EM: three lessons from the US west coast What the Emerging Markets Team has learned from conversations in Silicon Valley. September 2026 Art

AI demand will expand the emerging-markets opportunity beyond chipmakers because power, engineering capability and proprietary company knowledge are becoming scarce assets.

• Agentic coding tasks use about 1,000 times more tokens than conventional human-led code reasoning, helping explain why AI infrastructure providers are struggling to keep up.
• Power is becoming a binding constraint: models, chips, memory, construction and efficiency must improve together, while the team is researching opportunities from Brazil's Axia Energia to Indian power and infrastructure.
• Eskom, recently associated with electricity blackouts, is now discussing selling excess power to US hyperscalers, illustrating the potentially sharp change in emerging-markets electricity demand.
• AI-enabled engineering could improve physical production: CATL can screen thousands of battery-material combinations with automated laboratories, while manufacturers may reduce waste, shorten maintenance cycles and lower costs.
• Grid bottlenecks create investable needs because the IEA estimates that about 20 percent of planned datacentre projects could be delayed by grid risks, and transformer and cable waiting times have doubled in three years.
• Proprietary workflows can become 'token capital': Microsoft CEO Satya Nadella describes this as a moat, and Nubank's David Vélez is using AI learning loops to improve underwriting and customer service.

Takeaway: Investors should broaden AI research from semiconductors to power utilities, grid equipment, industrial engineering and data-rich businesses such as Nubank, while recognising that the article presents emerging research themes rather than investment recommendations and was written in August 2026.