No one can deny that Intel’s current 10nm was such a big problem that AMD and TSMC beat them to it. The 14nm have spread so much over time that Lisa Su’s have not had much resistance until the Core 12 with Alder Lake. But Intel has woken up and now important players in the industry say that it stepped on the accelerator to such an extent that it has managed to cut its cutting-edge process by 12 months 18A (2024)which is a knockout blow that could topple TSMC and AMD, or isn’t that enough?
There is currently a lot of controversy over the times and the technology that both Intel and TSMC will use to manufacture their cutting-edge nodes. TSMC is the most controversial in this, since the 3nm do not finish taking off, the 5nm seems that they did it only 6 months ago and will arrive a year after its staging and of the 2nm little is clear. But Intel is another story and could make up ground in just two years.
Intel Global Strategy, IDM 2.0 and IFS
Without a doubt, the turn of the wheel offered by Pat Gelsinger has caught TSMC off the hook. The accelerator stomp is forcing efforts to be maximized in Taiwan while Intel itself supports the Asian giant. Why would it do this if it is direct competition in semiconductors? There are two reasons: the first is, apparently, because contract. TSMC is very good at manufacturing, but he is very bad at designing and it needs American talent to improve on key aspects such as the transition to GAA transistor technology.
Second, Intel deals a blow to NVIDIA and AMD by reserving production for the help provided, so their graphics cards will have cutting-edge or optimized nodes for the occasion when they are on par with their rivals in performance. Thirdly, and more on the periphery of the arguments, Intel is assured of production capacity and can continue to produce its chips without affecting its FABs, a full-fledged “divide and conquer”.
Here you have a lot to do global IDM 2.0 strategy and of course its IFS foundry services now open to any company, a double whammy to its rival since you can mass produce more, at low prices and not lose slack with the best performing nodes. TSMC, for its part, is guaranteed production and support, which it is surely learning about Intel’s designs and technologies, but does it really pay off?
Intel advances its 18A to 2024
Intel has room with all these arguments to produce more chips, cutting-edge or not, and can also squeeze in key processes with more and more advanced nodes. The objective was to arrive first and, according to the latest rumors, they are achieving it, since their nodes 20A and 18A reduce their times considerably.
Both lithographic processes already include GAA transistors, something that TSMC will achieve with the 2nm and not without problems seen the rumors of almost a month ago, but this does not matter to Intel because even if TSMC arrived in a timely manner they have achieved a milestone : the 20A node could arrive next year for low power chips and the 18A would arrive in the second half of 2024entering mass manufacturing in early 2025.
This means that at least Intel would have gained 6 months to the clock, if not really 12 if it manages to meet the most optimistic forecasts. In addition to this, it must be taken into account that the traditional 18A or 1.8nm would arrive with GAA RibbonFET transistors and also being manufactured with the first scanners EUV High-NA, so the blow to TSMC could be greater than previously thought in terms of time, because they were supposed to arrive at the end of 2025 and would enter en masse at the beginning of 2026. Will TSMC be able to replicate Intel with its 2nm and in a timely manner?