ERCOT is still the most attractive power market in North America for new AI capacity, but the pricing is tightening. The 3-cent PPAs that defined the 2022-2024 mining buildout are gone for new contracts; large data center buyers are now signing in the 3.5-5.0 cent range, with structure mattering more than the headline number. Here's where we see pricing today and where it's heading.
Current rate bands
Across roughly 40 contracts we've reviewed or had visibility into over the past six months, ERCOT pricing for new data center load (50+ MW, 5-10 year terms) clusters in distinct bands by structure. The variance within each structure is small; the variance between structures is large.
- Fixed-price PPA, no curtailment: 4.5–5.5¢/kWh all-in, depending on substation and load factor
- Fixed-price PPA, curtailable with notice: 3.8–4.5¢/kWh; meaningfully cheaper but adds operational complexity
- Block-and-index hybrid: 3.5–4.2¢/kWh effective; better economics, more market exposure
- Behind-the-meter generation with grid backup: 2.8–3.8¢/kWh; capital-intensive, the cheapest path on a fully-burdened basis
- Bilateral with gen-tie: highly variable; depends on specific generator economics
Why ERCOT is still attractive
Three structural factors keep ERCOT cheaper than MISO and PJM despite the recent tightening. The most important is the deregulated market structure: data center loads can contract directly with generators or retail providers without going through a utility tariff. That eliminates the 1-2 cent regulated retail margin baked into most MISO and PJM transactions.
Second, the gas-heavy generation mix in West and South Texas provides a structural price floor that's lower than the coal-and-nuclear mix in PJM. Third, the queue dynamics — while painful — are still moving faster than PJM's effectively-closed queue.
Regional comparison
Same load profile (75-100MW, 80%+ load factor, 7-year term), here's what we're seeing across the major North American markets in Q2 2026.
- ERCOT West (Permian Basin, Big Country): 3.5–4.5¢/kWh, queue 30-42 months
- ERCOT South (Coastal Bend, Rio Grande Valley): 4.0–4.8¢/kWh, queue 36-48 months
- ERCOT Central (Travis, Williamson, Bell counties): 4.5–5.5¢/kWh, queue 42-54 months, transmission constrained
- MISO North (Iowa, Minnesota): 4.5–5.5¢/kWh, queue 42-60 months
- MISO South (Louisiana, Mississippi): 4.0–5.0¢/kWh, queue similar, gas-heavy
- PJM (Virginia, Pennsylvania, Ohio): 6.5–8.5¢/kWh, queue 60+ months and effectively closed
- WECC (Arizona, New Mexico): 4.5–5.8¢/kWh, queue variable, water constraints
Where curtailment is being priced
The most interesting structural shift over the past 18 months has been the willingness of AI operators to accept curtailable load in exchange for meaningful rate concessions. This used to be table stakes for mining and unthinkable for AI; that's flipped. Training workloads with elastic schedules can absorb 5-15% curtailment per year without meaningfully impacting model output, and the rate concessions for accepting that flexibility have widened.
We see 60-80 basis-point spreads between non-curtailable and curtailable contracts on otherwise comparable structures. For a 100MW load at 80% load factor, that's $4.2M-$5.6M per year of saved cost. Several major neoclouds are explicitly engineering their training schedules around curtailment economics.
Where prices are heading
Our base case for ERCOT through 2026: continued tightening at the lower end, stability at the upper end. The 3.5¢ floor is structural; we don't see it moving meaningfully lower without major generation buildout or a slowdown in load growth. The 5.5¢ ceiling is also structural at the upper end; load that prices above it migrates to MISO or WECC where the queue is the binding constraint rather than the price.
The wildcard is generation buildout. Texas added significant gas and wind capacity in 2024-2025; if that pace continues, prices may stabilize or modestly soften through H2 2026. If reliability concerns drive Texas to restrict new gas additions, the upward pressure resumes.
What this means for site selection
We're advising buyers to look at three structural decisions. First, accept curtailable load if your workload allows it — the math is compelling. Second, evaluate behind-the-meter generation seriously for 100+ MW loads with 5+ year horizons; the capital intensity is real but the rate certainty is valuable. Third, treat West Texas as the default destination and only deviate for specific reasons (latency, customer proximity, water).
