A nonbinding letter of intent in 2018? That is a limited downside unlimited upside kind of risk you take to fund extreme moon shots. Boom Supersonic did the same approach with United airlines, to fund a still undeveloped supersonic Airplane. A 1M investment like this is a hedge that has capped downside, restricted to Altman's personal money, and an infinite upside that gives an incentive to the founders to succeed. At the time the value of the investment is paper zero for Altman, but gives the founders, who were probably the most promising he has seen in the space, large help with fundraising, hiring, and credibility with future sales. OpenAI does not have to buy the chips unless they meet the criteria, In 2018 OpenAI was one of many startups plucking at a problem that seemed insurmountable. If you are trying to push into a problem and can't afford to do what Apple did to build both the hardware and software to solve it, you focus on the one and try to help the other however you can.
In WeWork, the CEO took a trademark that costs a couple of thousand dollars, and charged himself millions for it from company money - that's personal enrichment. Incubating a thing considered highly risky and expensive 5 years is not personal enrichment, its personal risk that helps more the likelihood of it being less impossible instead of a chance at enrichment from two companies you don't even know would be viable (OpenAI) 5 years from then.