1) Engineering (what software folks call hardware), especially cross-functional engineering across several fields, such as chemical and electrical.
2) Software, typically including applications of ML.
3) R&D. Research And Development, folks. The thing investors don't want you to do.
Those types of companies will almost never come out of venture capital. They start with a small founder investment (what VCs tell us is called 'bootstrapping'), significantly supported by federal government grants and contracts (the best source of funding for true R&D), and then product development and commercialization.
The ZIRP-era of pure software, VC-driven type of technology entrepreneurship tries to skip all the way to the end, scale beyond all bounds, and exit big and fast. No wonder we're experiencing groupthink, and why the Bay Area is hollowing out: in this form of technology entrepreneurship, there is no deep economic basis for growth.
In short: technology entrepreneurs should take on technology risk. That means raising capital late, not early, because investors will always be skittish about R&D and longer timeframes. And it means following your own vision first and foremost, rather than letting venture capitalists define how you see the world.
Investors are not where the action is. It is founders and early employees who built the likes of Sony, Qualcomm, General Atomics -- each of these, a real technology company.