Neocloud Brief · neocloud stocks · what is a neocloud · neocloud ETF
CORZ — Core Scientific: the bitcoin miner that became an AI landlord
Convert power and real estate into long-dated AI colocation leases, finance the build with debt, and hope the contracted backlog is worth more than the cost of delivering it.
01The 30-second read
Fast read
Core Scientific is the highest-profile example of a Bitcoin miner reinventing itself as an AI data-center landlord. It leases purpose-built, high-density colocation capacity on long-term contracts to two anchor tenants: CoreWeave (~590 MW, ~$10.2B over 12 years) and AMD (530 MW, $14B+ over 15 years, with rights to up to 2.5 GW) — roughly 1.1 GW and $24B+ of base contracted revenue.
The catch is capital. Q2 2026 revenue was $164.2M (+109% YoY) at $41.1M adjusted EBITDA, but the quarter produced a ~$1.16B GAAP net loss (mostly non-cash warrant revaluation and impairment), and the company carries ~$4.41B of debt against negative book equity of ~−$2.42B. Bulls see a backlog that dwarfs today’s revenue; bears see a business that must keep raising capital to deliver it.
02What Core Scientific actually is
Core Scientific builds and leases high-density data-center capacity. It is not a cloud and it does not sell GPUs — it sells the building, the power, and the cooling that GPUs need. The model is “AI data-center landlord”: sign a long-term lease with an AI customer, construct a liquid-cooled facility sized to their racks, then collect rent for a decade or more. Capital comes from debt and project financings; returns come from the spread between build cost and contracted rent.
The history matters. Core Scientific was a large Bitcoin self-miner that filed for Chapter 11 in December 2022 and emerged and relisted on Nasdaq on January 24, 2024. Since then it has pivoted from mining to colocation, converting its power and real estate into AI capacity. Mining is being wound down — self-mining at only two sites by mid-2026, with nearly 30% fewer miners online than at the end of Q1, and mining assets impaired in Q2.
Its campus footprint spans the U.S. The CoreWeave program runs across Denton, TX (~260 MW), Dalton, GA (~175 MW), Muskogee, OK (~70 MW), Marble, NC (~65 MW), and Austin, TX (~20 MW), while Pecos, TX and Hunt lead the AMD build-out. The CoreWeave relationship began as a 16 MW lease at Auburn in 2024 and grew into the ~590 MW program it is today — a useful illustration of how these contracts expand.
03The bull case
1 · Contracted revenue is enormous
- ~1.1 GW of contracted billable capacity and $24B+ of base contracted revenue
- CoreWeave: ~590 MW tied to ~$10.2B over 12-year terms
- AMD: 530 MW, $14B+ over 15 years (2.5% annual escalators) plus an exclusive right over up to 2 more GW
2 · Execution is ahead of schedule
- 437 MW billing by mid-July 2026 — nearly 200 MW more than Q1’s 243 MW, ahead of plan
- Four of five CoreWeave campuses substantially complete before end of summer
- Colocation revenue $137M at a 59% gross margin; guided higher in Q3
3 · Capital structure is deliberate
- $3.3B of 7.750% senior secured notes due 2031 closed Apr/May 2026 (priced 99.25% of par)
- $600M of committed senior secured credit facilities added Aug 27, 2026 ($100M revolver + $500M letter of credit), expected to release ~$300M of restricted cash
- ~$1.8B of liquidity at Q2-end
- AMD build (~$6B) expected to be funded at the project level; AMD-backed leases carry credit support for the full 15-year term
4 · The Street is constructive
- Consensus “Strong Buy” — 17 analysts polled by S&P Global
- Average target ~$37.13, range $28–$55 — roughly +106% vs. $17.99
- Other aggregators lower ($29.78–$31.96), but all see upside
04The bear case
1 · The balance sheet is stretched
- ~$4.41B of debt vs. ~$1.77B cash — net cash −$2.64B
- Book equity ~−$2.42B (negative)
- TTM net loss −$1.63B (−$5.07/share)
2 · Free cash flow is deeply negative
- TTM operating cash flow +$520M vs. capex −$1.48B → FCF −$958M
- Build cost $11–12M per MW; AMD’s 530 MW alone needs ~$6B
- Prepared to spend up to ~$1B ahead of contracts on ~500 MW
3 · Concentration cuts both ways
- Two counterparties each committed to 500+ MW — effectively two tenants deep
- CoreWeave is both largest customer and the rejected suitor: a ~$9B takeover was voted down and terminated Oct 30, 2025 (second failed attempt after June 2024)
- AMD is a chip designer, not a cloud operator — a newer tenant type here
4 · Report volatility, overhang, and the tape
- Q2’s ~$1.16B loss driven by a ~$1.05B non-cash warrant revaluation + $266.5M impairment; EPS −$3.32 badly missed −$0.06
- ~20.2% of shares sold short; beta 5.48
- Stock down ~4.6% in a month, ~41% off the $30.46 high
05The numbers
| Metric | Value | Note |
|---|---|---|
| Q2 2026 revenue | $164.2M | +109% YoY; beat ~$144.8M consensus |
| Q2 colocation revenue | $137M | GAAP; straight-lined over 12-yr CoreWeave leases |
| Q2 net loss | ~$1.16B | +$1.05B warrant revaluation; $266.5M impairment |
| Q2 EPS | −$3.32 | Missed −$0.06 consensus |
| Q2 adjusted EBITDA | $41.1M | vs. $28.5M a year earlier |
| Gross margin | 43% / 59% | Consolidated / colocation |
