Core Scientific was one of the most visible bankrupt Bitcoin miners of the 2022 crypto downturn. Less than four years later, it says it has more than $24 billion in potential long-term contract revenue, a market value of roughly $7 billion, and major ties with AI industry participants including CoreWeave and AMD.
The shift was not simply a recovery trade tied to Bitcoin. The company effectively took the same electricity, land, and facility footprint once used for mining and repositioned those assets as scarce AI data center infrastructure.
From public miner to roughly $4 million of cash
Core Scientific’s predecessor was incorporated in late 2017, and the company began operating under the Core Scientific name in 2018. In January 2022, it went public on Nasdaq through a merger with special purpose acquisition company Power & Digital Infrastructure Acquisition Corp., trading under the ticker CORZ.
At the time, its main businesses included self-mining Bitcoin, hosting services for third-party miners, and mining machine sales. That model depended on three things holding up at once: high Bitcoin prices, reliable machine uptime, and stable power costs.
In 2022, all three deteriorated. Bitcoin fell from a peak near $69,000 in November 2021 and briefly dropped to about $16,000 in November 2022. At the same time, network hash rate and mining difficulty kept rising, forcing miners to deploy more computing power and more electricity to produce the same amount of Bitcoin.
Energy costs added more pressure. In bankruptcy-related filings, Core Scientific said its electricity expense reached about $106 million in the first half of 2022. During the summer, power prices kept climbing, and the company also had to curtail electricity use repeatedly at the request of suppliers.
Customer default became another problem. After crypto lending platform Celsius filed for bankruptcy in 2022, it entered into a dispute with Core Scientific over hosting contracts. Core Scientific said Celsius had failed to pay about $7 million in related fees and continued to cause additional losses.
By the third quarter of 2022, Core Scientific posted a quarterly net loss of about $434.8 million. On Dec. 21, 2022, it formally filed for Chapter 11 protection in the U.S. Bankruptcy Court for the Southern District of Texas. On the filing date, it had about $4 million of cash on hand.
The problem was not a lack of assets. The company had spent aggressively to expand mining operations, using debt to buy mining rigs and build data centers. When Bitcoin prices, power costs, and financing conditions turned at the same time, those fixed assets could not be monetized quickly, while cash flow dried up first.
Restructuring preserved the assets that mattered most
Chapter 11 did not force an immediate liquidation. Core Scientific was able to keep operating during the bankruptcy process while renegotiating debt and equity with creditors, equipment financing providers, and existing shareholders.
In January 2024, the company completed its restructuring and exited bankruptcy. It said the process cut about $400 million of debt and preserved about 724 MW of operating capacity, along with data centers, land, grid interconnection, and related infrastructure across several U.S. states.
Existing shareholders were not wiped out completely. Under the restructuring plan, they received new shares and warrants in the reorganized company and were allowed to participate in a related rights offering. If all warrants are exercised and the rights offering is included, those securities could represent as much as about 60% of the reorganized company’s equity.
The new shares resumed trading on Nasdaq on Jan. 24, 2024. The stock opened at about $5.89 and closed at about $3.44 that day. At first, the market still viewed Core Scientific mainly as a post-bankruptcy Bitcoin miner trying to repair its balance sheet.
What survived the restructuring later became the base for its re-rating. Mining machines can lose value quickly, and Bitcoin output is exposed to coin prices and network difficulty. A data center site with grid access, substation infrastructure, and large-scale land is much harder to replicate in the short term.
CoreWeave changed how the market valued mining-site power
Core Scientific’s partnership with AI cloud infrastructure company CoreWeave marked the turning point in how investors looked at the business.
The two sides had already worked together on an early project of about 16 MW in Austin, Texas. In June 2024, Core Scientific announced a 12-year contract with CoreWeave to provide about 200 MW of high-performance computing infrastructure. The company said the agreement could generate more than $3.5 billion of cumulative revenue over its full term.
Under the initial deal structure, about $300 million of capital required for the project was to be funded by CoreWeave and offset through future hosting fees. That reduced Core Scientific’s early construction burden and gave it room to convert former mining sites into high-density data centers suitable for GPU clusters.
The partnership kept expanding. Long-term contracted capacity signed in 2024 rose from about 200 MW to about 502 MW through several additions of about 70 MW, 112 MW, and 120 MW. That figure excluded the earlier Austin project of about 16 MW.
In February 2025, the companies added about 70 MW at the Denton, Texas project. Core Scientific said it expected to fund about $104 million of capital expenditures for that expansion, with the rest of the related capital supplied by CoreWeave.
