Electronics Weekly Electronics Design & Components Tech News
It’s obvious why not – there aren’t any benefits. The goal of boosting EU chip output to 20% of the world market has zero chance of being achieved.
The aspiration of having leading-edge chips made on European soil depends on a single proposed initiative – Intel at Magdeburg – which looks increasingly unlikely of happening any time soon, if at all.
As usual, any money dredged up under Chips Act 1 appears to have been shuffled away to the usual suspects – the big companies – without any obligation on their part to deliver anything tangible in return.
It is reported that ESIA is now asking for a chip envoy whatever that may be, reduced export restrictions, renewed focus on areas where European companies already had advantages (i.e. more money for the big boys) and that aid should be awarded more quickly (hurry up dispensing the Slush Fund this time).
If market demand warrants investment in capacity then the industry will make it. If market demand doesn’t warrant investment then hand-outs won’t get companies to build overcapacity because there is nothing so expensive as unused capacity.
We’ve seen that with the US Chips Act – projects which have been funded under it are stalled waiting for a revival of demand to justify expansions in capacity.
This European Chips Act 2 idea smacks of an industry which has become too fond of sucking on the taxpayer’s tit.
As it happens Mike Leixlip’s best process is Intel 4 which was originally called the Intel 7nm process. While an advanced process, I don’t think you can say that 7nm is a leading edge process.
It is for Intel. It’s virtually the same as TSMCs N4 process which last year’s iPhones were made on so that’s only a year behind the absolute leading edge.
I believe all the equipment is suitable for Intel 3 which is the same as TSMC’s N3E which this year’s iPhones use so it can do that when the demand is there.
But Intel 3 and N3 are the end for FinFET at both places. After that you need the new NA machines, but neither company has those up and running in production.
TSMC’s N3 might be comparable to Intel 4 but N3 has been in volume production for at least a year longer than Intel 4 and presumably has multiple customers and much higher volumes than Intel 4 while N3e is 25% denser than TSMC N3 and Intel 4. So there’s a fair old gap still between TSMC and Intel.
TSMC will always have more volumes, and I agree Intel is about a year behind. But that’s still far better than it was not that long ago during the 10nm debacle.
N3e is designed for phones which Intel doesn’t do. Intel 3 is roughly same density as TSMC N3p, the performance node which is more what Intel need for server chips, and Apple are using in the next Macs.
But both have now reached the absolute end stops on FinFET so it will be interesting to see how they compare on the Nanosheet products.
Unfortunately it looks like Intel 18A hasn’t been that good so far on prototypes delivered, but we haven’t heard much about the equivalent TSMC N2 either – it’s certainly running too late for next year’s iPhone. Plus of course both depend on ASML NA machines of which I believe they have one each.
Intel has two
OK but the point was that the European Chips Act hadn’t delivered a leading process. And, BTW, are you sure Leixlip is running 3nm chips which is the current leading edge process?
Nothing in a “3nm” chip is 3nm as you well know. It’s pure marketing-speak.
Intel 3 = TSMC N3
Intel 4 = TSMC N4
You appear to be forgetting Fab 34 at Leixlip. Fully equipped with EUV machines, these can fab the Intel 4 process which is far in excess of anything else made in Europe. I know it’s not on European soil but it is in the EU.
Most of the Leixlip money – about $11bn – came from the Apollo private equity group with a much smaller amount – less than $1bn – coming from the Irish government. I’m not sure EC money came into it,
No it didn’t. But it’s still a European source of state of the art ICs at by far the most advanced fab in Europe. So leading edge ICs are made in Europe whereas your text implied there weren’t any.