OpenAI Samsung collaboration extends to next-generation chips

What it means for hiring

A memory supply agreement becoming joint work on next-generation processors is a different kind of relationship, and Samsung is an unusual partner to pick for it because it makes the HBM, the logic and the package. An AI lab that designs its own silicon with a partner who owns all three is not hedging against Nvidia in the abstract. It is building the option to stop buying.

For hiring this continues the most disruptive pattern in the market, which is AI labs recruiting chip people directly. They want physical design, design-for-test, packaging and signal and power integrity engineers, they want fifteen years of experience rather than five, and they pay on a software company's scale. Every semiconductor firm hiring for those roles is now benchmarked against a customer instead of a competitor, and the honest advice to a client is that the counter-offer will not work. What does work is the technical argument: at a chip company the engineer owns a product line for a decade, and at an AI lab they own one accelerator until the next architecture lands.

Read the original at eeNews Europe .