Texas Data Centers Will Raise Your Power Bill — and Two-Thirds of the Hit Comes from Wholesale Prices, Not Grid Upgrades

38.9 GW of announced data-center load in Texas by 2030. Residential bills rise ~1.64¢/kWh, with 68% from wholesale price spikes and only 32% from transmission costs.

Imagine you live in an apartment building that splits the water bill evenly. A brewery moves into the penthouse. Your bill goes up — but not mainly because the building needs bigger pipes. It goes up because the brewery is buying so much water that the price per gallon rises for everyone at the meter. That is the core mechanism this paper identifies for data centers and the Texas power grid. The committed claim: 371 announced data-center projects in ERCOT, totaling 38.9 GW by 2030, would raise modeled residential electricity costs by 1.64 cents per kilowatt-hour — and 68% of that increase flows through higher wholesale electricity prices, not through regulated transmission charges for grid reinforcement. This matters because public debate and regulatory attention have focused overwhelmingly on transmission infrastructure costs, while the bigger channel — demand-driven wholesale price inflation — gets comparatively little scrutiny. The architecture is a linked chain of four models: a generation and transmission expansion planner (capacity expansion), a wholesale electricity market simulation (production cost model), a transmission reinforcement cost estimator, and a retail cost allocation module. The paper does not use a single black-box optimizer; it chains established power-systems modeling tools to trace how a demand shock propagates through investment decisions, market clearing prices, grid upgrades, and finally onto residential bills. This pipeline approach is standard in power-systems planning but rarely executed end-to-end with this specificity for a single demand category. The ladder matters here. The authors test three policy-relevant scenarios against a no-data-center baseline: a reference case (1.64¢/kWh increase), a gas-colocation scenario where new gas generation is sited near data centers (reduces the hit to 0.92¢/kWh), and a transmission cost-exclusion scenario where data centers are carved out of the shared transmission cost base (raises the hit to 1.87¢/kWh). The gas-colocation finding is the sharpest result — it shows that generation siting policy can cut the residential cost impact by 44%, a lever that is actionable and quantifiable. Integrity deserves a careful read. The entire analysis is simulation-based — no observed market data validates the projections, because the 38.9 GW hasn't materialized yet. The authors are modeling announced projects, not contracted or built ones, which is the right forward-looking question but introduces substantial uncertainty about how much load actually arrives. ERCOT is a useful natural experiment because it is an isolated grid with transparent wholesale pricing, but the results are specific to Texas market structure and cannot be directly generalized to PJM or CAISO without re-running the pipeline. The paper acknowledges these limitations. The milestone to watch is straightforward: ERCOT's actual interconnection queue conversion rate. If even half of the 38.9 GW materializes by 2030, the wholesale price channel the paper identifies becomes the dominant cost pathway for Texas households. The concrete next number is whether realized data-center load crosses 15-20 GW in ERCOT by 2028, at which point the modeled wholesale price effects should become visible in actual settlement data. That is the falsification window. The obvious experiment the authors did not run is sensitivity to renewable-plus-storage colocation instead of gas colocation. They show that siting gas plants near data centers cuts residential costs by 44%, but the politically and commercially relevant question in 2025 is whether solar-plus-battery colocation achieves a comparable reduction. My honest read: this is being saved for the next paper, not omitted because it failed. The modeling pipeline supports it; the scenario just wasn't included in this round.