| TTM revenue | $440.3M | |
| TTM net loss | −$1.63B | −$5.07 per share |
| Free cash flow (TTM) | −$957.8M | OCF +$520.2M; capex −$1.48B |
| Cash | ~$1.77B | |
| Total debt | ~$4.41B | Includes $3.3B 7.750% notes due 2031 |
| Net cash | −$2.64B | Book equity ~−$2.42B |
| Credit facilities | $600M | $100M revolver + $500M LC; announced Aug 27, 2026 |
| Contracted capacity | ~1.1 GW | 437 MW billing as of mid-July 2026 |
| Contracted revenue | $24B+ | “Base contracted revenue" (company-defined) |
| CoreWeave program | ~590 MW | ~$10.2B projected / 12 years |
| AMD program | 530 MW | $14B+ / 15 years; up to 2.5 GW |
| Build cost | $11–12M/MW | AMD 530 MW ≈ $6B capex |
| Shares outstanding | 321.3M | Short interest ~20.2% of shares |
| Consensus | Strong Buy | 17 analysts |
| Avg target | ~$37.13 | Range $28–$55 |
| Implied upside | ~+106% | vs. $17.99 |
06What could change the story
Bull triggers
- Q3 revenue steps up clearly as the 437 MW billing for a full quarter
- AMD delivery hits schedule — Pecos’s first MW in H1 2027, ~half the contract in 2027, the rest in 2028
- Project-level bond financing for the AMD build-out lands on attractive terms
- Mining wind-down completes, producing a cleaner P&L into 2027
- A third customer is signed, easing two-tenant concentration
Bear triggers
- A construction, energization, or equipment delay pushing rent commencement out
- Punitive financing terms, or a capital-markets window that shuts for levered AI developers
- A customer-concentration event at CoreWeave or AMD
- Continued GAAP losses and warrant/equity dilution outrunning demand for the shares
- Higher-for-longer rates raising the cost of a project-finance-heavy model
07What’s next
2026
Q3 earnings (estimated), after close. The test of the “meaningful step-up” from $137M colocation revenue, plus any AMD-build financing.
2027
Pecos first megawatts. The first AMD capacity is due online; ~half the 530 MW contract is scheduled across 2027.
AMD ramp completion. The remaining half of the AMD program is due by end-2028; management targets a ~600 MW/year build cadence.
08FAQ
Is Core Scientific a Bitcoin miner or an AI company?
In transition. Mining is being wound down — self-mining at two sites by mid-2026, mining assets impaired in Q2. The growth story is high-density colocation leased to AI customers. The miner label should fade, but it has not yet.
How big is the contracted backlog?
Roughly 1.1 GW of contracted billable capacity and $24B+ of base contracted revenue — CoreWeave’s ~590 MW (~$10.2B / 12 years) plus AMD’s 530 MW ($14B+ / 15 years), with an AMD reservation right over up to 2 GW more.
Why is a $24B backlog worth only ~$5.8B?
Because the backlog is not free: delivering it needs billions of capex ($11–12M per MW), the company has ~$4.41B of debt and negative book equity, free cash flow is deeply negative, and the market is discounting execution, financing, and concentration risk.
Is CORZ profitable?
No. Q2 2026 was a ~$1.16B net loss, though mostly non-cash (warrant revaluation and impairment). Adjusted EBITDA was positive at $41.1M. This page is research, not a recommendation.
Sources & method
- Core Scientific Q2 2026 results, press release and 10-Q (July 28, 2026): revenue $164.2M (+109%), colocation revenue $137M, gross profit $70.0M, adjusted EBITDA $41.1M, net loss ~$1.16B (warrant revaluation +$1.05B; $266.5M mining impairment), EPS −$3.32, liquidity ~$1.8B
- Core Scientific Q2 2026 earnings call transcript (July 28, 2026): ~1.1 GW contracted / $24B+ base contracted revenue; 437 MW billable; $11–12M per MW build cost; ~$6B for AMD 530 MW; up to ~$1B ahead-of-contract investment
- AMD partnership announcement (July 27–28, 2026): 15-year agreements, 530 MW across five sites, $14B+ base contracted revenue, 2.5% escalators, up to 2.5 GW reservation
- CoreWeave lease history: February 2025 expansion to ~590 MW across six sites and ~$10.2B projected revenue over 12-year terms (company release; SEC EX-99.1)
- Financing: $3.3B 7.750% senior secured notes due 2031, priced 99.25% of par, closed May 6, 2026 (company release; Cooley; The Globe and Mail)
- Credit facilities: $600M of committed senior secured facilities — a $100M revolving credit facility and a $500M letter of credit facility — announced August 27, 2026, expected to release ~$300M of restricted cash (company release; Nasdaq; Cooley)
- Corporate history: Chapter 11 filed December 2022; emergence and Nasdaq relisting January 24, 2024 (company release; CNBC); CoreWeave ~$9B acquisition rejected and merger terminated October 30, 2025 (Reuters; Bloomberg; CNBC)
- Market data: price $17.99 (Sep 18, 2026 close), market cap ~$5.78B, EV $8.42B, 52-week range $13.36–$30.46, shares 321.3M, short interest ~20.2%, beta 5.48 (stockanalysis.com, Yahoo Finance, GuruFocus)
- Consensus: 17 analysts via S&P Global aggregation — Strong Buy, average target $37.13 (range $28–$55); other aggregators lower (MarketBeat ~$29.78–31.96, Public.com ~$31.02, TipRanks Strong Buy)
- Next earnings: estimated October 23, 2026, after close (stockanalysis.com, GuruFocus) — company-confirmed date not yet issued, so treat as estimated