You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Readers probably know of a somewhat countercyclic natures of the industry: economic downturns often lead to both demand shifts towards lower end, higher volume products (and thus revenue.)
But I believe few have full explanation for what's happening now. The industry saw both bumper profits, and huge revenue expansion at the same time.
The quotes for really old legacy nodes are on all time high, and even 200mm is back from the dead for non-marginal logic use.
Equipment purchase is on all time high. 30+ new fabs are coming online in the next 3 years.
I previously supported the notion that 2020 will be the beginning of a "long winter" for the industry as it runs out of growth drivers like smartphones, node race will leave the world with only a dozen companies to be able to afford the latest node, and the acquisition frenzy in big-micro will lead to further reduction in competition for "cash crops" of the industry. Now I am completely retracting this.
So, what's the real reason for the such strong, and unexpected upturn?
Covid created a semiconductor boom. Companies invested in remote work (PC demand), cloud (HPC demand), automation (IoT demand), and consumers have been buying TVs, gaming devices, and other electronic devices to keep themselves entertained in quarantine.
The economy has been weird through Covid. It's caused a recession in some industries but a boom in others.
In general, I don't think we'll be seeing less electronics any time soon.
I don't think silicon FET shrinkage for computers has much more than a decade left in it (clear skies till 2024, though) and it's growing exponentially more expensive. But with the increasing automation in industrial, automotive, medical, etc, and people's just incessant demand for more networking bandwidth, I think the field's just going to continue to grow.
In general, I don't think we'll be seeing less electronics any time soon.
I don't think silicon FET shrinkage for computers has much more than a decade left in it (clear skies till 2024, though) and it's growing exponentially more expensive. But with the increasing automation in industrial, automotive, medical, etc, and people's just incessant demand for more networking bandwidth, I think the field's just going to continue to grow.
3nm will probably the the last FinFET node if that's what you mean. After that I expect another 5-10 more years where GAA allows for continued progress on node size but at a slower rate. Eventually a wall will get hit where the economics of node transitions no longer make sense even for leading edge applications, and once the leading edge stops moving forward the Chinese fabs will catch up and potentially commoditize the market. Maybe we will have a new computing paradigm (quantum?) by then.
3nm will probably the the last FinFET node if that's what you mean. After that I expect another 5-10 more years where GAA allows for continued progress on node size but at a slower rate. Eventually a wall will get hit where the economics of node transitions no longer make sense even for leading edge applications, and once the leading edge stops moving forward the Chinese fabs will catch up and potentially commoditize the market. Maybe we will have a new computing paradigm (quantum?) by then.
There may not be decades of device scaling left, but surely decades of device improvements. For the last few nodes, the industry already didn't have as much of device scaling as device density improvements thanks to COAG, new metals, and routing, cells optimisation, isolation, etc.
My bet, we will see 1B devices per mm² in this decade.
I agree, there will continue to be improvements. I've often used the analogy of cars and internal combustion engines, where from the late 1800s to the mid 1970s engines became faster and more powerful. But this trend more or less peaked with the muscle cars of the 70s and since then improvements in cars has been more about things like safety, fuel efficiency, driver and passenger comfort, in car entertainment ect. Until now we're finally moving on to a new EV architecture and significant improvements are once again possible. So I'd expect something similar to unfold with computing and semiconductors. Computers will still get better, but it'll be about other features like security, power efficiency, additional features vs raw improvements in performance driven by density - until we transition to a new computing architecture.
3nm will probably the the last FinFET node if that's what you mean. After that I expect another 5-10 more years where GAA allows for continued progress on node size but at a slower rate. Eventually a wall will get hit where the economics of node transitions no longer make sense even for leading edge applications, and once the leading edge stops moving forward the Chinese fabs will catch up and potentially commoditize the market. Maybe we will have a new computing paradigm (quantum?) by then.
I guess post FinFETs we will have nanowire and nanotube FETs which are already lined up (CNTFET). So they will go another node or two to the least. That should take us to the 2030s.
And mean while how does 5G unfold willgive us the idea of what else will shape the entry into the 30s... I mean 2030s.
At a cloud computing conference when cloud was getting into shape, I remember someone showing an old ad stating that in the kitchen of the future will only have a microwave. We know how it played out.
So this SEMI, 5G etc will make room for themselves but still leave space for the others as the cookie crumbles.
