Ribbit Capital's Power Letter: AI's bottleneck is electricity

In its July 2026 Ribbit Capital Power Letter, the firm argues that the constraint on the AI buildout is no longer money but power. It says the United States alone will need to add more than 100GW of new power capacity by 2030, while data centres face long lead times for transformers, switchgear and other critical equipment.

Ribbit describes an AI infrastructure market where data-centre campuses are becoming grid-scale industrial projects. The letter cites a 2.6TW US electricity interconnection queue, average waits of more than five years and 20–60-month lead times for key equipment. It estimates that building an AI data centre can require roughly $50B per gigawatt, creating a large opportunity for companies that can shorten construction schedules, improve power density or make each watt more productive.

Ribbit applies a four-part framework to both electrons and intelligence. On the energy side, that means generating power, moving it through transmission and distribution, storing it across time, and pricing and delivering it to customers. On the intelligence side, it means creating models and chips, moving data through interconnects, holding context in memory, and bringing compute and tokens to market.

The firm says energy has moved from an early metaphor for its work in finance to a core investment focus. It compares the current energy market with the period when fintech began to challenge outdated financial infrastructure: both are capital-intensive, highly regulated networks where frustrated customers and technological change can create openings for new entrants.

Its nine investment beliefs cover vertically integrated compute factories; bottleneck-breaking suppliers across power, cooling, memory and networking; frontier generation; alternative approaches to intelligence; national-sovereignty infrastructure; businesses combining hardware and software; storage-led models; customer-facing energy propositions; and markets for buying and selling compute.

Ribbit is looking for founders who can turn a temporary constraint into a durable platform. The letter highlights speed, reliability, cost and positioning as potential advantages for compute operators, while emphasising customer ownership, flexible power capacity, abstraction and increasing returns to scale across the wider energy and compute systems.

Ribbit is looking for founders who can break constraints across the energy and compute stacks, particularly where hardware and software, electrons and tokens, meet. Its broader thesis is that the companies able to create more energy, or turn the same energy into more useful intelligence, could become foundational infrastructure businesses for the next decade.