They have such a large lead that obviously some people will catch up. TPUs handle LLM workloads in a more cost-effective manner. While NVIDIAs H200 is better suited for a wide range of workloads, which offers protection from the, albeit small, possibility the LLMs are a flavor of the month. They also perform fine with LLM workloads, but at a higher cost.

My understanding is the Blackwell chips have addressed the workloads that Google's TPUs are arguably better at now. The Blackwell's are going to lap the competition in nearly every aspect. Everyone else is going to compete in the price/performance category, which is inherently lower margin. I do think that moving off the cloud to the edge is clearly an important trend and NVIDIA has no real advantage there.

The market is growing so fast that you can lose market share and still grow like crazy. NVIDIA margins are quite high right now and I think it is very reasonable to expect some compression. However, until there is actual legitimate competition for them at the highest end of performance, I have no real reason to consider selling the position or believing it’ll drop relative to the market.

NVIDIA is expected to have a 97% increase in sales next year. That is just insane for a company of that size. And maintain gross margins of 76% that can compress a lot without me becoming overly concerned. The street doesn't see any meaningful compression through 2027, which is a very rosy outlook.

It’s definitely priced for perfection right now, but for me still a long term hold.