Core Scientific, Inc.

Core Scientific (Nasdaq: CORZ), headquartered in Austin, Texas, is one of the largest operators of high-density digital infrastructure in North America. The company designs, builds, and operates data centers across roughly a dozen sites in Texas, Georgia, Oklahoma, North Carolina, and elsewhere, with two distinct legs: a legacy digital-asset business (self-mining and hosted Bitcoin mining) and a fast-growing high-performance computing (HPC) colocation business that houses GPU infrastructure for AI workloads. The company emerged from Chapter 11 bankruptcy in January 2024 and relisted on Nasdaq.

The core asset is power. Core Scientific controls gigawatts of already-interconnected, grid-secured electrical capacity at sites originally built to mine Bitcoin — capacity that is now being converted, building by building, into liquid-cooled AI data halls. As of Q1 FY2026 (reported May 6, 2026), approximately 350 MW of HPC capacity was energized, with 243 MW actively billing. The verified power position is 0.35 GW online, 0.68 GW contracted, and ~0.6 GW planned, for ~1.63 GW all-in owned/committed, plus a further ~1.5 GW under evaluation that is excluded from the headline total.

Critically, Core Scientific is the landlord, not the tenant. It owns its sites and power and leases capacity to AI operators under long-term take-or-pay structures — the colocation-lease relationship runs the opposite direction from a typical neocloud. Its anchor and, today, essentially only HPC customer is CoreWeave, which leases ~590 MW across five sites under a 12-year arrangement representing $10B+ of contracted revenue.

The bull case is contracted, long-dated, and power-anchored. Core Scientific's edge is not GPUs — it is gigawatts of energized, grid-secured power at sites that already have substations, interconnects, and utility relationships. In a market where the binding constraint on AI is megawatts and time-to-power, converting an operating Bitcoin mine to AI colocation is dramatically faster than greenfield. The CoreWeave anchor — ~590 MW across Denton, Dalton, Muskogee, Marble, and Austin under a 12-year take-or-pay lease worth $10B+ — gives the company a decade of visible, credit-backed cash flow. Management is layering on optionality: stated ~1.5 GW gross expansions at both Muskogee (anchored by the $421M Polaris DS acquisition, 440 MW contracted with OG&E) and Pecos, plus a ~3.0 GW leasable development pipeline. The $3.3B project bond closed in Q1 funds the buildout without diluting equity.

The bear case is concentration and conversion economics. One tenant — CoreWeave — is ~100% of HPC colocation revenue and the great majority of forward revenue, and CoreWeave is itself a leveraged, customer-concentrated neocloud (Microsoft and a handful of hyperscalers). If CoreWeave stumbles, Core Scientific's take-or-pay is only as good as CoreWeave's balance sheet. The company still carries a shrinking, increasingly unprofitable self-mining business; Q1 posted a $347M net loss driven by $266M of impairments. The October 2025 collapse of CoreWeave's ~$9B all-stock buyout — rejected by Core Scientific shareholders — removed a backstop and left CORZ independent and reliant on executing its own conversions on time and on budget.

The honest read: this is a power-and-real-estate story wearing an AI multiple. The megawatts are real and largely de-risked through 2026; the planned pipeline beyond CoreWeave is thinner than the ~3 GW headline implies (significant slices are behind-the-meter concepts and load studies, excluded from our verified total). The stock works if management converts grid-secured power into signed, multi-tenant offtake. It re-rates down if CoreWeave remains the only name on the rent roll.

Owns grid-secured power and sites and leases to AI operators under take-or-pay; CoreWeave brings the GPUs. Sits firmly on the power/land side, not a neocloud — but more than a pure lessor, since it builds, energizes, and operates the data halls.