San Jose races to become Bay Area’s data center capital — PG&E customers could pay the price

San Jose has emerged as the Bay Area hub for AI data-center development with PG&E reporting requests to serve 11 projects totaling 1,630 MW (enough for ~1.2 million homes), while the utility’s Bay Area pipeline totals about 3,500 MW. PG&E estimates infrastructure costs of $500M–$1.6B per 1,000 MW (implying $1.75B–$5.6B for 3,500 MW) and projects up to ~9.6–10 GW of future data-center demand that it says could lower rates, though consumer advocates and state watchdogs warn customers may ultimately bear the costs; San Jose expects sizable property-tax/utility-user-tax revenue (a 99-MW center estimated to generate $3.5M–$6.4M annually). The story highlights material local risks — higher residential bills, heavy water use, diesel backup pollution, habitat and farmland loss — and regulatory scrutiny that could affect utilities, data-center operators and local real estate planning.

Analysis

Market structure: The Bay Area data-center surge concentrates economic rents with colo operators (EQIX) and incumbent utilities (PG&E/PCG) that secure long-term power contracts and rate-base treatment; PG&E’s 3.5GW local pipeline implies $1.75–$5.6B of incremental capex, boosting regulated earnings if approved. Hyperscalers (MSFT, AMZN) face higher upfront site, permitting and community costs that can compress incremental ROI on bespoke campuses, while municipal tax revenues and real-estate owners near sites capture outsized upside.

Risk assessment: Key tail risks are regulatory moratoria or punitive rate rulings (weeks–months) and a demand shock from AI hardware efficiency or slowdown (12–36 months) that could strand utility investments and depress colo utilization. Hidden dependencies include water availability, diesel-generator pollution/liability and rate-case outcomes that allow PG&E to pass costs through; catalysts are Little Hoover Commission findings, senator inquiries, and city council approvals over the next 30–90 days.

Trade implications: Favor selective exposure to regulated utility cashflow and incumbent colo capacity but hedge hyperscaler execution risk. Tactical trades: small equities exposure to EQIX via call spreads, conservative exposure to PCG via 2–5yr bonds/credit, and put spreads on MSFT/AMZN to hedge regulatory or demand pullback in 6–12 months. Rotate into energy services and industrials supplying gensets/cooling on weakness.

Contrarian angles: Consensus overstates community pushback as fatal — permitting delays are more likely than outright bans, which creates scarcity that benefits existing colo landlords (EQIX) and raises TTM FCF. Conversely, if AI model efficiency improves 20–50% in 12–24 months, power intensity per compute could fall materially, leaving utilities with stranded assets; price positions should size for both paths (small, hedged exposures).

Key Decisions for Investors

• Establish a 2–3% long position in EQIX using a 6–9 month call spread (buy ATM, sell +15% strike) to express demand-driven upside while capping premium; add on any headline-driven pullback >10%.
• Allocate 1–2% to PG&E (PCG) exposure via 2–5yr investment-grade muni-style bonds or senior notes if yields >150bp over Treasuries, reflecting predictable regulated cashflows from rate-base capex; sell if regulators deny cost recovery or spread tightens below 100bp.
• Open a 1–1.5% short-risk hedge on hyperscalers: buy 9–12 month put spreads on MSFT and AMZN (5–7% OTM buy / 12–15% OTM sell) sized to offset 50% of EQIX exposure, protecting against permitting/regulatory or demand shocks.
• Initiate a 1% tactical long in industrials/energy-services (gensets, cooling suppliers) via ETFS or individual names with >30% revenue exposure to data centers; take profits if these names rerate +25% or if municipal approvals accelerate meaningfully within 60 days.
• Set explicit triggers: add to colo/utility longs on any EQIX drawdown >15% or PCG credit spread widening >50bp; unwind or materially reduce long exposure if Little Hoover Commission or state regulators impose a moratorium or deny >50% of projected cost recovery.