US Nuclear Power’s Technological ‘Rebirth’ Will Take Some Time

Two major data center projects in Santa Clara, Nvidia's hometown, may sit empty for years because the local utility cannot yet supply enough electricity. The article highlights a supply-side infrastructure bottleneck that could delay AI-related capacity expansion and pressure near-term data center development timelines. The issue is negative for local real estate and infrastructure availability, but the broader market impact is likely limited.

Analysis

This is less a one-off permitting hiccup than a capital-allocation signal: the bottleneck is shifting from compute availability to power deliverability. That changes the economics of AI infrastructure from “who can raise capital and secure GPUs” to “who can secure interconnection, substations, and transmission rights,” which favors utilities, grid equipment vendors, and developers with legacy land banks near available capacity. The near-term loser is the data-center development pipeline that has already been underwritten on aggressive absorption assumptions; delays of 12-36 months can force lease-up slippage, covenant stress, and a repricing of pre-let deals.

Second-order effects likely show up in pricing, not just timing. Scarcity of powered sites should lift rents and make incremental MW more valuable, which ultimately transfers bargaining power to utility-scale power providers and firms that can add behind-the-meter generation, batteries, or gas peakers. It also creates a hidden capex tax for hyperscalers: every month of delay on a 100-300MW campus can burn option value and push customers toward alternative geographies with faster energization, even if those are suboptimal for latency or tax reasons.

The contrarian view is that the market may be too focused on the downside to AI infrastructure and underappreciating the scarcity premium in power-enabled real estate. If local permitting, transmission buildouts, or utility upgrades are accelerated, the perceived stranded-asset risk can reverse quickly; however, that is a months-to-years process, not a days-to-weeks catalyst. In the meantime, the cleaner trade is to own the enablers of the bottleneck rather than the bottlenecked developers.

Key Decisions for Investors

• Long NVT, ETN, and HUBB on a 3-12 month horizon: grid equipment demand should be the cleanest beneficiary of prolonged interconnection scarcity, with multiple expansion likely as orders extend beyond the current cycle.
• Long DUK / NEE versus short select data-center REIT exposure on a 6-18 month relative-value basis: utilities with regulated rate bases and transmission upside should outperform developers exposed to delayed power delivery.
• Pair trade: long power-infrastructure basket (PWR, ETN) / short highly levered data-center development names or REIT proxies if available; target a 15-20% spread over 2-4 quarters as schedules slip.
• Consider calls on utility-scale gas generation or turbine suppliers if local grid delays persist into the next earnings season; backup generation becomes a critical bridge asset when interconnection queues widen.
• Avoid chasing the headline weakness in AI capex broadly; use any selloff in hyperscaler-adjacent names as a staggered entry only if they have diversified siting and captive power arrangements.