Core Scientific’s $215.4M colocation quarter precedes Bitcoin exit

By Nathan Brooks

Core Scientific booked $215.4 million in data-center colocation revenue in the third quarter of 2026, about 92% of $232.9 million in total revenue. On Nov. 5, CEO Adam Sullivan said: “We expect to complete our exit from Bitcoin mining by year-end.”

CFO Jim Nygaard said the company should enter 2027 as a pure-play colocation provider, renting data-center space and power to AI customers. Management gave that year-end deadline on the earnings call, making it a statement of intent. The written release only says remaining mining sites will be repurposed for colocation as circumstances allow.

What has changed in the revenue mix?

Colocation supplied about 92% of revenue, up from 83.2% in the second quarter and 18.4% a year earlier. Self-mining revenue fell to $12.3 million from $57.4 million.

Mining income rises and falls with Bitcoin’s price and network difficulty, which measures how hard miners must work to earn new coins. Contract rent does not move with either.

Adjusted EBITDA reached $133.6 million, about 57% of revenue, versus $41.1 million, or 25%, in the second quarter. The $1.18 billion net income came from a non-cash gain on warrants linked to the stock price. That gain does not establish operating profitability.

How did the stock react?

Core Scientific shares fell 10.39% on Nov. 4 and 10.82% on Nov. 5. They closed Nov. 5 at $11.38 after reaching a one-year intraday low of $10.59, then rose 3.34% to $11.76 on Nov. 6. That price valued the company at $3.78 billion, far below its one-year high of $30.46.

Who is filling the capacity?

Every megawatt Core Scientific currently bills is rented by CoreWeave. CoreWeave has signed for 590 MW of capacity, and 485 MW is billing across five campuses. Another roughly 100 MW is being tested before transfer, while Dalton’s second phase is due by early 2027.

Core Scientific says this billing capacity equals more than $700 million of annualized colocation revenue. For CoreWeave, with 3.7 GW of signed power and a $104 billion backlog, the Core Scientific lease is a small piece of its business.

CoreWeave shares rose 6.28% to $46.40 on Nov. 6, but remained down 47.54% over the preceding month. Its next report is scheduled for Nov. 10, 2026.

Where does AMD fit?

AMD is the next tenant, with 530 MW across five leases. Delivery starts at Pecos in the first half of 2027, with about half online that year and the rest through 2028.

AMD also holds reservation rights on about 1.9 GW more, beyond the five leases. Those rights do not represent delivered capacity.

Sullivan targets $1.8 billion in annual run-rate revenue and $1.6 billion in net operating income. Those targets apply to roughly 1.1 GW once fully delivered, with the ramp running through 2028.

The contracts provide a path from more than $700 million toward the $1.8 billion target. The bullish valuation argument says that path is worth more than the $3.78 billion market value implies.

Can the balance sheet carry the ramp?

The revenue pivot is tidy; the balance sheet is not. Shareholders’ equity stood at negative $1.092 billion, while long-term debt was about $4.3 billion.

Liquidity totaled $1.74 billion, including restricted cash, with $803.9 million in cash. The debt is more than five times the cash balance.

Nygaard expects about $650 million from Tranche 1 warrants around their January 2027 expiration. Each exercised warrant adds shares, and those warrants account for part of the gap between roughly 322 million shares outstanding and about 509 million on a fully diluted basis. Exercises would supply cash while increasing dilution.

What has to happen next?

The investment case rests on having a single tenant for several more quarters. Management guided to a double-digit sequential revenue increase in the fourth quarter.

Debt sets the order of losses if delivery slips: shareholders would take the loss first. The CoreWeave agreement underwrites the near-term ramp, while AMD is needed to establish a second source of demand.

Two milestones matter. CoreWeave’s Nov. 10 report is the next test of its ability to keep paying, while AMD’s first Pecos megawatts should begin billing in the first half of 2027. If Pecos misses that date, Core Scientific would likely remain dependent on one tenant longer, weakening the diversification case.

The next proof point is billing rather than projection: CoreWeave’s Nov. 10 report and AMD’s first Pecos megawatts will test whether fixed colocation rents can replace volatile mining income without extending single-customer dependence.