3 ms·
I’m skeptical. This reminds me of Theranos, because it’s (to me) an incredible claim, yet “a company and its executives wouldn’t lie”- though this is a fallacy
by gymbeaux 3y ago
I’m skeptical. This reminds me of Theranos, because it’s (to me) an incredible claim, yet “a company and its executives wouldn’t lie”- though this is a fallacy that predates even Enron ~20 years ago.
On one hand, clearly “AI” is in vogue right now. It seems like everyone is trying to hitch a ride on the ChatGPT train. On the other hand, historically, these are bubbles. People eventually realize the space isn’t that large, and the magic is gone. Like with NoSQL DBs or “big data”, or even just the internet in 2000.
- Our_Benefactors 3y agoHave you tried ChatGPT? It is the antithesis of theranos.
- gymbeaux 3y agoI use it all the time. It writes probably around half of not more of my code these days. It’s great… but I’m not seeing how that equals “all sorts of other companies also need a bunch of GPU power”
- adam_arthur 3y agoThe AI boom is very real. And much more pragmatic than the crypto or metaverse booms. The problem in the long run is the valuation. Its effectively a greater fool trading scheme for the next while here, as earnings are almost certainly not going to grow to support current valuation within the next decade or two. If you think all the FAANGs and AMD will sit by and accept NVDA being the monopoly provider of AI compute (not to speak of recent success using CPU+SIMD), then by all means, buy and hold at these levels. The higher NVDA margins go, the more antagonistic the rest of the market will be to undercut them, e.g. via better support for AMD cards, building their own, or otherwise.
- gymbeaux 3y agoI love that AMD is kinda-sorta paying attention to its AI offerings e.g. their CUDA competitor, ROCm, but I’m not seeing the next few years unfold as a massive W for enterprise hardware OEMs like AMD, NVIDIA, Dell, etc. as far as profits go.