Tata Semiconductor manufacturing cannot be used for AI
Tata Group JV with PSMC may roll out first semiconductor chip by 2026, but here's the catch - it's a 28nm chip, not the 3nm chip used for AI. In comparison, Apple had announced in 2023 that it will buy every 3nm chip that TSMC can make for the next-gen iPhones.
You cannot directly jump to a 3nm chip. We don't even have the tech to make it. And even if we did it will cost upwards of 60-80 billion to even start. And by the time it completes the world would have moved to better chips.
Could you please share your source mentioning the cost to be 60-80 Bn?
Coz it says that 3nm would cost $15-20 Bn.
https://www.granitefirm.com/blog/us/2023/04/29/cost-of-chip-foundry/
Also we don't have the tech to make it - hence the JV
Haha thanks for fact checking. I didn't mean the cost for only the 3nm chips. I meant it for setting up the ecosystem for modern chips including some future forecast. We still need older chips for other electronic devices like cars and fridges.
See more comments
Something is better than nothing.
Just google about the cost of creating a 3nm fab.
Know about EUV lithography tools cost which asml sells ? Asml mcap u know??
Even usa and china don't have one.
After 28nm, we will do finfets as well.
1st let planar tech stabilize.
Ah, lemme take an interesting story.
When XUV700 was launched, people were all excited, “damn! such a posh car..blah blah”. Sales were off the charts, waiting periods were almost 2 years. People were like - This is the best Indian car ever, this will rule the markets. Down the line, Scorpio got re-launched, it was not so fancy but can do all terrain hunts. Some may complain, highway driving isn’t that smooth with scorpio but it all depends on the driver and surface.
So, here we are 28nm is scorpio which can be used for all kind of terrains.
Infact, a SoC has multiple technology nodes in a chip. Memories (sram, dram) being fastest (3/5-nm) and other logic devices in higher nodes to balance the performance & costs.
After telling all these (boring) stories. India will mainly start with OSAT which is mainly on assembly & test. 28-nm has wide number of applications and save insane $$. The whole EV ecosystem relies on mature technology nodes as space isn’t a constraint. You will be surprised to know only few OEMs can afford 3-nm technology (mainly mobile fone manufacturers).
Happy Learning!
Tata's 28nm chip? A dinosaur in the AI world. Apple's already on 3nm. We're playing catch-up, not leading. What's our next move?
Tata's 28nm chip, a baby step, not a moonwalk. But remember, even Rome wasn't built in a day. So, what's your next big idea?
Discover More
Curated from across
Software Engineers on
by Kendall Hyrum
TCS
Coding will be dead soon
But this isn't the first time a tech exec has predicted the death of coding.
https://www.tomshardware.com/tech-industry/artificial-intelligence/jensen-huang-advises-against-learning-to-code-leave-it-up-to-ai