AEP Raised Guidance Into a 5.42% Long End: The AI Buildout Is a $78 Billion Rate-Base Bet, and the 30-Year Decides Who Pays

AEP raised full-year 2026 operating earnings guidance to $6.25–$6.55 a share on July 30, 2026, from $6.15–$6.45 — $0.10 at both ends, about 1.6% of the mid-point. Second-quarter operating earnings of $742 million, or $1.36 a share, took year-to-date operating EPS to $3.01, and management kept the 7%–9% annual growth target and a compound rate above 9% through 2030. Moves the guidance mid-point 1.6% while the contracted load grew 6 gigawatts in a single quarter.

The raise itself is the least interesting line in the release. The interesting part sits on the balance sheet: AEP holds roughly $5 billion of Department of Energy loans and nearly $400 million of grants that cut the interest cost of the buildout, and it pre-funded the equity with a $3 billion forward sale executed in May 2026 that settles by May 2028. Locks in the funding for rate-base growth before the load arrives, which is the one thing the last utility capital cycle did not manage.

AEP's load story is now a conversion story, and the conversion rate is the number that matters. Of the roughly 195 GW sitting in the utility's active interconnection queue, 69 GW is contracted through 2030 — 45 GW in ERCOT, 18 GW in PJM and 6 GW in SPP — so about a third of the pipeline is under signed agreement, and nearly $2 billion of customer cash and collateral stands behind the 45 GW submitted into ERCOT's Batch Zero process. Signs up about 35% of the 195 gigawatts in its queue and builds $78 billion of rate base on the smaller number.

That conversion is now being audited. AEP Ohio's load forecast as of February 2026 reflected signed contracts for 17.8 GW of data-center demand running through 2035, and a March 2026 regulatory staff filing alleged double-counting in the utility's expected growth. Turns a 69 gigawatt contract book into a contested forecast that only completed interconnections can settle.

AEP's large-load tariffs are the mechanism that decides who pays, and they hardened through the summer. The company has approval in five of the eight states where it has filed, with Michigan, Oklahoma and SWEPCO Texas still pending; the filings require customers to take or pay for 80% to 90% of contracted demand under contracts running as long as 20 years, with termination-fee collateral. In Ohio the threshold is 25 MW and the term is 12 years; in Virginia it is 100 MW individual or 150 MW aggregated for 14 years; in Kentucky, 150 MW buys 20 years.

Here is what the framing gets wrong. These terms are excellent for the utility and harsh for the data center: a 12-year obligation to pay for at least 80% of contracted megawatts, against hardware that turns over every three to four years, is a fixed cost sitting on a depreciating asset. Moves volume risk off the ratepayer and onto the hyperscaler's balance sheet, and that is precisely the trade a 5.42% long end makes necessary.

This is where the 5.42% enters the story. AEP's 2026–2030 financing plan funds a $77.9 billion capital and joint-venture equity contribution with $47.1 billion of cash from operations, $45.5 billion of net debt issuance and $9.7 billion of equity — dividend reinvestment, an at-the-market program and settlement of the March 2025 and May 2026 equity forwards. Covers less than half the capital plan with internal cash, leaving debt to do the rest at the long end.

The cost of that debt is already visible. Interest expense rose 19.2% in the second quarter to $585 million while revenue grew 7.0% to $5.45 billion; total debt was 14.8% higher year over year at $53.5 billion, and diluted share count rose 2.7% to 550.6 million as the forward sales settled. Moody's put FFO/debt at 14.0% at June 30 against a 14% to 15% target, with S&P at 14.6%. Leaves no credit headroom to absorb a rate shock without more equity.

The hedge is the federal government, and it is large. AEP holds roughly $5 billion of Department of Energy loans — including up to $3.3 billion supporting nearly 100 AEP Texas transmission projects, worth $685 million in customer interest savings — plus nearly $400 million of grants. Subsidizes the cost of the money that builds the grid, which is the quiet reason the ratepayer fight has not become a ratepayer fight.

Strip the story down and only one subtraction matters. AEP's most recently settled returns — 9.84% in Ohio, 9.75% in West Virginia and Kentucky, 9.65% in Arkansas — against a 30-year Treasury at 5.42% leave an authorized equity spread of roughly 4.2 to 4.4 points. A year earlier, with the 30-year at 4.95%, the same authorizations implied 4.7 to 4.8 points. Compresses the authorized equity spread by about 30 basis points in twelve months, and AEP's three pending cases — 10.75% on $95 million for SWEPCO Texas, 10.5% on $299 million for PSO Oklahoma, 10.5% on $61 million for APCo Virginia — are attempts to win it back.

The market has already voted on the discount rate rather than on the earnings. Utilities fell 13.0% in the third quarter of 2026, the worst of any S&P 500 sector, while the 30-year ended that quarter at 5.63% against 4.95% at the end of 2025. AEP is the most rate-exposed version of the trade: the highest trailing multiple in its peer group, near the highest leverage, and a dividend yielding 3.05% against a 5.42% risk-free bond.

The buildout has an order book upstream, and it is concentrated. AEP has secured about 13 GW of gas-fired turbine capacity for deployment through 2031, including 3 GW added in the second quarter, and holds options on up to 10 GW more through 2035 under framework agreements with GE Vernova and Mitsubishi. Puts 13 gigawatts of generation demand behind a regulated balance sheet, with $24 billion of generation capital behind it and GE Vernova's gas backlog at 116 GW in July.

Grid equipment is the constraint that rations AEP's rate base, and it shows up in order books rather than headlines. But the margin line is where the read-through gets interesting: Eaton's Electrical Americas operating margin fell to 27.5% from 29.5% while backlog grew a third, so the equipment suppliers are converting AI demand into volume faster than they are converting it into profit.

Watch one line in the release: new contracted gigawatts. Six were added in the second quarter. If the third quarter adds less, the ~11% rate-base CAGR becomes a function of rate-case outcomes rather than of demand, and the 5.42% long end stops being an environment and starts being a constraint. Turns the next print into a test of the demand assumption, not the earnings assumption.

That third risk is the one no rate case can fix. If hyperscaler capital intensity slows, the take-or-pay contracts keep paying while the load behind them does not, and a utility that has spent $33 billion on transmission is left holding a 12-year obligation from a counterparty whose hardware turns over in three to four years. Turns the utility's largest rate-base story into a credit exposure once the steel is in the ground.

The demand behind those contracts is not a forecast; it is a spending spree already under way, and it is funded in the same capital markets that AEP needs. Amazon raised 2026 capital spending guidance to $220 billion while its trailing free cash flow turned negative, and Meta's second-quarter capital expenditure of $31.1 billion was funded out of $31.9 billion of operating cash flow. Funds AEP's rate base and its own obligations from the same pools of money, which is the concentration risk nobody prices into a 12-year take-or-pay.