Conviction · May 2026
Using AI as a chatbot is one experience. Living inside Claude, Codex, and the agentic loops they enable is another. New models land every few months. New tools ship every week. Every release compounds. You feel the acceleration.
This is hard to transmit through prose. The voices below come closest. They run capital and ship product at scale, and they are putting both behind the conviction.
“I went home one Friday, actually fairly depressed. You could just see how this was going to have such a dramatic impact on society.”
Ken Griffin Founder & CEO, Citadel
At Davos in January he called AI 'all garbage' and dismissed the jobs panic as hype to justify datacenter spending. By May, finance work that took MS/PhD teams weeks was being done by AI agents in hours.
Stanford Leadership Forum · May 2026
“We're not investing approximately $200 billion in capex in 2026 on a hunch.”
Andy Jassy CEO, Amazon
They see demand from two sides at once: as the cloud running frontier models through Bedrock, and as a major equity holder in both Anthropic (~17%, ~$33B) and OpenAI ($50B committed).
2025 shareholder letter · 2026-04-09
“The technology is so powerful, it's worth the trillion-dollar investment.”
Jamie Dimon Chairman & CEO, JPMorgan Chase
Said AI 'will change almost everything' and that society should prepare for displacement now.
Fortune · 2026
“We're in a massive 10-year AI buildout, and only 2 years into it.”
Lisa Su Chair & CEO, AMD
Time-horizon from someone underwriting the buildout on the merchant-silicon side.
AMD Q1 earnings · May 2026
“That's why I think it's probably a hundred to two hundred gigawatts a year of terrestrial chips, and probably on the order of a terawatt of chips in space. Just because of power constraints on the ground.”
Elon Musk CEO, Tesla / SpaceX / xAI
Terafab plans 100-200 GW of terrestrial chips per year and 1 TW of chips in orbit. The entire US grid runs about 1.2 TW. Producing intelligence on demand at that scale forces the buildout off the planet, which is why 80% of Terafab's output is targeted for space. Even then, current global fab capacity is only about 2% of what Tesla and SpaceX say they will need.
Terafab announcement · 2026-03-21
“The models, they just want to learn.”
Ilya Sutskever Co-founder, OpenAI · Founder, Safe Superintelligence
Ilya said this to Dario before OpenAI existed. It's a description of how the architecture behaves, not a marketing line. Build the model bigger and it gets smarter on its own. Nobody has to teach it new tricks. Every scale-up for the last decade has confirmed it.
Relayed by Dario Amodei on Dwarkesh Patel · 2024
“We're standing on square 40 out of 64.”
Dario Amodei CEO, Anthropic
The chessboard parable: an inventor asks a king for one grain of rice on square one, doubling on every square through 64. Square 32 totals about 4 billion grains, still countable. Square 64 totals 18 quintillion, more rice than has ever been grown. Dario's point: AI capability is already in the second half of the board. The next doublings happen on top of what models can already do.
TMT Breakout · Mar 2026
“Our AI data center portfolio can double from here.”
Stephen Schwarzman Chairman & CEO, Blackstone
Said on Blackstone's Q1 2026 earnings call after declaring Blackstone the world's largest investor in AI infrastructure. The current book is already over $150B in data centers with another $160B in the pipeline. Doubling translates to roughly $300B of additional AI buildout from a single private allocator.
Blackstone Q1 2026 earnings call · 2026-04-23