Market structure
Compute is being financed as an asset. Power still arrives as a filing.
Four dated public shifts in capital, grid access, listed prices, and large-load rules.
Four shifts, summer 2026
- Capital
- >$500B platforms
- Grid access
- Texas queue audit
- Listed prices
- CME futures Oct 5
- Large-load rules
- FERC to six RTOs
Source boundary
These events are public third-party announcements and reporting. They are not GreenCIO measurements, customer results, or a claim that every cited source is ingested. Reported dollars and gigawatts stay attributed to the named outlet.
- >$500B
- Third-party capital platforms
- ~300
- Projects in the Texas audit
- Oct 5
- CME compute futures
- 6
- RTOs sent FERC show-cause
NVIDIA Newsroom, 10 Aug 2026. Not a GreenCIO figure.
Texas Tribune, 14 Aug 2026. Mostly data centers, per that reporting.
CME Group and Silicon Data announcement, 11 Aug 2026.
FERC §206 coverage, June 2026. Not a count of filings we ingest.
The four shifts
Each event leaves a diligence gap
Capital
NVIDIA announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to set up independent compute financing platforms meant to mobilize over $500 billion of third-party capital. NVIDIA described compute as becoming an investable asset class. Financing can close while power, residual value, and interconnection stay in the memo.
NVIDIA Newsroom, 10 Aug 2026
Open the case briefGrid access
Texas paused new data-center grid connections while the governor ordered an audit of interconnection projects. Reporting put the audit at about 300 projects, mostly data centers, and interconnection requests at 474 GW. ERCOT has a batch framework for large loads. The Texas Tribune wrote that the audit still lacked specifics, and that it is hard to find information about Texas data centers.
Texas Tribune, Utility Dive, POWER Magazine, Aug 2026
Review the screening methodListed prices
CME Group and Silicon Data announced compute futures for 5 October 2026, a listed contract on monthly GPU costs. A listed GPU-hour price does not say whether the megawatts arrive, on what topology, or on what start date.
CME Group / Silicon Data, 11 Aug 2026
Read the Compute Index methodLarge-load rules
FERC directed all six RTOs and ISOs to justify or reform large-load integration rules. Reporting on PJM covered interruptible treatment in heat waves and a data-center-linked power-price surge. Virginia ordered Dominion to assign some transmission costs to data centers.
FERC §206 coverage, Jun–Aug 2026
Open the regulatory panelAgainst those shifts
What we publish
Shipped methods and editorial analyses. Coverage still depends on permitted sources and qualification floors.
Compute Index
A listed GPU-hour price still needs a published method, a quote-age cutoff, and a refusal when coverage is thin.
No print when the panel is too thin or too stale
Read the index methodPower Diligence
Time-to-power is a separate question from the contract price. Screen region, stage, and declared constraints before capital commits.
Approved account required
Review the screening methodAnalyses
Public financing structures and tariff decisions, taken apart with a source boundary. Not customer results.
Editorial archive of third-party events
Open the analysesSource feed
Permits, capacity changes, and financing items, each with a publication date and a link to the original source, when the feed is current.
Omitted on this site when rights or freshness cannot be verified
Open the feedCase brief
What the NVIDIA announcement does not disclose
On 10 August 2026 NVIDIA said it was partnering with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to establish independent compute financing platforms. The stated aim was to mobilize over $500 billion of third-party capital. NVIDIA framed compute as becoming an investable asset class.
That announcement is a capital-markets event. It is not a power interconnection, a residual-value schedule, or a GPU-hour settlement. The named firms are not GreenCIO counterparties. This brief does not claim a relationship with NVIDIA or with those platforms.
What the announcement establishes
- Independent financing platforms, not a single NVIDIA fund.
- Third-party capital as the intended source of funds.
- Compute described as an asset that can be financed at scale.
What it leaves undisclosed
- Which sites get power, on what queue, and on what date.
- Who holds residual-value, construction, and grid-upgrade risk.
- The GPU-hour price, topology, and term behind the boxes.
A diligence file still needs the same three questions a listed GPU-hour contract cannot answer: can the megawatts arrive, what residual value sits with the lessor, and what price the compute actually clears. The same gaps show up in Meta's Hyperion financing and in a listed GPU-hour contract.
Regulatory panel
Large-load rules are being rewritten in public
Show-cause orders, interruptible treatment, and cost assignment. Not legal advice, and not a predicted outcome.
FERC large-load show-cause
FERC issued Section 206 orders directing all six RTOs and ISOs to justify or reform large-load integration rules. Interconnection treatment is now a national underwriting input, not a single-state anecdote.
McGuireWoods, Utility Dive, June–July 2026
PJM interruptibility
Reporting described a path to put data-center load onto backup power in heat waves, and a 76% PJM power-price surge tied to data centers. Interruptibility belongs in the credit file, not the footnote.
E&E News by POLITICO, May 2026; TechCrunch / Maryland Matters, Jun–Jul 2026
Who pays for the wires
Virginia ordered Dominion to assign some transmission costs to data centers. Ohio already converted requested capacity into minimum billing demand. Cost allocation is moving in parallel with queue freezes.
Utility Dive, 12 Aug 2026
Ohio tariff analysisRelated modules
Modules that bear on these shifts
Bring the site, the purchase, or the filing those shifts affect.
We will show which modules apply and name the evidence still missing.