It's critical to be on the right side of a technological trend & to choose wisely. Typically when you pick it right, problems get solved with time automatically (cost, performance, size, etc) but if you pick it wrong you get stuck for longer or find yourself in a shrinking market

For instance, blockchain scalability ended up not being as fast as people hoped limiting transaction volume keeping costs high.
Meanwhile GPUs perf improved 26x in 6 years.

Bandwidth improved linearly over a 20 year period so it was worth betting on but not too optimistically.

These days cloud compute (CPUs) is not trending well. The processor manufacturers are adding more cores at incremental improvement per core. Long lead times and delayed product cycles (Intel is behind 12 mo).
Cloud vendors are being aggressive about payback periods.

AI research is trending well: billions is being poured into research by large institutions. That doesn’t strongly correlate with utility but it means that the breakthroughs can sometimes be used by startups. There is power in knowing something is possible vs the uncertainty.