A traditional, technology-first PhD lab focuses on answering unknown scientific questions, not scaling market solutions. Because of this, trying to force a commercial application out of purely accidental, curiosity-driven breakthroughs is highly risky and often happens entirely by chance.
If your explicit, long-term goal is to found a hard-tech startup, you should lean toward a problem-first lab. Starting with a massive real-world problem (like industrial decarbonization) ensures that whatever scientific breakthrough you work toward already has a built-in market demand. While universities have great infrastructure to help spin out research, it is much easier to transition from lab to market when your research was designed to solve a specific problem from day one.