Of course, the competition in the GPT space will be absolutely brutal. There's no reason to think the 10 unicorns listed there can not roll their own GPT bots. Or that 10 new GPT startups won't appear between now and the end of this year. Ah, ... make that 100.
Still, if you are a VC, you invest in what exists right now.
[1] https://www.analyticsinsight.net/top-10-ai-unicorns-that-are...
What businesses have the secret sauce being the proprietary features of one vendors chips? Heck, a number of the big computing companies (in various areas) also have custom silicon.
Here's the thing: building a GPT bot is not that difficult, you can do it with a few lines of keras, for example [1]. The difficult part is training one. Presumably once you trained one, you can train another one. That's why the maker of Stable Diffusion is valuable.
[1] https://keras.io/api/layers/attention_layers/multi_head_atte...