Including the earlier Austin work, the contracted capacity between the two sides reached about 588 MW, which the company disclosed as about 590 MW. The contract term remained 12 years, and potential cumulative revenue rose to about $10.2 billion.
Those contracts changed the market’s frame of reference. Investors who once focused on Bitcoin output, hash rate, machine efficiency, and BTC holdings began watching different indicators: contracted capacity, billable MW already online, long-term contract value, and delivery timelines.
The same megawatt can be valued in very different ways. In Bitcoin mining, its worth depends on coin prices, network difficulty, and machine efficiency. In AI hosting, that same power can be tied to contracts that run for more than a decade and produce a longer-duration revenue stream. The assets did not fully change, but the use case and the cash-flow profile did.
A $9 billion acquisition that never closed
CoreWeave was not only one of Core Scientific’s largest AI hosting customers. It also tried twice to buy the company.
In 2024, CoreWeave offered $5.75 a share in cash for Core Scientific, but the proposal was rejected by the board. Core Scientific said at the time that the bid materially undervalued the company and its growth prospects.
In July 2025, the two sides reached another acquisition agreement. Under that deal, Core Scientific shareholders would receive 0.1235 shares of CoreWeave Class A common stock for each CORZ share they held.
Using the methodology described in the deal announcement, and based on CoreWeave’s five-day volume-weighted average price, the transaction valued Core Scientific’s fully diluted equity at about $9 billion. Based on CoreWeave’s closing price on July 3, 2025, that implied a value of about $20.40 per Core Scientific share.
But it was a fixed exchange-ratio deal, not a fixed-cash offer. As CoreWeave’s stock declined, the implied value for Core Scientific shareholders fell with it. By Sept. 2, 2025, that implied value had dropped to about $11.41 a share.
Some Core Scientific shareholders worried the transaction would give up the upside of remaining independent in exchange for stock in another AI infrastructure company facing its own high capital spending, customer concentration, and debt pressure.
On Oct. 30, 2025, Core Scientific shareholders voted down the deal, and the companies later terminated the merger agreement. Even so, the roughly $9 billion valuation became an important reference point for how the market measured the value of Core Scientific’s power and data center assets.
From a $5.75-a-share cash bid in 2024 to a stock deal worth about $9 billion in 2025, CoreWeave’s two attempts showed how far Core Scientific’s center of gravity had moved. The company was no longer being valued mainly for mining rigs and Bitcoin production, but for power capacity that could be delivered on a relatively fast timeline.
AMD pushed the story into its next stage
The CoreWeave contracts helped establish that a Bitcoin mining site could be converted into an AI data center. The AMD relationship added a second layer by reducing reliance on one customer and one project set.
In 2026, Core Scientific announced a large-scale infrastructure partnership with Advanced Micro Devices, or AMD. Based on the company’s second-quarter 2026 disclosures, the relationship could support as much as about 2.5 GW of leasable capacity. Within that, signed 15-year agreements covered five projects totaling about 529 MW and represented more than $14 billion of potential base contract revenue.
Beyond the signed capacity, AMD also secured rights to reserve about 1.925 GW of additional capacity. If all of those rights were converted into formal projects, the total potential scale of the partnership could reach about 2.5 GW. Those reservation rights, however, are not the same as signed revenue contracts. They still depend on site selection, grid interconnection, construction, and end-customer demand.
As part of the arrangement, AMD received warrants to purchase up to 30 million shares of Core Scientific stock at an exercise price of $23.47 per share. After a related lease was signed on July 27, 2026, about 6.5 million of those warrant shares had vested and become exercisable. The structure made AMD more than a technology and industry partner; it also gave AMD a path to share in any future stock appreciation.
Within the AMD-related signed capacity, about 152 MW was leased by AI infrastructure operator Neocloud. Core Scientific, Neocloud, and AMD also signed a three-party credit support agreement mainly intended to protect AMD equipment deployed at the sites and to give AMD the right to step in under certain Neocloud default scenarios. The documents did not describe AMD as an unconditional guarantor of all Neocloud payment obligations.
As of July 2026, Core Scientific said it had contracted about 1.1 GW of customer power capacity tied to more than $24 billion of potential contract revenue. Of that, about 437 MW had already begun billing, representing about $635 million of annualized GAAP hosting revenue.
From roughly 590 MW under CoreWeave-related contracts to about 529 MW linked to AMD and Neocloud, Core Scientific was beginning to move away from a single-customer story and toward the profile of a data center developer serving multiple AI industry participants.
