Constraints

• DIGITIMES analyst Sabrina Yu reported at Tech Forum 2026 that AI data centers are increasingly adopting "behind-the-meter" on-site power to bypass persistent electrical grid bottlenecks. • The shift is driven by four critical energy challenges: rising GPU thermal design power, the need for new high-voltage direct current architectures, grid constraints, and strict carbon-emission mandates. • This trend matters as cloud service providers struggle to maintain the massive energy requirements of AI infrastructure while meeting sustainability goals.

• Crude oil futures surged above $85 per barrel Thursday, driven by ongoing supply constraints from geopolitical tensions and maintenance shutdowns at major U.S. refineries affecting production capacity. • Energy stocks in the S&P 500 gained 2.1% on the session, with exploration and production companies outperforming as rising commodity prices improve profit margins and cash flow forecasts. • Analysts noted that refinery capacity utilization fell to 87% this week, the lowest level since January, suggesting tighter margins ahead for integrated energy companies dependent on downstream operations.

• Broadcom announced supply chain bottlenecks across the tech sector, particularly at TSMC, due to skyrocketing demand for AI chips on March 24, 2026. • The constraints affect networking, custom silicon, and data-center hardware critical for AI infrastructure beyond major model developers. • This highlights the AI hardware race creating opportunities for startups to innovate in chip manufacturing tools amid industry-wide shortages.