AI infrastructure spending accelerates despite valuation repricing on rates

AI infrastructure is becoming a separate asset class. American Electric Power, one of the largest US utilities, is raising $2.6 billion in equity to fund AI-driven power demand. Fervo Energy, a geothermal developer, raised $1.89 billion in its IPOInitial Public Offering - a company's first public sale of stock. above range, betting that AI data-center clusters will drive electrification tailwinds for years. Venture Global is surging on new LNG supply deals and Louisiana export expansion. The market is pricing in that AI will require unprecedented capital investment in power generation, transmission and cooling infrastructure, not just chips.

The semiconductor capex impulse remains intact despite Tuesday's valuation shock. Broadcom reported strong bookings. NVIDIA is guiding above expectations despite China export headwinds. TSMC is expanding capacity in Taiwan and the US. AMD is investing in memory partnerships. The message is clear: AI capex cycles are real and multi-year, and suppliers are confident enough to commit capital even as macro uncertainty rises.

However, higher-for-longer rates are starting to weigh on project economics. Data-center operators are facing higher financing costs for new cluster buildouts. Power utilities are seeing debt issuance spreads widen. Smaller infrastructure players may struggle to access capital if yield curves invert further. Investors are rotating selectively, favoring large, highly-leveraged capex stories (Broadcom, Nvidia) while avoiding smaller infrastructure players with weak balance sheets.

The bull case is that AI capex is non-discretionary; companies must build or risk falling behind. The bear case is that valuations have run ahead of actual revenue realization, and rising rates will slow project starts in late 2026 and 2027. If energy prices normalize and Fed cuts materialize, infrastructure trades could re-rate higher. But near-term volatility is likely as investors price in both capex upside and rate downside.