Summary

  • Corvex reported approximately 1.5MW of current critical IT capacity. Its approximately 8MW figure is an expectation for the end of 2026, with both deployments expected ready for service in Q4 and expansion revenue anticipated from Q1 2027.
  • The company executed definitive agreements for the two site deployments, yet separately identified capital, customer commitments, equipment delivery, utility energisation and local permitting as conditions that can still prevent completion.
  • Corvex has a 12.5MW Midwest right of first refusal, not 12.5MW of contracted capacity. It says total capacity could exceed 20MW only if it exercises the right and contracts for that capacity.

The useful correction to Corvex’s “5x” and “13x” framing is not skepticism about scale. It is a ledger of what each number has earned. A current operating measure, an expected service-ready measure, a right to make a later deal, and a financing close answer different questions. The fact that the private placement later closed makes one of those ledgers more concrete; it does not settle the other three.

The current receipt is approximately 1.5MW

The release says Corvex currently has approximately 1.5MW of critical IT power capacity. That is the reported operating baseline. It is the correct starting point for a reader asking what capacity exists now, rather than what might be installed, sold or commissioned later.

Corvex says it has definitive agreements to double capacity at its existing Mid-Atlantic facility and establish operations at a second existing enterprise-scale Midwest data centre. Both facilities have live utility power, a meaningful development advantage over a greenfield site. But a site agreement inside a live building is not the same receipt as GPUs delivered, racks installed, power made available to Corvex equipment, clusters accepted, or a customer placed on service. The company itself sets the distinction by describing the 8MW result as expected and the sites as expected to be ready in Q4.

Eight megawatts is a dated expectation

The two deployments are expected to lift critical IT capacity from 1.5MW to approximately 8MW by the end of 2026. Corvex expects the Midwest cluster to add roughly 2,000 latest-generation GPUs and the Mid-Atlantic expansion another 1,000. It anticipates revenue from the expansions beginning in Q1 2027.

These are not trivial disclosures. They specify a timetable and an intended equipment scale. But the operative verbs are expected and anticipated. The same release names the gates: capital availability, customer commitments, equipment delivery, utility energisation and local permitting. It also says there is no assurance that the expansion will be pursued or completed. A capacity figure is therefore not an operating-capacity receipt merely because it is tied to an existing building with live utility power.

The right monitoring sequence is physical: procurement, delivery, installation, energisation, testing, customer activation and revenue. A press release can promise the endpoint before each intermediate receipt exists. It cannot replace them.

The 12.5MW is a right to choose, not a lease of capacity

The Midwest right of first refusal is the point most likely to be silently converted into capacity. Corvex says it secured a right of first refusal on an additional 12.5MW, anticipated to be ready for service in Q3 2027. It says total critical IT power could exceed 20MW if it exercises that right and contracts for the capacity.

“If” appears twice in the economic architecture. Corvex must elect its right, and it must contract for the capacity. It also says it will market the capacity to prospective customers and that no customer agreement, price or duration is assured. A right of first refusal can be strategically valuable: it preserves a choice on scarce powered space. Yet it is not the same thing as a binding capacity purchase, an energised deployment, a customer take-or-pay commitment or a revenue-producing GPU cluster.

That difference matters for capital planning. The option may create a future claim on space while leaving the company exposed to the later question of whether customer commitments, price, hardware supply and funding justify its exercise. Presenting it as 12.5MW of “contracted” capacity removes precisely the decision Corvex still has to make.

The PIPE moved from expected to closed—but gross is not an allocation

On August 31 Corvex described a private placement expected to generate approximately US$33m in gross proceeds before fees and expenses. Its September 4 Form 8-K records that the private placement closed on September 2. That is a real state change. It is no longer accurate to call the placement unclosed.

It would be equally inaccurate to treat the US$33m headline as a completed project budget for the 8MW deployment or the 12.5MW right. The company describes multiple intended uses: critical IT capacity expansion, Amplified AI Cloud, Token Factory and Assured AI. The June 30 10-Q, which predates the closing, reported about US$21.7m in cash and cash equivalents, US$3.8m of quarterly revenue and continuing operating losses. It does not allocate the later financing to a particular facility or asset.

Closing proves one financial receipt, not every allocation or delivery receipt downstream of it. Net cash after fees, timing of use, equipment commitments, working-capital needs and the division between physical capacity and product development all remain separate evidence questions.

Four ledgers prevent a capacity headline from doing too much work

The operating ledger starts at approximately 1.5MW. The deployment ledger asks whether the two sites reach service-ready 8MW on the stated schedule. The option ledger asks whether Corvex exercises and contracts for the 12.5MW right, then obtains the necessary customers and execution evidence. The capital ledger records the completed PIPE, then asks how net cash is actually allocated and whether it funds the intended work without creating a new constraint.

Each ledger can advance independently. A PIPE close does not commission a cluster. A definitive site agreement does not exercise an ROFR. A customer conversation does not create take-or-pay revenue. Keeping the distinctions visible gives the next filing room to add real evidence instead of forcing one headline to carry every future outcome.

Sources