Core Scientific Just Became a 57%-Margin Landlord. The Market Is Pricing It Like a Warning.
Colocation EBITDA more than tripled and the Bitcoin wind-down is nearly done — yet the stock has halved in ninety days on a sector-wide funding scare that management is openly trying to buy into.
CORZ · Earnings Call · 2026-11-05
A business inflecting while its stock halves
Core Scientific delivered, on paper, the quarter it had been promising for two years. GAAP colocation revenue rose nearly 60% sequentially to $215 million. Adjusted EBITDA more than tripled to $134 million from $41 million. Adjusted EBITDA margin jumped to 57% from 25% in a single quarter. Billing capacity went from 120 MW at the start of the year to 485 MW of the 590 MW contracted across five campuses. On any operating scorecard this is the breakout quarter of the company's public life.
And the stock is down 48.6% over the last ninety days, a clean seventeen-week slide that put it 48.6% below the June high of $29.16 — giving back every bit of a +336% run since January 2024. The divergence is not a Core Scientific problem. It is a funding-market problem. Across the whole market, the AI data centers theme averaged -30.7% over the last thirty days across 66 tickers, with 55 of them down; data center revenue fell -38.7% across a 17-name basket; high bandwidth memory names dropped -40%. The sector that Core Scientific now almost purely belongs to is being re-priced on cost of capital, not on demand.
CEO Adam Sullivan walked straight into it, and I suspect he is right:
the overarching theme right now in the market is capital is the throttle on all new development. Capital will be the throttle to bringing any compute online... What we're seeing in the market today are not only one-off data center developers, but also some private equity-backed data center companies who are just running out of capital.
That is the honest read of the tape. Fewer bidders, same assets.
The second-tenant question finally has an answer
For three straight quarters, analysts asked Core Scientific the same thing: when does a customer other than CoreWeave show up? In March the answer was a work-in-progress — “today, we sit with 500 megawatts under exclusivity arrangements with a large investment-grade counterparty” — Adam Sullivan, Executive (likely CEO or President) · 2026-03-03. In May that exclusivity lapsed and management pivoted to re-marketing Pecos and Muskogee, insisting “We are in conversations with all of the hyperscalers, chip makers, AI labs, Neo Clouds” — Adam Sullivan, Chief Executive Officer · 2026-05-06.
This quarter the waiting ends. AMD signed for an initial 530 MW across five campuses, with a reservation over roughly 1.9 GW more at Muskogee, Hunt and Pecos. Note the keyword history: exclusivity was one of the company's biggest gainers two quarters ago and is now one of its biggest decliners — because the word has stopped describing the business. Contracted capacity is now 1.1 GW across two customers, which management expects to produce $1.8 billion of average annual run-rate revenue and $1.6 billion of NOI. The strategic point buried mid-call is the credit one:
“Though they represent less than half of our total contracted capacity today, the AMD contract, directly or through credit support, represents over half of our contracted stabilized NOI.” — Adam Sullivan, Executive · 2026-11-05
That is the answer to the CoreWeave-concentration critique that has dogged this name since 2024 — and it is a better answer than a second neocloud would have been.
Buying when the bidders thin out
Here is the part I find genuinely non-consensus. Management could easily have leaned into the story that powered land is appreciating because AI demand is insatiable. Sullivan said the opposite: valuations on powered land are "generally still in line with where they were over the course of the past 12 months," but "the amount of bidders on those projects is getting smaller." Core Scientific, with $1.7 billion of liquidity and roughly $650 million of cash-only warrant proceeds expected around January 2027, is positioning as the buyer. It already showed the template by paying about $440 million for Polaris earlier this year.
The financing leg is less settled. Project bonds remain the expected path for the AMD build, and the CFO was explicit that the market is functioning — “the project bond market is absolutely open” — Jim Nygaard, Executive · 2026-11-05 — even as he conceded diligence has become "a bit more comprehensive." Five separate leases means five ways to package and sequence a deal. Management also committed to not issuing more equity or equity-linked securities to fund phase one of AMD.
What the call glosses over is how much has already been spent. Capital expenditure hit $565 million in the June quarter, up 366% year over year. Free cash flow was -$602 million. And effective net cash has gone from +$821 million at the end of 2024 to roughly -$1.8 billion — a swing of $2.6 billion in six quarters. That is what financing risk means in practice.
Politics, Batch Zero, and the pure-play
This was the first Core Scientific call where the political question outranked the commercial one. Five of the eleven analyst questions were about Texas, Georgia, permitting, or the midterm elections. Management's line was that the outlook is unchanged — "along the same lines as it was three days ago prior to the election" — and leaned on process: “in Texas, things are never done retroactively. It's a business-friendly state.” — Adam Sullivan, Executive · 2026-11-05 The substantive detail is that Pecos and Hunt are already through stormwater permits, are pursuing permit-by-rule for backup generators, and that Hunt has entered ERCOT's November Quarterly Stability Assessment — the final gate before energization. The sector-wide phrase for that queue, Batch Zero, was the single highest-momentum keyword in the entire market last quarter, which tells you the regulatory bottleneck is now the industry's shared obsession, not one company's. New to this company's vocabulary too: Governor elect Tuberville in Alabama and Governor-elect Mazzei in Oklahoma, both cited as supportive.
And then the quiet structural change. Nearly 50% fewer miners were online at quarter-end than at year-end 2025; the wind-down completes in Q4, and 2027 begins as "a pure-play, high-density colocation provider." The AMD contract is the reason that sentence is now safe to say. Core Scientific did not lose a debate about Bitcoin mining — it outgrew the business, and the revenue mix made the decision for it.
The verdict. Operating execution is no longer the question; capital formation is. If the project bonds price well in early 2027 and AMD exercises its chunkier reservation deadlines through next year, this quarter's $134 million of EBITDA is a waypoint, not a peak. If the funding window narrows further, a 48.6% drawdown becomes the opening move. Management is betting the first, and for the first time in this company's history it has contracted revenue, delivered megawatts, and a credit-worthy second customer to back the bet with.