Too Few Gen AI Dollars Chasing Too Many Gen AI Efforts
The faster U.S. companies scale LLM buildouts, the better for Gen AI. However, the effort to scale LLMs is too fragmented.
While I like what Google is doing in terms of designing its own chips (in addition to using Nvidia chips) to build proprietary LLMs, I do not like the fact that there are probably a dozen or so other companies building competing LLMs. From my perspective the LLM buildout is really a two-horse race between Google and OpenAI with Amazon’s AWS unit worthy of honorable mention. META will prove to be an also-ran in the space (Zuckerberg knows how to waste capital and resources), Oracle is subscale and Groq (founded by ex-Googlers), is doing interesting things on the chip side. Microsoft is nothing in Gen AI without OpenAI.
I would like to see Google acquire OpenAI, put all Gen AI resources into OpenAI and spin it out with Sam Altman as CEO and Google as the largest equity holder with 51% control. OpenAI is not doing itself any favors by partnering with Microsoft and is only slowing the pace of Gen AI innovation. Gen AI investment capital would be better put to work investing in a combined Google / OpenAI effort. Google ought to acquire Groq while it is at it. Groq’s founders founded Google Tensor AI chip effort.