This statement assumes the following axioms:

- The engineers that go to work for such startups are capable of devising/deploying solutions towards climate change: Some engineers & researchers may be able to do this, but most are focused in CompSci/SoftEng, with no experience in the field that you want them to be in.

- No roadblocks exist throughout the deployment of said solutions (NIMBYs, politicians, perfectionist "environmental activists"/virtue signalers, bureaucracy, supply chain issues)

Despite the "Sounds good"ness of your statement (which I partially agree with), the practical reality is that the chances of succeeding in something like crypto is still way higher than in building infrastructure, as there are near-0 NIMBYs in crypto that will protest in the next town hall meeting, nor will there be weather issues during the deployment/upgrade of a network/smart contract.

"Software will eat the world" still rings true today, only because the deployment of software is still orders of magnitude easier than deploying something into the real world. In order to make something like infrastructure deployment that much more enticing, the (chance_of_success*windfall - chance_of_failure*investment) must be greater than investing in software, which has significantly lower development costs, does not get bottlenecked by supply chain issues, & can be deployed globally within minutes.