Company is challenging industry leader ASML with its nanoprint lithography tech.
See full article...
See full article...
Canon plans to disrupt chipmaking with low-cost “stamp” machine
- Thread starter JournalBot
- Start date
I tried to find more info on the process.
Only found a press release from a quick Google.
I’d read a full deep dive into their process in a heartbeat
Only found a press release from a quick Google.
I’d read a full deep dive into their process in a heartbeat
Upvote
145
(146
/
-1)
It’s sort of ironic that Canon—a company most associated with cameras—would come up with an alternative to light-based chip manufacturing.
Upvote
184
(193
/
-9)
“
The electric car has existed since like the late 1800's. Yet here we are in an EV boom more than a 100 years later. Analysts in general are idiots who can't see two inches past their feet imo. The tech may have existed in the past, but maybe it needed time to mature, for the infrastructure to be in place, for the need for a cheaper alternative to develop.
Upvote
310
(315
/
-5)
Like marketing, they drive the car while relying exclusively on the rear view mirror. That way, their predictions for past events are perfect!
Upvote
151
(153
/
-2)
Well the analysts talk to upper management, usually MBAs, all of whom believe that research involves sending idiots into a lab to pick up stuff which just happens to be lying around, and why the hell does it cost so much?
Upvote
162
(165
/
-3)
Speaking of obvious, a 90% savings will do a lot to overcome momentum.
Now, about those yields you do not wish to discuss...
Upvote
117
(119
/
-2)
Upvote
87
(91
/
-4)
I have to assume that if it works even for the low end market, with those savings, this would still be a huge opportunity. The market isn’t shrinking by any means.
Upvote
87
(87
/
0)
So what kind of sizes of structures can it generate? Similar detail as euv? Or only larger structures?
Why is it cheaoer? Just to buy or to run? Why is it cheaper, is it Faster or just the lower power consumption?
Lots of questions left?
Why is it cheaoer? Just to buy or to run? Why is it cheaper, is it Faster or just the lower power consumption?
Lots of questions left?
Upvote
26
(37
/
-11)
Actually nope, canon's camera expertise translates into 2 things, optics and Micromechanics.
The nanoimprint thing is the micromechanics part of the operation taking the steering wheel
Upvote
123
(126
/
-3)
Yea, Japan has no interest in seeing a stronger China, the only reason there is a "might" is because up to now they did not have a product (or announced plans for a product), with this, the export restrictions will come.
Upvote
58
(58
/
0)
If it costs a digit less than ASML's EUV machines valued at 150,000,000 USD as stated in the article, even low income nations can have thriving semiconductor industries. This is a win even for TSMC, samsung, Intel etc since this frees up their production lines for more profitable, more complicated products like CPUs, GPUs etc. Memory chips, simpler CPUs etc can be made cheaper by newer players
Upvote
108
(108
/
0)
This will probably end sloted nicely between Inmersion DUV and High-Numerical-Aperture-EUV. You see, contrary to popular beliefs, top end chips are not made SOLELY using EUV, that would be too expensive.
Instead, leading edge processes use multiple process nodes for the same chip. The most critical digital features are made using EUV (say, 5nm or 3nm), and other features, like the analog parts (the connections to/from DIMMs and PCIe for example), or the bigger interconnects (say, power) and pads, are done in Inmersion DUV (10nm), so, two process nodes used on the same chip. But we just talk about the smaller node when refering to it.
Currently, there are a bunch of things that are done using EUV out of necessity, because Inmersion DUV stopped being economic around 10nm. A process like this one would slot nicely between HNA-EUV and Inmersion DUV for nodes below 2nm, so you could do the logic and 1st logic metalization in HNA-EUV, then do the other logic metalization layers and some other features with nanoimprint, and then do the power, pads and analogue in Inmersion DUV.
Upvote
162
(163
/
-1)
Whenever someone says this, I think that industry is ripe for disruption. Surprising comment for someone in the extremely fast moving technology industry.
Upvote
93
(94
/
-1)
It also opens up enormous opportunities for startups. Ex-TSMC/Intel employees might gamble on something new.
Upvote
47
(47
/
0)
5nm according to the article. Even if they can't make complex structures, and only simpler ones like memory chips. Moving memory chips production out of more expensive lines into 5nm cheaper lines would be really great.
Upvote
89
(90
/
-1)
The expensive part of the $40B plants aren't just the ASML machines. Making the silicon wafers is also really expensive.
