In partnership with
⚡ THIS WEEK'S SIGNAL
The federal transmission money everyone was waiting on already landed. On September 24, DOE named 31 projects in 26 states for $1.9 billion under SPARK, its speed-to-power program. Read the cost-share column before the headline: recipients are putting up $3.35 billion of their own, and for a regulated utility, "their own" means rate base. A program sold as bill relief just added $3.35 billion of utility capital that someone has to earn a return on.
━━━━━━━━━━━━━━━━━━━━━━━
📰 MAIN STORY
The list exists. DOE's Office of Electricity announced on September 24 that it intends to help fund 31 grid-improvement projects across 26 states under SPARK — Speed to Power through Accelerated Reconductoring and other Key Advanced Transmission Technology Upgrades — at $1.9 billion federal against $5.25 billion total (DOE announcement). These are selections, not signed grants: DOE's project summary labels every figure "subject to change pending finalization of formal financial assistance agreements," with awards expected between October 2026 and January 2027. We flagged this program last month in Everyone's Arguing Over Who Pays. The Transformer Line Is Three Years Long — that the program named for speed funds conductors, not transformers. The list confirms it.
Read what each party is optimizing for. DOE wants capacity inside existing rights-of-way, because new corridors take a decade it does not have; applicants wrote to that preference. Utilities want federal dollars against capital they would otherwise fund entirely from rates. Four states — Colorado, Oklahoma, Ohio and Indiana — came in as prime applicants themselves, a different instrument than an investor-owned utility filing alone. No hyperscaler is a recipient, and none needs to be: Alabama Power's project names data centers as the load it serves, Duquesne Light's names customers "the existing system cannot accommodate," and TVA's is titled Rebuilding the Transmission Backbone for Energy and AI Dominance. Mag 7 campuses get headroom on lines that federal money and other people's rates are upgrading.
Here is the contrast. The four small cooperatives on the list received a 75% federal share — Kit Carson Electric in Taos gets $38.1 million of a $50.8 million project. The largest investor-owned utilities hit a $50 million ceiling, so Duke Energy Carolinas' federal share is 24.8% of a $201.4 million project and the New York Power Authority's is 18.5% of $270 million. Across all 31, federal dollars cover 36%. The other 64% is recipient cost share, and for a rate-regulated transmission owner that capital enters the asset base it earns a return on and recovers from customers. DOE says the selections will "help reduce consumer electricity bills." Both things are true at once.
The numbers worth keeping. $1,893,241,306 federal, $3.36 billion in cost share, $5.25 billion total, 31 projects, 26 states. DOE expects more than 1,500 miles reconductored or rebuilt and grid-enhancing technologies across nearly 21,000 miles, making over 23 GW available — though the Assistant Secretary's quote in the same release says "more than 20." Four Topic Area 3 projects take $810.4 million, 42.8% of all federal money, and two of those target the seam between the Eastern and Western interconnections: Colorado's $250 million of a $1.21 billion intertie modernization, and Oklahoma's Three Corners Connector, $250 million of $832 million, developed by Grid United. That is 26.4% of the program spent moving power between the two halves of the country. For scale on the demand side, Dominion's Virginia data center requests alone exceed 70GW.
Scenario A — the agreements execute on schedule. If DOE finalizes the agreements by January with cost share and domestic-content terms intact, advanced conductor and grid-enhancing technology suppliers get something this market has never had: a named, multi-year, federally anchored order book across 26 states. Analysis: that is a volume signal for a value chain constrained by order visibility, not demand. On the operator side, a co-op like Kit Carson energizes capacity at a quarter of what its own members would otherwise carry.
Scenario B — negotiations slip or scopes shrink. DOE wrote that caveat itself. Then 23 GW becomes an announcement figure rather than operating capacity, the gap we have tracked since 90GW were announced with 2GW running in July. Recipients that already booked the work into capital plans fund it from rates, and customers pay for the whole project instead of 64% of it.
One thing to watch: whether DOE posts executed agreements — not selections — for the two interconnection-seam projects before January 2027. Those two carry a quarter of the program's federal dollars and the longest negotiation risk.
━━━━━━━━━━━━━━━━━━━━━━━
🧾 DEAL DECODER
The instrument: a DOE financial assistance agreement with non-federal cost share.
Who signs what. DOE's Office of Electricity and the prime applicant — here, parties as different as Duke Energy Carolinas, the Colorado Energy Office and a Taos cooperative. Not a grant yet, not a tax credit, not a loan. A selection is an invitation to negotiate, which is why DOE marks every figure in Selected Applications for DE-FOA-0003580 (SPARK), September 2026 subject to change.
Who fronts the capital, and who gets it back how. Federal funds are generally reimbursed against incurred cost, so the recipient spends first and claims after. The match must be non-federal — like an employer match that only arrives once you contribute. For an investor-owned utility, that matching capital is ordinary plant: rate-based, earning a return over decades.
Who eats the loss if the load never materializes. Nobody, contractually, and that is the decode. Unlike the slot reservation agreements we took apart on September 1, no deposit is forfeited if the data centers never arrive. The capacity sits in rate base.
Which precedent it sets. A state energy office can prime a billion-dollar interregional project no single utility would file alone.
