Question Global foundries future?

[TnTecks]

Why is there only one company on the planet selling bleeding-edge process tech? Global Foundries has a deal to make money from TSMC using its IP for the next 10 years. Intel is SOL and Global foundries abandoned development of 7, 5, and 3nm euv. Do you think there will be another company to take on the financial risk of process tech? Or is the international community happy with letting TSMC take all the risk, and most of the profits?

 

[turtile]

Why is there only one company on the planet selling bleeding-edge process tech? Global Foundries has a deal to make money from TSMC using its IP for the next 10 years. Intel is SOL and Global foundries abandoned development of 7, 5, and 3nm euv. Do you think there will be another company to take on the financial risk of process tech? Or is the international community happy with letting TSMC take all the risk, and most of the profits?

Unless the US government steps in, I don't think GF will ever be at leading edge again.

Europe is working on it though:

News - EU Signs €145bn Declaration to Develop Next Gen Processors and 2nm Technology

https://www.eetimes.eu/eu-signs-e145bn-declaration-to-develop-next-gen-processors-and-2nm-technology/ "In a major push to give Europe pride of place in the global semiconductor design and fabrication ecosystem, 17 EU member states this week signed a joint declaration to commit to work together...

forums.anandtech.com

I don't think they will be successful considering Intel, TSMC and Samsung will have massive capacity build out that manufactures cutting edge much more efficiently. Catching a moving target is never easy.

 

[gdansk]

Requires significant capital and an outside party to license some better technology. If anyone, I suspect Samsung to be most likely given their past dealings. They are perhaps more interested in Fab 1 than Fab 8 because of the EU efforts.

Otherwise it seems their plan is to cater to customers who have strange requirements or don't need the latest processes.

And no, it doesn't seem the industry is happy with TSMC domination. Samsung seems to be close/cheap enough that they are winning over some large customers (e.g. Qualcomm on 5nm and Nvidia on 8nm).

 

[darkswordsman17]

Requires significant capital and an outside party to license some better technology. If anyone, I suspect Samsung to be most likely given their past dealings. They are perhaps more interested in Fab 1 than Fab 8 because of the EU efforts.

Otherwise it seems their plan is to cater to customers who have strange requirements or don't need the latest processes.

And no, it doesn't seem the industry is happy with TSMC domination. Samsung seems to be close/cheap enough that they are winning over some large customers (e.g. Qualcomm on 5nm and Nvidia on 8nm).

Yeah GF's gave up. They straight up said they weren't going to chase cutting edge process tech at all and instead build for other markets.

Not sure I agree or rather they really don't have that much choice. Qualcomm did that before and left because Samsung just plain is not competitive to TSMC. I'm not sure the gap is closing either, but we'll see. Nvidia went with Samsung because of two reasons, they couldn't secure enough TSMC 7nm production and because Samsung gave them a ridiculously good deal (which many others turned down simply because Samsung has not shown they can really produce beyond 10/8nm).

 

[gdansk]

Yeah GF's gave up. They straight up said they weren't going to chase cutting edge process tech at all and instead build for other markets.

Not sure I agree or rather they really don't have that much choice. Qualcomm did that before and left because Samsung just plain is not competitive to TSMC. I'm not sure the gap is closing either, but we'll see. Nvidia went with Samsung because of two reasons, they couldn't secure enough TSMC 7nm production and because Samsung gave them a ridiculously good deal (which many others turned down simply because Samsung has not shown they can really produce beyond 10/8nm).

I'm pretty sure the gap is closing if only because the difference between TSMC's nodes is declining and Samsung wasn't a year late this time. Secondly, TSMC's 5nm achieved much worse than advertised scaling in revealed designs so far. But everyone still uses that to compare to Samsung's 5nm.

Samsung's is likely worse in practice too, but I wouldn't know how much. The TSMC 5nm vs Samsung 5nm designs available so far are so wildly different in other ways there is little chance to compare the process from what we consumers see.