Do they need EUV to make RAM? Doing a small amount of searching leads me to 2025 press releases from companies saying they're first to a new process node, and mentioning EUV like it's an innovation.
I assume they could scale faster with more machines of the older, more understood, lithography technology.
In the coming years, EUV mastery will almost certainly be necessary to impact the global RAM market. For context, Samsung started using EUV for DDR4 in 2020. However, DDR4 is generally fine without EUV, it's just as you move into producing the latest DDR5 and HBM3 RAM, the "multi-patterning" techniques that CXMT uses in lieu of EUV result in very low yields (economically uncompetitive). It might be possible to use ultra-low PPP to simply outproduce the rest of the world at competitive prices and maintain reasonable margins. For example, it can cost 10x more in CapEx to build a chemical plant in the USA vs. China. With respect to marginal profits, if China's solar ambitions result in extremely low cost of electricity, their cost for synthetic fused silica could end up being very cheap, and potentially use electrical plasma chemical vapor deposition to create super high quality truly anhydrous silica (if electrical plasma chemical vapor deposition is used). A low price for synthetic silica would also allow CXMT to be competitive at lower yields than EUV achieves.
But over the next 5 years as technology marches on, and high tech RAM production moves from EUV to High-NA EUV to Hyper-NA EUV processes...multi-patterning will likely cease to be feasible. So while China is very much on track to achieve truly astonishing RAM market impact from 2025 to 2030...it then hits a pretty hard wall and China really will need to finish developing and deploy EUV at scale over the next 5 years to achieve market impact past 2030.
I personally wouldn't bet against China achieving a well-scaled EUV deployment by 2030. It's not guaranteed by any means, and reasonable minds might be skeptical because EUV is fucking hard. But they're doing all the necessary things to achieve it: poaching the right engineers from ASML and TSMC, stealing technical documents, funding it generously, and removing red tape. China reported that they got EUV working in pilot plants by the end of 2025, and those reports are considered credible. Another 5 years to polish their EUV tech and scale out EUV deployment seems very reasonable to me.
1. People who are currently buying AI services realizing it's not all that useful to them and discontinuing their subscriptions. Note that this can come from a changing ecosystem as much as anything to do with the products themselves. I know a couple people running AI propaganda operations where a single person can now do what previously took a major media conglomerate; this is great for them, but if I personally know a couple folks doing this, it indicates that there are probably hundreds of thousands worldwide, and people are simply going to stop trusting anything they read on the Internet.
2. Rising interest rates from the Iran war. Suddenly the cash flows needed to finance all this datacenter and AI model expansion are much higher, and combined with #1, may not be viable.
1. Most AI datacenter plans and valuation are not tied to subscriptions, but from a more vague promise of "AGI," so this isn't likely to pop the bubble IMO (even if it does happen)
2. Historical precedent holds that governments are more likely to suppress rates to spur the economy during wartime.
This is great. Showing only even hours would be less text, maybe as an option.
As I’m looking at it to plan late night activities it shows this mornings forecast where I kind of expect tomorrow’s. Maybe make a continual S curve that can be scrolled by touch? Maybe less elegant, maybe more?
They define the negative consequences for justice as “unequal distribution of harm, inaccessibility, and consumption of space, time, and resources”
Which fits with my understanding.
The ones mentioned in this thread all use Kiwix for off-line wikipedia, OSM for maps, Khan for educational videos. It looks like internet-in-a-box is aimed at working well on low-powered devices, whereas nomad expects beefy hardware and includes local AI. Not sure how WROLPi differs from internet-in-a-box.
Maybe it's like linux distros: all based on the same software, but optimized for different use-cases or preferences.
I mean, technically they use Kolibri for educational videos and exercises. A lot of them do come from Khan Academy, but we do a lot of work to make an offline first education platform, and also bring in a huge swathe of other open educational resources.
I think you’re right about the impetus for the piece. I’ll say that branding is for things consumed in public, I expect vendor lock in before branding. But as you know this site is awash in speculation about how the lack of differentiation will play out
Way down on my list of projects to vibe code is tags for HN users. I.e `Elon Stan` , `smart about aeronautics` , `grumpy` , `reasonable` etc etc. I like reading different opinion but if I formed an opinion about a user id like to record that without using my brain
I wish Twitter had this. It's always frustrating to read someone I've been following for a long time saying some bit of lore or unusual position that would change the whole way I think about them, except I know I won't be able to remember to associate it with their username next time I see them.
reply