The S&P 500 and Nasdaq snapped three-week winning streaks this week, down roughly 1.43% and 2.05% respectively, with pressure coming from rising long-term Treasury yields and renewed inflation concerns. The Dow Jones Industrial Average slipped 0.85%. Much of the volatility centered on AI and infrastructure plays: the Nasdaq's tech-heavy composition bore the brunt of the decline, while infrastructure names tied to data center buildout—including GE Vernova and Eaton—fell 10% and 7.2% respectively for the week.
Political and regulatory headwinds added to the weakness in AI infrastructure. Pennsylvania Governor Josh Shapiro issued an executive order on Tuesday imposing tough standards on data center developments in the state, marking a significant regulatory barrier to buildout. Broader concerns about state-level restrictions on data center siting have begun to weigh on investors' conviction in continued AI infrastructure expansion. The Treasury Department's Wednesday announcement that it would more than double the size of its buybacks of longer-dated government debt provided brief relief—stocks and yields fell initially—but the move proved short-lived as yields climbed again Thursday and Friday.
Broadcom shares fell 4% on Wednesday after rival Marvell announced an expansive partnership with Google, diversifying Google's TPU supplier base away from Broadcom's historical co-design relationship. Elsewhere, chip designer Cadence Design Systems was added to the Bullpen following CEO appearances on major platforms; the stock had fallen roughly 23% from early June highs on concerns that AI could disrupt traditional chip-design software, though Cadence sees agentic AI as a driver of greater design-tool adoption.
For practitioners: data center regulation is now a material factor in project planning. State-level restrictions on buildout, combined with rising funding costs for infrastructure projects, are reshaping the timeline and location decisions for new facilities. Watch for company guidance updates on capex commitments and timeline push-outs, particularly from cloud providers and specialized data center operators. The memory cycle may prove more durable this time due to AI-driven customer commitments and long-term supply agreements, providing some offset to infrastructure headwinds.