This makes sense for chips that integrate sensing and processing on the same chip. The military tends to need things like that. LIDAR sensors need both sensing and fast counting, which often use different semiconductor technologies. The first flash LIDAR chips had an InGaAs sensor chip and a CMOS counting chip, both ball grid arrays, soldered back to back. If different technology layers on the same chip can do the job, that would simplify the parts.
It doesn't just simplify, it improves capability and saves power because it makes it much easier and cheaper in power to build an extremely wide interface between them.
Wouldn’t any form of government funding produce a lazy company vs companies standing on their own two feet with the free market acting as drill sergeant?
Suppose you have to do some R&D in order to make something happen. There is no way to keep it a secret and China isn't going to enforce a patent on it, so if you pay to do the R&D and then have to charge prices high enough to recover it, they undercut you on price and you go out of business. But if the government covers the R&D then you can do domestic production at a competitive price.
Meanwhile the subsidy should be going to every company in the industry so then they still have to compete with each other.
Or to put it a different way, what's really the difference between a subsidy and a tax cut?
Chip stacking today is limited by thermal management. Can't stack multiple high TDP (e.g. logic) chips together, can only do one + low power chips (e.g. memory). Wonder if this is something DARPA is trying to solve here.
I was curious if this was just via tax breaks and whatnot, but apparently not.
If my light research is correct, Texas founded its own CHIPS act in 2023, and shortly after appropriated 552 mil to the UT TIE program. The DARPA money commitment came a year later.
> I was curious if this was just via tax breaks and whatnot, but apparently not.
People always try to distinguish these as if they're completely different, but a refundable tax credit and a subsidy are just two different terms for the same thing.
In essence, you're right. To a sufficiently large company, absolutely.
But there'd be a huge difference between giving me a 500mil investment vs giving me 500mil in tax breaks.
Obviously most fall into the former camp, but it's also optics. One feels like taking money from taxpayers and giving it out, the other less so. It's a little more bold to hand out money than to just not collect it IMO.
Refundable tax credits are the thing invented to erase the distinction. That's the "refundable" part. If you would have paid $100 in taxes and you got a $1000 refundable credit then you're paying -$900 in taxes and the government sends you a "tax refund" check.
Which is effectively the same thing that happens even with non-refundable credits for large conglomerates, because they were paying enough in taxes from their other business units to eat the whole thing, while they're still consuming other government services. In order to make that fair to smaller companies and individuals, someone with a non-refundable credit that they couldn't use should be able to sell it to an unrelated business who could use it. But then it would just sell for approximately its face value and affect the government budget in exactly the same amount as if it was a refundable credit, which is why tax credits should always be refundable.
If we are thinking about world models and embodied AI as the next big wave, I feel we need to factor the size of this market across military and civilian applications. When I hear Tesla AI4-6 plans for autonomous driving and Optimus, Apple’s home robot, and many such initiatives, I’m also thinking of an autonomous military drone arms race and autonomous space exploration and moon colonization. Edge compute will be a key innovation and I feel Apple is gonna be best placed to exploit their power efficient M-series architectures.
Battlefield robots already exist. They just don't look like you and me. And mostly they fly. That word "embodied"is a broad hint at misunderstanding. Robot surgeons don't stand on humanoid feet next to the operating table, and they don't have humanoid "hands."
Boston Dynamics has had humanoid robots doing parkour for years now and nobody needs a robot to do parkour. This was also a hint.
Do we need to distinguish between Atlas, the control theory based humanoid tech stack vs say Optimus that may run on an end-2-end visual reasoning world model? I feel these are two different paradigms. Same for military drones. The next gen prototypes may have launch-and-forget levels of autonomy, however ethically questionable.
This makes sense for chips that integrate sensing and processing on the same chip. The military tends to need things like that. LIDAR sensors need both sensing and fast counting, which often use different semiconductor technologies. The first flash LIDAR chips had an InGaAs sensor chip and a CMOS counting chip, both ball grid arrays, soldered back to back. If different technology layers on the same chip can do the job, that would simplify the parts.
It doesn't just simplify, it improves capability and saves power because it makes it much easier and cheaper in power to build an extremely wide interface between them.
Skywater Technology offers similar services as well. Competition is good but does the US risk funding too many subscale entities?
https://www.skywatertechnology.com/heterogeneous-integration...
