Best Case
15%Steel River stays on schedule, demonstrates that solar plus batteries can meet a meaningful share of new data-center load, and triggers multiple copycat hyperscaler-backed projects in solar-rich U.S. regions.
Forecast dossier
Hyperscaler power procurement will shift from renewable credits to anchor-tenanted solar-storage megaprojects
Google and Cypress Creek Energy broke ground on the Steel River Energy Center in Arkansas, with Google securing the first two phases through a power purchase agreement. The project is planned for 2.5 gigawatts of solar and 2.9 gigawatt-hours of battery storage by 2029, with LG Energy Solution batteries, First Solar modules, U.S.-made steel, and domestic assembly. The durable change is that large AI and cloud buyers are starting to underwrite grid-scale generation, storage, and domestic supply chains as a single package rather than buying unbundled renewable attributes after the fact.
Verdict: Likely. Steel River is large enough to become a reference transaction for AI-era power procurement, but replication depends on interconnection, permitting, battery supply, and whether utilities let corporate offtake reduce rather than shift grid costs.
Date
Jul 15, 2026
Reliability
82
Harm potential
Medium
Scenario odds
Baseline
50%The first phases proceed, hyperscalers increasingly sign anchor PPAs for paired solar and storage, and the model becomes a common supplement to utility tariffs and gas-backed data-center power.
Adverse Case
25%Interconnection delays, equipment costs, tariffs, or local opposition slow completion, making hyperscalers diversify back toward gas, nuclear, and utility special-rate structures.
Wildcard
10%A major reliability event or policy change forces new data-center projects to include directly deliverable firm capacity, accelerating hybrid portfolios that combine solar, batteries, gas, and demand-response obligations.
Timeline projections
1-Year
More anchor PPAs
Developments: At least several hyperscaler-backed solar-storage deals are announced, with projects framed around AI load, community benefits, and grid reliability.
Risks: Developers may overstate timelines because interconnection queues and transformer procurement remain bottlenecks.
Outlook: The procurement model gains attention but remains concentrated in a few favorable markets.
2-Year
Utility negotiations intensify
Developments: Utilities and regulators begin asking whether corporate PPAs reduce system costs or shift transmission and balancing costs to other customers.
Risks: Cost-allocation disputes could slow approvals or force new tariff designs for large loads.
Outlook: Solar-storage offtake becomes linked to broader data-center tariff and grid-planning debates.
3-Year
First operational proof points
Developments: Initial phases of large projects begin showing whether co-located batteries materially improve the match between solar output and data-center demand profiles.
Risks: If battery duration is too short for evening peaks or weather variability, buyers may need more firming resources.
Outlook: The model is validated for partial load coverage, not full 24-hour clean power.
5-Year
Portfolio procurement standardizes
Developments: Hyperscalers bundle solar, storage, firm capacity, transmission upgrades, and local benefit funds into repeatable regional procurement templates.
Risks: Supply-chain shocks or policy reversals could make domestic-content commitments more expensive.
Outlook: The market shifts from one-off clean-energy claims to structured infrastructure procurement.
10-Year
Corporate buyers shape grid buildout
Developments: Large technology buyers influence where generation, storage, and transmission are built, especially near data-center clusters.
Risks: Public backlash grows if residential and small-business customers perceive that data centers receive privileged grid access.
Outlook: Corporate energy demand becomes a major planning variable for U.S. regional grids.
20-Year
AI campuses become power-system actors
Developments: Large compute campuses operate with contracted portfolios of generation, storage, demand response, and backup resources that resemble private utility systems.
Risks: Regulators may impose stricter reliability, emissions, and cost-sharing duties on large-load customers.
Outlook: The boundary between data-center developer, power buyer, and grid participant becomes increasingly blurred.
50-Year
Compute and electricity planning converge
Developments: Major compute infrastructure is planned alongside generation, storage, water, land, and transmission from the outset.
Risks: Long-term electricity demand could be lower if AI efficiency improves faster than workload growth, stranding some assets.
Outlook: The durable lesson is integration: large digital infrastructure cannot be planned separately from physical energy infrastructure.
Planning prompts to verify
- Track whether other hyperscalers announce co-located solar-storage PPAs above 1 gigawatt in the next 12 months.
- Monitor Arkansas interconnection milestones, construction progress, and any cost-allocation disputes involving regional grid upgrades.
- Compare future data-center energy deals for domestic-content clauses covering modules, batteries, steel, trackers, and transformers.