Upvote
45
(47
/
-2)
Upvote
11
(22
/
-11)
Exactly, and I am glad someone replied right away because it’s important history. Before ASML became the lithography leader with EUV, the space was dominated by Japan companies like Nikon and Canon, which are known to most people as just a camera company. Japan was known as a semiconductor powerhouse up through even the 1990’s, and this was part of the reason. ASML was rising by then until the climax of ASML vs Nikon in court over lithography related patents in the 2000’s. Also, when you look at a lithography machine, you see it’s like a very complex projection system.
Upvote
117
(117
/
0)
No joke.
Out of curiosity, I searched this company that they quote and it seems to be a one-person outfit (though they list some other company that actually produces their reports?) and it doesn’t seem to have any specialization in chip manufacturing technologies; it’s a generic mobile and digital consultancy.
Not sure why Ars is quoting this guy. They‘d have been better off going to, for example, an ex-GF/TSMC/Intel guy, or, I dunno, Yield Engineering and asking. Maybe they owed him a favor?
Upvote
91
(94
/
-3)
Asianometry is great for chip, OS, and other history, too. Really great job
Upvote
35
(35
/
0)
yup I'm pretty sure a certain neighbor of Poland would really like a lot of low cost chips.
Upvote
15
(19
/
-4)
Wow, I'm surprised I have not heard of this before.
It is interesting (?) to note that Canon was a leader in the field back in the early 1970s with their contact and hold-off tech.
But the Perkin-Elmer stuff murdered them in the mid-1970s. And after many more years, that tech ended up being owned by... ASML (well, sort of).
Turnabout is fair play!
It is interesting (?) to note that Canon was a leader in the field back in the early 1970s with their contact and hold-off tech.
But the Perkin-Elmer stuff murdered them in the mid-1970s. And after many more years, that tech ended up being owned by... ASML (well, sort of).
Turnabout is fair play!
Upvote
15
(18
/
-3)
Upvote
59
(61
/
-2)
Not really, they have been involved in lithography before, they just stopped trying to compete at the high end for a while, like everyone else not named ASML. At that scale it's all optics, regardless of what they call it or what new tech they come up with to make it work.
Upvote
33
(36
/
-3)
No.
Yeah, this is completely wrong.
There were three key advances needed to make a practical EV:
1) the li-ion battery, more specifically, the polymer layers that allow lithim migration.
2) large-format thyristors and/or SMPS systems, which allow the DC from the battery to be efficiently converted into AC for the motors.
3) the control systems had to be fast enough to react to control all of this.
(3) was available in some format in the mid 1990s. (2) emerged in cost-effective form in the late 1990s, early 2000s. (1) emerged in practical form in the mid-2000s.
EVs emerged a couple of years after these advances emerged. None of them could have been developed much faster, and none of them would have been influenced by the petroleum industry, one way or the other.
It's a nice conspiracy theory, but not a true one.
Upvote
111
(116
/
-5)
Everyone's obsessed with cutting edge nodes, but I wonder if this could help backfill legacy node capacity in an economical way? For instance, 28nm has a massive presence in all sorts of tech and yet nobody really wants to build those fabs.
Upvote
56
(57
/
-1)
Upvote
10
(13
/
-3)
Mitutoyo sandbagged on the quality of their metrology to get around ITAR. Companies do not give a shit about anything but making money.
Upvote
18
(18
/
0)
Oh sure the chipmaking MACHINE is dirt cheap but how much does the chipmaking INK cost?
Upvote
135
(141
/
-6)
Exactly, and I am glad someone replied right away because it’s important history. Before ASML became the lithography leader with EUV, the space was dominated by Japan companies like Nikon and Canon, which are known to most people as just a camera company. Japan was known as a semiconductor powerhouse up through even the 1990’s, and this was part of the reason. ASML was rising by then until the climax of ASML vs Nikon in court over lithography related patents in the 2000’s. Also, when you look at a lithography machine, you see it’s like a very complex projection system.
View attachment 73048
Looking at that graph and the part of the article about how Canon has been working on this for 15 years. It looks like they bet on this instead of EUV some 15 years ago, and that problems into getting into production really bit them in the ass.
Upvote
43
(43
/
0)
They are also longtime makers of inkjet printers. Nanoimprint lithography is probably completely unrelated to their bubblejet printer technology, but if you read the details some of the nanoimprint mask making sounds weirdly similiar.
Upvote
15
(15
/
0)