OpenAI is partnering with $Taiwan Semiconductor Manufacturing(TSM)$ to develop custom AI chips to bypass $NVIDIA(NVDA)$ .
I think this is a terrible mistake for OpenAI.
OpenAI is still a sensor company, an application build on infrastructure. It hasn't reached critical mass to tip the market and still has competitors.
Shifting focus to infrastructure before you win the market comes with two costs:
1) It allows competitors to catch up building on the standard infrastructure.
2) Skyrockets the costs as chip design is extremely capital intensive.
This could result in distraction and ultimately losing the LLM market. If this happens, chip design efforts will be already doomed as it'll lose its revenue generating segment.
It's best practice to shift focus to owning the infrastructure after you win in the sensor market, just as $AMZN, $META, $GOOG and $MSFT have done.
This is too soon for OpenAI and it can't afford such distraction right now.
Image
Comments