The US will fund as many people who will develop this tech as need be. The national champion approach produces a too big to fail lazy company.
Wouldn’t any form of government funding produce a lazy company vs companies standing on their own two feet with the free market acting as drill sergeant?
Suppose you have to do some R&D in order to make something happen. There is no way to keep it a secret and China isn't going to enforce a patent on it, so if you pay to do the R&D and then have to charge prices high enough to recover it, they undercut you on price and you go out of business. But if the government covers the R&D then you can do domestic production at a competitive price.
Meanwhile the subsidy should be going to every company in the industry so then they still have to compete with each other.
Or to put it a different way, what's really the difference between a subsidy and a tax cut?
The funding can get companies spun up and competing.
Assuming they are picking winners. Western government's usually just throw money at existing legacy organizations, not up and coming companies.
In this case it's an institute attached to a big university (University of Texas) that is scaling up an R&D idea, that happens to be useful for the military https://www.statesman.com/story/business/2024/07/18/semicond...
Chip stacking today is limited by thermal management. Can't stack multiple high TDP (e.g. logic) chips together, can only do one + low power chips (e.g. memory). Wonder if this is something DARPA is trying to solve here.
TFA did mention associated microfluid cooling research.
This sounds like a cool initiative, but I am curious where Texas is coming up with the $552M for this investment.
I was curious if this was just via tax breaks and whatnot, but apparently not.
If my light research is correct, Texas founded its own CHIPS act in 2023, and shortly after appropriated 552 mil to the UT TIE program. The DARPA money commitment came a year later.
> I was curious if this was just via tax breaks and whatnot, but apparently not.
People always try to distinguish these as if they're completely different, but a refundable tax credit and a subsidy are just two different terms for the same thing.
In essence, you're right. To a sufficiently large company, absolutely.
But there'd be a huge difference between giving me a 500mil investment vs giving me 500mil in tax breaks.
Obviously most fall into the former camp, but it's also optics. One feels like taking money from taxpayers and giving it out, the other less so. It's a little more bold to hand out money than to just not collect it IMO.
Refundable tax credits are the thing invented to erase the distinction. That's the "refundable" part. If you would have paid $100 in taxes and you got a $1000 refundable credit then you're paying -$900 in taxes and the government sends you a "tax refund" check.
Which is effectively the same thing that happens even with non-refundable credits for large conglomerates, because they were paying enough in taxes from their other business units to eat the whole thing, while they're still consuming other government services. In order to make that fair to smaller companies and individuals, someone with a non-refundable credit that they couldn't use should be able to sell it to an unrelated business who could use it. But then it would just sell for approximately its face value and affect the government budget in exactly the same amount as if it was a refundable credit, which is why tax credits should always be refundable.
Interesting. I didn't know that was how they worked, so thanks for explaining.
Looks like it's part of a larger effort to invest in semiconductors: https://gov.texas.gov/business/page/texas-chips-office
The Texas state government's annual budget is well over $300 billion, so it's not a huge amount of money for them.
$500M is always a huge amount of money.
If we are thinking about world models and embodied AI as the next big wave, I feel we need to factor the size of this market across military and civilian applications. When I hear Tesla AI4-6 plans for autonomous driving and Optimus, Apple’s home robot, and many such initiatives, I’m also thinking of an autonomous military drone arms race and autonomous space exploration and moon colonization. Edge compute will be a key innovation and I feel Apple is gonna be best placed to exploit their power efficient M-series architectures.
Battlefield robots already exist. They just don't look like you and me. And mostly they fly. That word "embodied"is a broad hint at misunderstanding. Robot surgeons don't stand on humanoid feet next to the operating table, and they don't have humanoid "hands."
Boston Dynamics has had humanoid robots doing parkour for years now and nobody needs a robot to do parkour. This was also a hint.
Do we need to distinguish between Atlas, the control theory based humanoid tech stack vs say Optimus that may run on an end-2-end visual reasoning world model? I feel these are two different paradigms. Same for military drones. The next gen prototypes may have launch-and-forget levels of autonomy, however ethically questionable.
Isn't Optimus the robot made by that guy who thinks it's going to be a "fantastic surgeon?"
Humanoid robots are a joke to pump stocks. Don't fall for the hype.
Because America can no longer guarantee the safety of Taiwan or bear the risk of having TSMC shutdown?