━━━━━━━━━━━━━━━━━━━━━━━
10 Stocks at the Center of AI’s $1 Trillion Boom
AI’s growth story is much bigger than chatbots.
The market was already valued at an estimated $800 billion by late 2025, and continued adoption could push it beyond $1 trillion in the years ahead. That growth will require more data centers, computing power, software, networking and intelligent devices.
But the most obvious AI names may not be the only companies positioned to benefit.
MarketBeat’s free 10 Best AI Stocks to Own in 2026 report identifies ten publicly traded companies helping power the next phase of the AI buildout. Some are established leaders. Others occupy less obvious corners of the expanding AI ecosystem.
This free report names 10 stocks positioned to ride that wave before the rest of the market catches on.
━━━━━━━━━━━━━━━━━━━━━━━
⚡ QUICK HITS
FERC Sent PJM's Data Center Backstop Auction Back, and the Chair Was Blunt. FERC only partly approved PJM's reliability backstop procurement on September 29, faulting its cost allocation, transmission owner exit rules and collateral terms, and PJM shelved an auction due to open the next day; Chairman Laura Swett said the commission would not accept "a deeply flawed, eleventh-hour procurement mechanism with billion-dollar implications for consumers." For anyone serving load in PJM, the number to sit with is the roughly $2 billion in collateral Northern Virginia Electric Cooperative told FERC it would have had to post. [Utility Dive]
A Bipartisan Senate Bill Would Assign Data Center Transmission Costs Outright. The Bipartisan American Affordability and Jobs Act, introduced October 1 by Whitehouse, Heinrich, Lee and Capito, would expand FERC's project-by-project permitting authority, scrap National Interest Electric Transmission Corridors, and require transmission costs be fully assigned to data center owners, with mandatory exit charges above 20 MW. If it advances after the November elections, it federalizes the question Virginia and California are each answering one docket at a time. [Utility Dive]
MISO Wants to Fast-Track Large Loads That Bring Their Own Generation. MISO filed its Large Load Addition Resource Study process at FERC on October 2, asking approval by December 2: a 120-day paired review for loads above 200 MW matched with new generation in the same resource zone, ten applications per study period, non-refundable deposits, 15-year pairing. For planners, the real filter is the requirement that load and generation commercial operation dates align within three years. [Utility Dive]
Alabama Ratepayers Are Funding $134 Million of a Data Center Enabler. Alabama Power was selected for $50 million of a $183.9 million project whose stated purpose is unlocking capacity "for large-scale data centers, industrial manufacturers, and major commercial facilities." The federal share is 27%; the remaining $133.9 million is utility capital recovered through Alabama rates, which makes this the clearest small-print example of today's main story. [DOE selected applications]
Google's 250 MW San Jose Resolution Is Still Waiting on a Vote. Resolution E-5455, PG&E's exceptional-case agreement to energize Google's 250 MW San Jose load under Advice Letter 7785-E, was last calendared for the CPUC's October 8 voting meeting after coming off three prior agendas. California is the one place where a hyperscaler's cost responsibility is being set deal by deal rather than by tariff, and no California utility appears anywhere on the SPARK list. [CPUC docket]
🧭 THE EXPOSURE MAP
If you hold equities. Analysis: the segment structurally strengthened is advanced conductors, dynamic line rating and power flow control hardware, plus the firms that install them — a category constrained by order visibility rather than demand. Selected transmission owners get partial federal funding of capital they would otherwise rate-base entirely, which changes the financing mix more than the growth story. Confirming or contradicting events: Q3 earnings calls in late October, and whether any selectee revises capex guidance to reflect federal share. Companies named here are participants in a structure, not recommendations.
If you run projects at a utility. If you were selected, the question your organization probably cannot answer yet: what is the fully loaded cost of grant compliance — domestic content documentation, milestone reporting, wage determinations — and was it inside the budget you submitted in May? If you were not selected, a neighboring system just bought capacity cheaper than you can build it.
If you set or influence policy. Federal co-investment is now a live variable in state cost allocation. The next venue is each selectee's rate filing, where a commission decides whether the federal share is credited to customers or absorbed into the revenue requirement.
If you pay a utility bill or buy AI compute. The mechanism is rate base, reduced at the margin. A federally funded dollar earns the utility no return; a cost-share dollar does, recovered through a transmission component typically 18 to 36 months after the spending decision. Nothing here changes a bill this quarter or reprices your AI tools.
━━━━━━━━━━━━━━━━━━━━━━━
🔧 TOOL / RESOURCE OF THE WEEK
DOE's Selected Applications for DE-FOA-0003580 (SPARK). The primary document: all 31 projects with prime applicant, location, federal funds and total project value. Utility program and portfolio managers can see what a peer system committed to, and at what cost share, before the agreements are signed. → energy.gov/documents/selected-applications-de-foa-003580-spark
━━━━━━━━━━━━━━━━━━━━━━━
💬 CLOSING THOUGHT
This is a real program doing a sensible thing. Squeezing capacity out of lines that already exist beats waiting a decade for new corridors. For the optimism to hold, the agreements have to execute, and commissions have to decide how the federal share shows up for customers instead of letting it vanish into a revenue requirement.
The money moved faster than the rules for crediting it. Will that get settled in rate cases, or in the next permitting bill? I think that's what a lot of us are still either wondering or working out.
