I think it's somewhere in between tbh. Not a full bubble but definitely frothy.
Here's my take: the infrastructure spend IS real. Companies are genuinely investing billions in AI compute because they have to or they'll get left behind. But the timing of when that spending translates to actual revenue/profit is super uncertain.
Like, Meta spent $40B+ on AI infra last year. That's real revenue for NVIDIA. But is Meta gonna make that back in 2-3 years from AI products? Unclear.
The dot-com comparison is kinda fair but also different:
- Back then: internet was new, unclear use cases
- Now: AI has proven use cases (Claude/GPT are legitimately useful), question is just ROI timeline
What worries me more is the concentration risk. If you're NVIDIA, literally every single major tech company is your customer AND they're all trying to build competing chips (Google TPU, Amazon Trainium, etc). That moat narrows fast.
I wouldn't be shocked if NVIDIA is up another 30% in 6 months OR down 40%. That level of volatility usually means the market doesn't actually know what's happening lol.