Financial results show the transition, but not a cash windfall
Core Scientific’s revenue mix has changed in a visible way.
In the second quarter of 2026, total revenue was about $164.2 million. High-density hosting contributed about $136.7 million, or about 83% of total revenue, while self-mining digital asset revenue came in at about $21.5 million.
High-density hosting generated about $80 million of gross profit in the quarter, with a gross margin of about 59%. By contrast, the self-mining business posted a gross loss of about $12.2 million, with a gross margin of negative 56%. Adjusted EBITDA for the quarter was about $41.1 million.
That means high-density hosting is no longer just a future plan. It has become Core Scientific’s main source of revenue and gross profit, while the mining business that once carried the company has become a smaller and less profitable residual segment.
Still, there is a clear gap between contract size and accounting profit. Core Scientific posted a net loss of about $1.155 billion in the second quarter, including about $1.046 billion tied to changes in the fair value of warrants and contingent value rights. Those losses were mainly driven by the company’s rising share price, and they were non-cash accounting items rather than equivalent cash outflows.
Even so, the company still reported a GAAP operating loss of about $78.5 million for the quarter, showing that it had not yet reached a stable stage of accounting profitability.
In the first half of 2026, net cash provided by operating activities was about $230.9 million. That figure included about $208.2 million from sales of digital assets, along with customer prepayments and other working-capital changes. A positive operating cash-flow figure therefore does not automatically mean the data center business can independently cover all construction and financing needs.
Heavy capital spending and a larger debt load remain key risks
The AI data center transition still requires a large amount of capital.
In the first half of 2026, Core Scientific spent about $954 million in cash on property, plant, and equipment, of which about $181 million was provided by CoreWeave. It also paid about $233 million to acquire land and development rights. Together, those two uses of cash totaled about $1.187 billion.
To support construction, the company issued $3.3 billion of senior secured notes in May 2026 with a 7.75% coupon and a 2031 maturity. As of June 30, 2026, long-term debt had increased from about $1.06 billion at the end of 2025 to about $4.3 billion.
That is why more than $24 billion of potential contract revenue should not be read as $24 billion of cash already in hand. Those revenues would be recognized gradually over more than a decade, and they depend on projects being built on schedule, customers continuing to perform, and data centers operating reliably.
As of the close of trading on Aug. 5, 2026, Eastern Time, Core Scientific shares stood at $21.77, up about 533% from the $3.44 close on Jan. 24, 2024, the first day of trading after its relisting. Based on about 321.3 million shares outstanding as of July 23, the company’s market capitalization was roughly $7 billion.
That rally reflects, to a large extent, how the capital markets are pricing in future cash flow ahead of realization. Investors are not buying profits the company has already booked. They are buying the possibility that its power footprint can be converted into AI data center revenue over time.
What Core Scientific is really selling is energized delivery time
The turnaround is not a case of a bankrupt miner simply recovering because Bitcoin rebounded. It is a change in infrastructure use and in the framework investors use to value the business.
In the Bitcoin mining era, Core Scientific’s core assets were mining machines, hash rate, and sites with access to low-cost power. In the AI data center era, the relative importance of mining rigs falls, while land, grid interconnection, substations, fiber, cooling systems, and project delivery capability move to the center.
With demand for AI data centers rising in the U.S. and grid access often taking time, customers are not just buying electricity. They are buying the ability to obtain large-scale energized capacity within a defined schedule. A mining site that already has location work done, grid access secured, and room to expand may be able to enter service faster than a data center built entirely from scratch.
In that sense, Core Scientific is not only selling megawatts. It is selling what can be described as energized delivery time. By controlling land and power resources early and then signing long-term contracts with AI customers, it is turning infrastructure once tied to Bitcoin mining into longer-duration hosting cash flow.
The risks did not disappear; they changed form. Where the company was once exposed mainly to Bitcoin prices, electricity costs, and mining difficulty, it is now more exposed to capital spending, financing costs, customer concentration, construction delays, and contract performance. If projects are not delivered on time, or if customer demand and financing conditions shift, more than $24 billion of potential contract revenue may not be fully realized.
From about $4 million of cash and Chapter 11 protection to a roughly $7 billion market value and more than $24 billion of potential contract revenue, Core Scientific has clearly gone through a major re-rating.
But the deeper shift is not that it suddenly controls more Bitcoin. It is that the company has moved from being a miner tied to crypto price cycles to being a data center developer whose main product is power capacity and delivery timing. Whether it has truly moved past the high-leverage risks that defined the previous cycle will depend on whether those power plans are turned into billable data center assets on schedule.