Canon ships its first nanoprint lithography machine, rivals ASML | Hacker News Reader

Canon ships its first nanoprint lithography machine, rivals ASML

datacenterdynamics.com

67 pointsmgh238 comments

For those like me who haven‘t heard of ASML before. There was a recent NPR Planet Money episode (26 minutes) about the company. I found it quite interesting:

https://www.npr.org/2024/11/13/1212604208/asml-euv-extreme-u...

When Canon first started developing these machines, I wonder if they knew that one of the biggest markets in the world, China, would be completely cut off from them.

I can see these machines succeeding if China is allowed to buy them. But with just a handful of semi fabs remaining and the most cutting edge chips in much higher demand, what is the market for these machines?

> if China is allowed to buy them

The US controls what Japan can and can't do, so it will be a NO.

The question isn't if Canon can sell to China now. The question is when Canon decided to invest R&D into this machine, did they have any idea that China would be cut off from this product?

I assume that Canon definitely wanted to sell to China.

But now the market is severely limited. TSMC uses ASML. Intel uses ASML. Both aim to manufacture the most cutting edge chips. That's where the money is. In addition, both TSMC and Intel own a fairly decent amount of shares in ASML.

>The question is when Canon decided to invest R&D into this machine,

They have been doing this R&D since the 90s. I doubt they think much into it whether China have access but more whether it could compete with traditional tools.

Canon bought the tech in 2014 after working with the original startups from the 90s:

https://www.youtube.com/watch?v=8UdNB3ZY4Ks

Since the tech can't compete with EUV in volume, their market is probably niche/lower production runs at a significantly lower install price.

If there's one thing you want if you're highly dependent on a single vendor or supplier, it's a competitor to keep them in check. Expect the big players to buy a bunch of these, even if they do own shares in ASML.

It would be sad to see the competition die just because it's a little late, especially in such an important and cornerstone piece of technology today.

Japan has their own disagreements with China, don't think they would even if the US would allow it.

Don't think that's true. There were many Japanese firms who were displeased when the US government forced them to stop serving the Chinese market.

There are many US firms that are displeased not being able to serve the Chinese market...

Nvidia, AMD, etc they all build products specifically for designed to go around export controls. Their shareholders would be pissed if they didn't.

For-profit corporate interest trump whatever historical geopolitical disagreements countries may have.

That's why government always step in with regulations, because corporations only care about selling to the highest bidder for profit no matter how that would affect national interests.

And today China is the most profitable county you can sell to if you're in the lithography business. Banning sales of equipment to China is something all lithography companies hate.

Absolutely. The point I was making though, is that regulations could very well come from an initive of the Japanese government.

Of course the reality is that both the US and Japan are involved in chip restrictions, as both have to gain by restricting China's access to advanced technology.

https://www.ft.com/content/aad4f7b3-0a0f-46ca-9449-14fedce0e...

The US has had to apply heavy pressure to both the Dutch and the Japanese to get them to restrict exports to China. They wouldn't have done it if not for the US demanding them.

https://www.ft.com/content/3fa44901-33e4-4ab4-9f7b-efe1575a6...

Japan is a sovereign country. They are closely allied to the US now but that doesn’t mean it will always be the case nor that they have to silently agree to every stupid mandate the US tries to impose.

>Japan is a sovereign country.

So is China. That didn't stop the US from retaliating economically when they do things the US didn't like. Just because countries are sovereign doesn't mean the US can't force them into compliance, allies or no allies.

Why do you think the US spends more on military than all the other powers combined? So they call all be sovereign from the US? Or for the US to have enforcement over world geopolitics and influence over the other powers to keep being the world reserve currency?

No of course they don’t have to, but for the foreseeable future it’d be criminally stupid not to (on this particular point).

This good be a big leap forward using the chips act and access to more players good usher in a golden age of development. The cottage industry of chip production. Cheaper access to tools is what drives innovation from my point of view.

"The lower price point of Canon’s nanoprint machines could potentially allow smaller manufacturers to make leading-edge semiconductors."

> the price would be “one digit less than ASML’s EUVs.”

if thats true it will surely be a game changer, if not immediately surely in the medium term?

geopolitics, oligopolies and oligopsonies notwithstanding, if a technology slices an order of magnitude it does change the landscape.

The ASML machines are modular and each generation gets new upgrades that pushes Moore’s law further. For example, with the previous generation, the Chinese SMCI modified the machines to be competitive with the newer chips from EUV machines. So there is a lot of expectation on the new EUV machines still. Also in terms of productivity.

With the stamping approach there might be less upgradability possible. But it might also still unlock new markets.

Due to the way it works speed is also several digits off.

I think that’s a sign that ASML hasn’t had much competition at that scale. And that the new tech requires buyers willing to take a risk and invest in learning the new process.

Actually, Canon used to be the one of the main players, whereas ASML was the new kid on the block. It with new technologies and high reliability that made ASML win.

There are hundreds of additive steps in a process. Even if one step has zero price, don’t expect the total to collapse.

Does this come under the US controls on high tech equipment that use IP from US companies?

This appears is a direct write on wafer machine. Is it likely way way slower than a lithography machine?

How do they make the mould?

see e-Beam system Wikipedia page

Thanks! Here is the link:

https://en.wikipedia.org/wiki/Electron-beam_lithography

hostile de-globalisation eroding single supplier dominance through the SC supply chain. Good opportunity for more players to take smaller slices of a growing pie

nobody was stopping them before the hostile de-globalisation which increases the risk of conflict by reducing what each side has to lose.

Artificial incentives are artificial

I wonder about the TCO of nanoprint. The ASML TwinScan is only a part of the CapeX and part of the cost in OpeX of a Fab. You will need to take into account how many wafers it can do per month including down time and maintenance. Yield. Cost of Design and EDA. Whether tools are portable. BOM and running cost or basically electricity usage. EUV TwinScan is extremely power hungry.

And then you need to take into account of its other design constraint. Does it work equally well with analog. Obviously there would be less ready made design blueprint available etc. And if it was that much cheaper you would have seen it being used in commodity like DRAM and NAND. It is not which suggest there are missing / downside information we dont know.

Volume is only one route to profitability. Smaller scale is fine for smaller markets with larger margins.

I am not sure how my comments not relates to volume. NAND and DRAM is precisely where all the volumes are.

I'm not exactly sure where we are misunderstanding each other. I just wanted to point out that volume is a goal generally of commodity goods. I see opportunities in niche markets that don't rely on volume for profitability.

Specific NAND and DRAM are only needed because general purpose computing production currently relies on large volumes of RAM and techniques developed a market to meet the need.

A niche market could solve computing problems that avoid NAND and DRAM by solving computing problems with another method.

The PC market as it stands right now has developed through commodity solutions but that isn't the only way to skin a cat.

Reply on news.ycombinator.com