The Grid That Cannot Keep Up

Key Takeaways

• PJM capacity price jumped to $329/MW‑day, a 1,037% rise
• Interconnection queues now span 3‑7 years, transformer lead times 160 weeks
• 30% of data‑center power expected from on‑site generation by 2026
• Major utilities admit they cannot meet demand through 2026
• On‑site generation offers first‑mover advantage over grid‑dependent rivals

Pulse Analysis

The dramatic surge in PJM’s capacity market price is more than a headline; it signals a structural mismatch between soaring AI‑driven compute demand and a legacy electricity grid that was designed for slower growth. Historically, capacity prices hovered under $30 per megawatt‑day, providing predictable cost inputs for data‑center planners. The 2026‑27 price spike to $329 reflects not only heightened demand but also the scarcity premium utilities charge when they must stretch existing transmission assets to accommodate gigawatts of new load. This price volatility forces investors to reconsider the economics of building new facilities in traditional grid‑constrained zones.

Compounding the price shock are the protracted interconnection timelines that now dominate project schedules. Securing a queue position, completing studies, and waiting for utility upgrades can consume three to seven years, while critical hardware such as substation transformers faces lead times of up to 160 weeks. These delays translate directly into opportunity costs for AI firms that need compute capacity on a near‑term basis. The equipment market, projected to grow from $20 billion in 2026 to $65 billion by 2030, cannot accelerate fast enough to offset the grid’s inertia, making the interconnection bottleneck a long‑term strategic concern.

Faced with these constraints, the industry’s most advanced players are bypassing the grid altogether. On‑site generation—ranging from natural‑gas turbines and small modular reactors to co‑located renewable assets—offers a path to immediate power availability and price certainty. Companies like Constellation Energy and Exxon Mobil are already signing long‑term offtake agreements, while modular data‑center concepts such as Crusoe Energy’s West Texas deployment illustrate how compute can be paired directly with dedicated generation. As on‑site power is projected to supply 30% of data‑center demand by 2026 and potentially half by the decade’s end, firms that master this energy‑first model will secure a durable competitive moat, reshaping the landscape of AI infrastructure investment.