bottleneckScore 85/100Research
Sam Altman: OpenAI's biggest bottlenecks are 'transistors and then electrons in that order' — chips and energy will constrain AI scaling through the decade
Sam Altman· OpenAI· AI· 2026-07-25· about OpenAI
“What do you think your biggest bottleneck is? Uh if you just say like we want this to continue scaling unabated. What is the biggest bottleneck that you can see? >> Transistors and then electrons in that order.”
Why it matters
OpenAI's CEO has identified semiconductor supply and power generation as the primary constraints limiting AI model scaling — more critical than talent, data, or algorithms. This directly signals where capital and supply chain stress will concentrate over the next 3-5 years.
Investment implication
Semiconductor manufacturers (NVIDIA, TSMC, Samsung), advanced packaging providers (CoWoS, chiplet integrators), and power infrastructure companies (utilities, data center operators, renewable energy) will see sustained pricing power and capex demand. Conversely, AI software companies and model developers face input cost pressures.