Hacker News Reader

Comment by 7thaccount | Hacker News Reader

Do they actually plan this, or is it just to make the current administration happy? $60B sounds like a lot of money.

They’ve steadily expanded over the years, I wouldn’t be surprised if this was real. It’s not just calculators, TI chips show up in a lot of military hardware in particular.

There are a lot of markets where TI is king, such as in battery management ICs.

which generally cost pennies.

If you want to build a fleet of infantry drones and you want to source components from the US there’s a lot of these and many others to buy. Not sure if there’s $60B of them there but they’ll figure out how to waste this much I’m sure.

Well if you look at how Feds has being pumping money into Musk's bank account and what brought down the SVB, then $60B may not be that much. At this point as the country that can prints/(adjust db ledger) $60B is nothing.

Which is why they are not planning to make cutting edge semiconductors in the new plant but "foundational semiconductors" aka old nodes that now cost pennies

These have super low profit margins.

US military may be willing to pay more for chips made in the US

They apparently make enough money for TI to have continued investing, over many years, in R&D and manufacturing of those devices.

Not really?

The r&d isn't expensive because these are legacy/old type chips. There isn't much r&d happening. They're made using old processes.

Any r&d is likely using different materials. Military stuff often uses different materials for specific applications.

The profits come from raising yields, which is much easier on older nodes.

They make really good VCO+PLL chips. Ultra low phase noise.

DSPs as well

It's 1/3rd the valuation of the company... its not real lol

I just want them to make a newer version of the TI-99/4A.

Wouldn't it be funny to see that CPU produced on a modern node?

Given the low transistor count, you could probably achieve speeds in the multi-10s of GHz, it would be so small.

A multi-core 99/4A would be hilarious to witness.

With MESI cache coherence, maybe you could migrate the whole workspace for your subroutine into a line of your core's L1D cache, and make it perform like hardware registers while retaining the pleasantly parsimonious TMS 990 architectural semantics?

Their announcement seems to be a mix of new and already announced expansions.

They’re making a political advertisement to get favors for things they already did. The meat of the announcement is that the plants started under the Biden administration are continuing as planned – which is an accomplishment of sorts given how complex fabs are – and they’d really like to continue getting the CHIPS act funding which Trump periodically talks about cancelling because it was a Biden-era law even though it had strong support from Republicans who favor investing in important domestic industries, especially in their districts:

https://www.texastribune.org/2025/03/12/texas-congress-corny...

Based on every one of these promises ever, yes. It's probably somewhere between vaporware and a small tertiary facility.

What are some things TI promised on, but failed to deliver?

https://spectrum.ieee.org/the-inside-story-of-texas-instrume...

Number of new jobs for one. They just laid off ~300 people working on an older node in Lehi, Utah, while now claiming they’re going to bring new jobs to Lehi, Utah. Thus “new jobs” don’t really track the actual change in the number of their employees even just at the local level across even moderate timescales.

Similarly hyping up total R&D spending without a tight timeframe is meaningless, they will eventually spend X$ assuming the company survives indefinitely.

More charitably there’s meaningful differences between nodes, but it’s likely a significant portion of the workforce will be rehires.

Reply on news.ycombinator.com