Intel negotiating for over $10B in subsidies under the Chips & Science Act

'Intel’s potential financial package is expected to consist of loans and direct grants, although the exact distribution between these is still under negotiation.

Earlier, Intel had reportedly demanded a greater share of these subsidies because it was ‘putting the business at stake’ by committing to fully local R&D and manufacturing in contrast to the internationally hedged bets by TSMC et al

Does anyone want >= 10nm chips? Intel have not yet shown they can reliably shrink below that. TSMC, Samsung et al have proven that they can and are thus more desirable for the USA imho.


To answer your question. Yes, lots of people want chips in 10nm or even older. Remember the chips shortage a couple of years ago. It was mostly because of the inability to get enough chips from old FABs that were low cost and thus not good candidates to redesign for newer FAB tech, such as chips used in cars.



I’ll take a stab at this:
Intel_4 is shipping in volume, and technical reviews rate it as better than TSMC N5. Intel_18A has around 40 customer test chips running in the factory today. Intel has said Intel_18A is better than TSMC N3, while TSMC has said they are similar. Jensen Huang of NVIDIA has said “that the company is working to diversify its chip manufacturing and had recently received good test results for an Intel test chip based on the company’s next-gen process node.” Intel has lined up both Microsoft and Open AI CEO’s for their IFS direct connect event this Wednesday. Also appearing are ARM, Broadcom, and Mediatek presenters.

From a futures point of view, Intel_18A has both ribbon fet and backside power delivery which are not due from TSMC until their N2 which is 2025/2026. Intel is the only company with a high_NA EUV stepper in house.

Based on past performance it is reasonable to assume Intel processes won’t be ready until at least a year later than Intel projects. Even given this assumption Intel_18A is very near if not ahead of TSMC N2.


Alan, your rational well-informed perspective is always much appreciated, but sometimes it’s depressing to this AMD “fanboy”. :wink:


I accept alan81 knows quite a bit more about what Intel is trying to do, than myself. Given their recent history though, there must be some doubt that Intel can deliver.

Hi Sorcery - thanks for your response
Here’s a chart showing the share of manufacturing by node for TSMC in '22
~77% was in >10nm node chips

So yes, the vast majority of such chips are for more mature technologies, where competition is much more intense due to the diversity of vendors and use cases

1 Like

There is certainly more than “some doubt”. I do think it is more a matter of when, than if. Given the slowing of Moore’s Law, time becomes a little less critical than it has been historically. There are certainly some smaller milestones along the way, but I don’t think we know if Intel has pulled it off or not until the end of 2025.

1 Like

Where TSM excels is in yields. Even Samsung has not been able to match TSM
So, success for INTC is not about getting test chips functional, but whether they are able to get manufacturing yield at scale.


I would be a little careful with the yield belief. It has been widely reported that TSMC has been “struggling” with N3 yields. It is also important to note that for a typical cell phone die of 80mm^2 typical yields will be around 90%, while with a much larger X86 CPU die they will drop to around 75%. This is an area where the AMD chiplet strategy has served them well as they are making X86 devices out of a collection of much smaller dies. Given Intel is currently manufacturing some very large die in large quantities we can infer their defect density (data underlying the final yield %) is actually quite good.

Intel is in the middle of transitioning to chiplets, and will begin to benefit by manufacturing smaller die and “gluing” them together.

Samsung’s yield problems have been legendary though:-)