Sam Altman reveals investment lessons behind OpenAI's largest strategic bets and capital partnerships.

Sam Altman says the best investment decisions rarely look smart in the moment, and he's crediting two of Silicon Valley's most polarizing figures for teaching him that lesson.

During an interview for the "Invest Like The Best" podcast, the CEO of OpenAI stated that Peter Thiel and Paul Graham, in different ways, showed him that companies and investments with the most potential usually never seem popular when they start.

While following consensus can yield good results, Altman argues that achieving spectacular outcomes requires betting on things that others ignore or criticize.

Drawing on his experience leading Y Combinator, founded by Paul Graham in 2005, Altman emphasized that good startup investing is more like learning to fly than absorbing lectures.

He tied this philosophy directly to OpenAI's history. After GPT-4's release, Altman knew they were "on this exponential of model improvement" and began reaching out to cloud service providers, chip manufacturers, and energy firms to book future capacity. The reaction was almost uniformly cynical: there's simply no way an industry can move at that speed.

Altman compared the experience to raising money for an early-stage startup, where most doors close but only one or two need to open. This mindset, he suggested, connects his investing philosophy to OpenAI's compute strategy: betting against consensus when the underlying signal is strong enough to justify it.