your verdict flip?

``` Resolved: Dylan Patel's July 2026 claim is correct — by mid-2028, solar plus battery will be cheaper than gas for powering new US hyperscale data centers behind the meter. The test is the COST CLAIM, apples to apples: both systems must do the same job — run a new 100+ MW data center campus's 24/7 load. Within that, each side gets its least-cost real-world configuration. - "Cheaper" = lower all-in levelized cost ($/MWh) at prices a developer could actually contract in the US — including tax credits, subsidies, and tariffs actually in effect: the post-OBBBA credit phase-out as it applies to a project breaking ground in 2027-2028, IRA 45X domestic battery manufacturing credits, and current tariffs on imported Chinese cells, modules, and batteries. Real procurement prices, not policy-free abstractions. - "Solar plus battery" = utility-scale PV plus storage, architected however a competent developer would: size the battery as you see fit, keep gas or diesel backup or a grid connection for the tail if that is the least-cost way to reach data-center-grade service — but price every component and state plainly what fraction of annual MWh is non-solar. A 90%-solar hybrid is a legitimate answer; a hidden backstop is not. - "Gas" = the least-cost gas configuration actually deployable behind the meter on the same timeline: combined-cycle turbines (with real order-book lead times and scarcity premiums priced in), aeroderivative or industrial turbines, or fleets of gas-converted reciprocating engines. Do not strawman gas as backlogged CCGT if reciprocating fleets are deployable now at volume, and do not use legacy grid fleet averages. - Service requirement: the availability a hyperscale operator would actually accept for the campus. State the availability your design achieves and defend it. If your number differs from another advocate's, expect to be cross-examined on whose assumption reflects real 2026-2028 procurement. - Region: answer for the US Southwest (NV/AZ/West TX) and Mid-Atlantic (Virginia) separately if they diverge. Deliverables, in order: (1) VERDICT — YES (Patel is right) or NO, with your central $/MWh estimate for both least-cost configurations in mid-2028, per region, and the availability each design achieves. (2) CROSSOVER — if NO: your median crossover year with an 80% confidence interval, or "not before 2040" with the specific mechanism that keeps gas ahead. If YES: the year it happened or happens. (3) CRUXES — the 2-3 assumptions your verdict most depends on, each with the numeric threshold that would flip it. Address at minimum: - battery pack and module price trajectories, and whether US tariffs on Chinese cells break the curve for US deployment; - subsidy decomposition — how much of your cost gap is policy? Does your verdict flip with zero subsidies and zero tariffs? Does OBBBA phase-out timing change the answer for a 2027-2028 groundbreaking? - reliability sensitivity — at what required availability, if any, does your verdict flip? Patel's own caveat ("enough battery to get through the night" versus "three days of rain") lives here; - solar/battery supply-chain capacity against 10-30 GW/yr of US data center demand — price scarcity premiums the same way you price gas turbine backlog premiums. (4) FLIP CONDITION — the single piece of evidence that, if produced in this debate, would change your answer. Numbers over narrative: every cost or price-curve claim needs a source and a date. Prefer 2024-2026 actuals — signed solar+storage PPA and behind-the-meter deal prices, Lazard LCOE+, NREL ATB, EIA data, announced turbine and reciprocating-engine lead times, battery pack price surveys — over extrapolated curves. ```