In partnership with
⚡ THIS WEEK'S SIGNAL
Two regulators tested the same proposition this month — that hyperscalers pay for the wires that serve them — and got two different answers. Virginia's SCC said yes, going forward, on July 31. California's CPUC pulled its Google item off the August 13 calendar and moved it to September 3. A signed national pledge is not a rate mechanism. The docket is the rate mechanism, and the docket is still open.
━━━━━━━━━━━━━━━━━━━━━━━
📰 MAIN STORY
The vote moved again. On August 12 the California Public Utilities Commission (CPUC) pulled Draft Resolution E-5455 from its August 13 voting meeting and rescheduled it to September 3. E-5455 would approve, with modifications, PG&E's Advice Letter 7785-E — the agreement to energize a 250 megawatt (MW) Google load at 230 kilovolts (kV) on Disk Drive in San Jose. PG&E filed it December 18, 2025, requesting approval by June 1 to hold a January 2027 construction start and a December 2028 in-service date. It is now mid-August.
Everyone here wants something the filings state only partly. PG&E wants the load on its system on schedule, and argued in its January reply that the earlier data-center resolutions are expressly non-precedential and that there is no basis for assigning any share of broader network upgrade costs to a single customer. Google wants December 2028; it also signed the White House Ratepayer Protection Pledge on March 4, which commits signatories to pay for the power delivery infrastructure their data centers require. The Public Advocates Office wants the 75% refund cap the Commission applied to the STACK and Microsoft cases. And FERC is now in the room: the June 18 show cause order to CAISO (Docket EL26-71) drew a response deadline of August 17, and Commissioner LaCerte's concurrence explicitly asked state commissions to check their large-load retail tariffs against that same pledge.
Here is the shape of it. Google's refundable capital advance in this case is likely below $80 million, per the draft resolution. The Newark–NRS 230 kV line its facility connects to and depends on is a project of more than $1 billion, with LS Power's portion capped at $813 million and an approved revenue requirement of $102 million in year one declining to $41 million in year forty — roughly $2.8 billion over four decades, carried by ratepayers across the CAISO footprint. Google's 250 MW is about a quarter of that line's roughly 1,000 MW capacity.
The load forecast is the part worth screenshotting. CAISO's San Jose area long-term forecast moved from 2,100 MW in the 2021–2022 transmission plan to roughly 3,400 MW base and 4,200 MW sensitivity in the 2024–2025 studies — mostly data center driven, in CAISO's own words. Google's facility is about 20% of that 1,300 MW of growth, and Energy Division confirmed the project depends on more than ten additional South Bay upgrades. The draft's response is structural rather than numeric: raise the annual refund cap to 100% of net revenues actually received, extend the refund window from ten years to fifteen, and defer the network-upgrade question to the Rule 30 proceeding (A.24-11-007), with a Tier 2 advice letter due within 60 days of that decision. The draft does float two rough attribution benchmarks — about $1 million per MW, or about $650,000 per MW — without adopting either.
Scenario A: The Commission adopts E-5455 as drafted on September 3. The real cost-allocation event then shifts to A.24-11-007, and the Tier 2 advice letter becomes where the dollars actually land. For investors, PG&E's large-load capital program carries a recovery path contingent on that decision rather than settled by this one; the Rule 30 decision is the confirming or contradicting event. For ratepayers, nothing appears on a bill until Rule 30 resolves and the FERC-jurisdictional portion flows through PG&E's transmission owner formula rate.
Scenario B: The item holds again, or the Commission imports the 75% cap, or it decides to wait on whatever CAISO and the participating transmission owners filed into EL26-71. Then December 2028 becomes hard to hold, and California hands Nevada and Arizona siting teams a slide they will use for two years.
What to watch: the September 3 CPUC voting agenda — specifically whether Ordering Paragraph 4 still reads 100% of annual net revenues or reverts to 75%.
━━━━━━━━━━━━━━━━━━━━━━━
Your prompts are leaving out 80% of what you're thinking.
When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning — constraints, edge cases, examples, tone — and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.
89% of messages sent with zero edits. Used by teams at OpenAI, Vercel, and Clay. Try Wispr Flow free — works on Mac, Windows, and iPhone.
━━━━━━━━━━━━━━━━━━━━━━━
⚡ QUICK HITS
Virginia Just Told Dominion To Directly Assign Transmission Costs: In a July 31 order in the Rider T-1 case, the SCC concluded that new large-load data centers cause the transmission costs built to serve them, and directed Dominion to develop a policy assigning those costs directly. The order does not set the number — the follow-on tariff filing does, and that is the document to read. [Virginia Mercury]
The Pledge Is Getting A Statutory Backstop: The House Energy and Commerce Committee advanced the Ratepayer Protection Act (H.R. 9340) 52-0 on July 21, directing state commissions to consider requiring 100 MW-plus data center loads to cover full grid upgrade costs; two days later the White House expanded the pledge to 187 utilities, cooperatives and developers plus 23 governors. All seven original signatories are Mag 7 or adjacent — which makes the California docket a live test of what "pay for all new power delivery infrastructure" means when a utility argues the opposite on your behalf. [POWER]
A Governor Just Became A Party To A Merger Case: Abigail Spanberger announced August 6 she will formally intervene in the SCC's review of NextEra's $67 billion acquisition of Dominion — a first for a Virginia governor. Hearings are set for November with a ruling due January 15, and the intervention puts affordability questions into the record of the largest utility merger in U.S. history. [Virginia Mercury]
Ohio's Tariff Screened Out Three-Quarters Of Its Own Pipeline: Under the AEP Ohio large-load tariff approved in July 2025, more than 30 GW of preliminary interest narrowed to roughly 13 GW that paid for engineering studies and about 5.6 GW signed under the tariff. That is the clearest public number anyone has on the gap between announced capacity and contracted capacity, and every planner sizing a 2027 workplan should have it. [Forbes]
SPARK Selections Still Have Not Posted: DOE's Office of Electricity indicated August 2026 selections for the roughly $1.9 billion SPARK opportunity (DE-FOA-0003580), with awards October 2026 through January 2027; the program page still shows only the March key dates and was last modified in April. Applicants carrying projects in capital plans without a confirmed federal share are now two-thirds through the selection month with nothing to book. [DOE Office of Electricity]
The Hyperscaler Metric Quietly Changed: In Q2 2026 calls, Microsoft, Alphabet and Meta shifted emphasis from aggregate capex toward time-to-energy, power procurement and how fast a campus converts into revenue-generating compute. When the reported KPI moves from dollars spent to megawatts energized, the interconnection queue stops being an operations problem and becomes an investor-relations one. [Data Center Knowledge]
━━━━━━━━━━━━━━━━━━━━━━━
🔧 TOOL / RESOURCE OF THE WEEK
Draft CPUC Resolution E-5455: the full text of the Google San Jose energization resolution, including the Commission's own worked example of how the refund formula actually pays out. If you pay a California electric bill, this is the rare public document that shows the arithmetic connecting one customer's advance to your transmission rate — read pages 4 through 14. → [CPUC Published Documents]
━━━━━━━━━━━━━━━━━━━━━━━
💬 CLOSING THOUGHT
The pledge, the show cause orders, the Virginia order and the California draft all point the same direction, which is genuinely new. What none of them has yet is a number. For the optimism to hold, Rule 30 has to produce a method rather than another deferral, and the Dominion tariff has to survive its own filing. Three delays is a pattern, not a schedule.
So I'm curious what you're seeing. Is anyone building the Rule 30 outcome into a 2027 workplan yet?
