Doesn't feel all that long ago when Intel pushed Nvidia out of the chipset market, and how that felt almost existential for Nvidia at the time.

Tables have turned a fair bit since then.

Now we know why the Lip Bu Tan CEO of Intel declared they would stop competing with Nvidia.

lunar lake is pretty good

Maybe because it is manufactured by TSMC? Was that your point? https://en.wikipedia.org/wiki/Lunar_Lake

Nvidia needs more silicon manufacturing partners to prevent TSMC from charging a significant premium. Their choices are very limited today, and almost non-existent outside of Asia.

Given a real geopolitical risk and NVidia’s current valuation - 5 billion seems cheap if it secures them an alternative manufacturing partner.

As part of this deal, is Intel doing any GPU fabrication?

Isn't it limited because TSMC is just so good at their job? I thought I read places that their error rates are lowest in the industry.

Which leads to multiple questions - if NVidia can't manufacture elsewhere, will the best in the industry get much more expensive? is the GPU value in CUDA or TSMC?

At best, the two companies are locked into a monopsomy/monopoly market for the next 2 years where TSMC needs NVidia to sell chips and NVidia needs TSMC to make them - in the long term, these arrangements are rarely stable. NVidia wants more manufacturing partners, and TSMC wants more GPU makers.

More discussion at https://news.ycombinator.com/item?id=45288161

> Under the new agreement, Intel will build custom central processing units for Nvidia to build into its AI platforms for data centres, as well as a separate project to develop PC chips.

So this will let both NVidia and Intel increase sales for data center applications.