As we know well from the number of topics we have covered in this regard, it is ASML that has the upper hand and the one that sets the pace for the industry. The complexity and advancement of its scanners are of such caliber that by a pure investment, no one envisages making them a real and direct competition in the most advanced technologies.
It is true that in the most common scanners there is some competition, but in the rest, which is important in semiconductors, ASML stands alone and therefore knowing what he has prepared for us is synonymous with understand how far Intel, AMD go. go and NVIDIA.
ASML updated its roadmap with important news
The sections where the company will once again explode the market will this time be divided into three very clear and differentiated sections: DUV systems, EUV systems and EU-High-NA systems.
Obviously, these three market options are aimed at different types of products, but the truth is that in any case, the market for DUVs is more and more rare and logically everyone is oriented towards the UVU not to really be lagging behind its competitors. The problem, as has been mentioned many times, is that the number of scanners is very limited each year, and although ASML is increasing the volume of production, there is an open war to buy more units.
For this reason, in the three systems that we have named, there are novelties, which will mark both the technologies, as well as the yields as well as the availability of wafers in production, crucial in the midst of the semiconductor crisis. that we are living. .
Although it sounds strange, the reality is that TSMC, Samsung or Intel are still using DUV in a mixed process with EUV. As we know, wafers and chips are created in layers, so immersion technology is still needed in some masks.
For this reason, ASML introduced its fourth generation of NXT scanners, in particular the new NXT: 2050i. The improvements focus on the new wafer manager, the wafer stage, the projection lens, the laser pulse extender and the lens hood itself.
These improvements allow for more accurate and faster measurements and a higher degree of wafer deflection correction, which not only results in more functional chips, but also increases throughput per hour (Wph) by 275. pads to all 295 pads.
This if we are talking about ArFi technology (immersion), but if we are talking about dry technology (ArF), the news focuses on a new NXT: 1470 what offers 300 WpH, being a little faster than its counterpart in ArFi, but it will not become the best version of itself, since it is a second generation arriving with the same n ame and a performance of 330 WpH
This will to some extent prevent the supply from being so scarce, the problem is that in the midst of the semiconductor crisis, leading companies such as those mentioned may be content with their current scanners in order to purchase EUV.
EUV at 0.33 NA
Or just EUV, since we are currently at this wavelength. Intel already has scanners with this technology and this length, like obviously Samsung and of course TSMC, from which will come among others the 7 nm + and 5 nm, as well as the 1z RAM of the Koreans.
We have been working with scanners for a year and a half NXE: 3400C and now in the midst of a shortage ASML announces the new NXE: 3600D, which are already in the process of being booked and will be sent from the end of this year if all goes well.
The improvements are not clear in this regard, there is a bit of controversy and speculation as the scanner is arriving with a certain delay, but if we have to quote something, the rumor is that it will have more power, better. optics, a faster phase change and will be suitable for the 3nm system and below.
This is also without counting the increase in WpH, since we went from 135 in the already veteran NXE: 3400C to 160 in the new scanner. NXE: 3600D.
High EUV-NA 0.55
The high wavelength for EUV was a challenge not too long ago and is now more of a tangible reality than anything else. The slide shows this, with a WpH between 185 and 150 depending on the mj / cm2 applied for the engraving.
The news is not known, there aren’t even any leaks on how it was possible to reach and climb to 0.55 NA, but we have an arrival date for the scanners. . EXE: 5000 and especially new ones EXE: 5200 this will happen in 2025 and 2026, depending on the evolution of wavelength techniques.
We know that the 0.33NA can record at a step of 30 nm with a single exposure, while work is underway to reduce this figure to 28 nm and possibly end of life up to 26 nm, at least in several models.
This can also produce the 3nm TSMC with the so-called Multi pattern, but from there it is not believed that they can be reduced without a greater wavelength. So much so that everything is complicated, that new scanners will be needed and the engravings with them will have to be directly double-patterned, at least for the 2 nm TSMC and 5 nm from Intel.
But also, although the difficulty is greater, the performances will also be done by scanner, since they work on EUV masks with greater capacity and absorbent layers, also on the improvement of photoresists, key part for an engraving more perfect.
Therefore, the future does not seem easy, but at least ASML has in mind and in design new scanners that will go down from the 26nm of Pitch Metal, which will produce nodes with smaller transistors, below the 2 nm and 5 nm on TSMC and Intel. The problem on the other hand is obvious: investment is skyrocketing, costs for more of the same, and node-to-node time increasing, so scaling performance is going to be really tough for processors. and GPUs.