Xebec
Well-known member
+1 - AI5 has been delayed about an entire year from its original date.
Musk says Tesla's mega AI chip fab project to launch in seven days
- Thread starter Daniel Nenni
- Start date
hist78
Well-known member
6 days left now.
hist78
Well-known member
And the good news is that 6 days from today is not April 1st.
one of my startup 20 years ago closed door on 4/1. Of course, we all thought it is a management poorly placed jokeOh man, that would have been a good one!Exciting times.......
Japanese vendors are unlikely to fully back Elon Musk’s Terafab project, and the reason is simple: Panasonic. After overextending resources for the Gigafactory and accepting razor-thin margins, they felt "discarded" once Tesla pivoted to in-house battery tech. For Japanese firms, being treated as a disposable tool rather than a strategic partner is the ultimate Japanese business taboo. Without a foundation of mutual trust, the Japanese ecosystem will fully support Musk.
Especially in today’s environment of acute capacity shortages, the choice for suppliers is clear. If you are a Japanese vendor with limited output, would you prioritize a long-term, high-trust partner like TSMC, or a brand-new, volatile venture from Tesla
Especially in today’s environment of acute capacity shortages, the choice for suppliers is clear. If you are a Japanese vendor with limited output, would you prioritize a long-term, high-trust partner like TSMC, or a brand-new, volatile venture from Tesla
Last edited:
Xebec
Well-known member
FWIW, Panasonic is still a large chunk of Tesla's battery business, even today, and are probably selling Tesla more cells than when the JV formed.
Panasonic has also failed to keep up with the Chinese battery makers in terms of tech -- they have no LFP offerings for example (big mistake), nor do they seem to be keeping up with CATL in terms of charging performance for NCM / NCA batteries. Tesla had to go to Panasonic's competition for LFP, and in-house to try to advance tech beyond what Panasonic offers today (4680 dry process, and "battery as structure").
For all of his companies sans the rockets, he went with straight technology licensing, and tech transfer, not unlike what PRC companies did.
Battery manufacturing, and car making, and power electronics/solar making, and chip manufacturing are such niche businesses that just hiring few senior people off the labour market and hoping them to monkey something with endless money will never work.
He can license a few years old node with really enormous moneys, sure. He can try to hire a very senior grey hair exec, and try to give him full authority to build a semblance of a working fab business (SMIC way,) but that assumes very long runway, and determination to at most be No. 3-5. It worked for SMIC because they aimed to be TSMC from decades ago – mass market foundry without pretensions that targets cost above all.
Battery manufacturing, and car making, and power electronics/solar making, and chip manufacturing are such niche businesses that just hiring few senior people off the labour market and hoping them to monkey something with endless money will never work.
He can license a few years old node with really enormous moneys, sure. He can try to hire a very senior grey hair exec, and try to give him full authority to build a semblance of a working fab business (SMIC way,) but that assumes very long runway, and determination to at most be No. 3-5. It worked for SMIC because they aimed to be TSMC from decades ago – mass market foundry without pretensions that targets cost above all.
Last edited:
Mooredaddy
Well-known member
Let’s look at this logically. Let’s take into account all of Elons promises and how frequently they don’t come true? Let’s go through some of the big ones:
In July 2025 “I think we’ll probably have autonomous ride hailing in probably half the population of the U.S. by the end of the year.”
- idk who ever believed this one, come on
In summer 2025 “ thousands of Optimus robots made by the end of the year”
- no evidence of this whatsoever
In 2024 he said xAI would have full AGI in 2025
- wrong again
In the summer of 2025 “ most epic demo ever by the end of the year”
- no demo materialized
In 2025 he said that he is confident that Tesla will deliver “unsupervised full self-driving” in consumer vehicles by the end of the year
- Again no
In 2024 “foresees volume growth increasing by 20-30% in 2025“
- In reality every Tesla metric has deteriorated by the way. Revenue down, cars sold down, cars delivered down, yearly profit massively down from $20.85 billion in 2023 to just $4.8 billion in 2025.
I could go on, there are literally websites that track hundreds of wild claims he's made. I think its useful AND necessary to be skeptical and clear eyed when wild claims are made. We are all logic oriented people here and we should not be blindly fanboying over Elon tweets. Logically Tesla is declining in market share and financially as well. China is making cars with better build quality and at better price points. Teaming up with Intel as suggested, when they are also struggling doesn't exactly scream dream team to me. Further, Dojo was shut down last year because it was just getting stomped performance wise by AMD and Nvidia. I just don't see a reasonable scenario where a company that has zero experience fabricating semiconductors, teams up with Intel of all companies to all of a sudden mass produce chips at a TSMC scale. They dont have the expertise, they dont have the financials, and they don't have the experience. I truly think this is another case of Elon massively underestimating how difficult a task is, except this time its the most complicated things humans have ever constructed.
I see no way to go it alone. Just the unique materials supply chains alone requires loads of expertise he doesn't have. What to do about HP sulfuric acid in a domestic environment where TSMC AZ is shipping it in from Taiwan? Does Elon know he will need a partner to set up a hydrogen peroxide ion exchange purification plant next door to his? Too much to re-invent. Given the span of Elon's ambition includes packaging, I would suggest Intel is the only choice.