adequate return on the capital required. Sector-level demand doesn't answer that company-level question.
There's also a political angle, and it deserves a practical read, not a hand-wave. Local resistance isn't just a messaging issue if it delays access to power or construction. At the same time, political attention doesn't prove every proposed data center will get blocked. The outcome depends on grid conditions, planning decisions, community concerns, and whether projects can address them.
Key takeaways
- Core Scientific expects to finish its Bitcoin mining wind-down by the end of 2026 and enter 2027 focused on colocation.
- Mining generated 7.5% of third-quarter revenue, while colocation generated $215.4 million of the $232.9 million total.
- The pivot responds to weaker mining economics, but conversion costs and facility readiness still matter.
- AI hosting brings different risks, including customer utilization, power availability, project execution, and local opposition.
Conclusion: a new business, not a risk-free one
Core Scientific's planned exit from Bitcoin mining comes after its business mix already shifted. Mining contributed just 7.5% of third-quarter revenue, while colocation services accounted for most of the total. Its earlier AI hosting effort has now effectively become the company's main plan.
The real test is execution. Core Scientific has to complete the wind-down, convert facilities to customer needs, and make that capacity commercially usable. The decision may fit with the mining pressures it's reported, but it doesn't turn data center hosting into a low-risk business. If you're assessing the company, the most useful next signals are facility readiness, customer utilization, what additional investment is needed, and whether the company can secure reliable power.