Asianometry: Intel Should Second-Source Nvidia

[Brady]

Generally this video explores the hypothetical of: what if — like how IBM forced Intel to give AMD an x86 license so that IBM wouldn’t get stuck with a single supplier, successfully creating a competitor in the market — the USG forces Nvidia to give Intel (and realistically I bet it would require AMD, too) a “Tensor core and CUDA” license.

I think it’s a really interesting idea. Like with x86 it has direct precedent with success. It also creates a new double-digit billion dollar revenue stream for Intel if successfully rolled out, helping Intel get the cash it needs without taxpayer bailouts and also spreads Nvidia’s historic (and unsustainable) 80+% GM to the wider industry players.

Seems too smart for government (and their lobbyists puppeteers) to allow to happen. Just fun to think about.

 

[tooLongInEDA]

Usually an excellent and well-informed channel. But the host admits this is essentially an interesting brainstorming thought that came into his head rather than something he's really thought through.

He never really explains what's in this for nVidia - just why they should give up their IP and dominant market position for nothing in return ("nothing" being the theoretical right to produce x86 devices, which they almost certainly neither want nor need to do).

Nor does he ever examine exactly what it means to "second source" in the complex world of modern chips and their associated software and tooling ecosystems. Is it even practically possible to produce what we would historically regard as a true second source for something with this degree of hardware and software complexity ? But I suspect he really means cross-licencing rather than true second sourcing.

On the @Brady comment above, if nVidia's 80+% gross margin is "unsustainable", then the "problem" here is going to sort itself out anyway without this sort of invervention.

The fact that so many of us are thinking and talking - more generally - about these sort of market distorting and crony capitalism measures without much thought for the unintended consequences really ought to be telling us something.

 

[hist78]

Usually an excellent and well-informed channel. But the host admits this is essentially an interesting brainstorming thought that came into his head rather than something he's really thought through.

He never really explains what's in this for nVidia - just why they should give up their IP and dominant market position for nothing in return ("nothing" being the theoretical right to produce x86 devices, which they almost certainly neither want nor need to do).

Nor does he ever examine exactly what it means to "second source" in the complex world of modern chips and their associated software and tooling ecosystems. Is it even practically possible to produce what we would historically regard as a true second source for something with this degree of hardware and software complexity ? But I suspect he really means cross-licencing rather than true second sourcing.

On the @Brady comment above, if nVidia's 80+% gross margin is "unsustainable", then the "problem" here is going to sort itself out anyway without this sort of invervention.

The fact that so many of us are thinking and talking - more generally - about these sort of market distorting and crony capitalism measures without much thought for the unintended consequences really ought to be telling us something.

I watched this video last night. I think that if Intel can't solve its design and manufacturing problems, having the rights to use CUDA would be meaningless.

On the other hand, if I can buy authentic Nvidia products, why would I buy an authorized Nvidia clone from Intel—unless it's significantly cheaper than Nvidia’s? But in that case, it wouldn’t provide higher margins to support Intel’s design and Intel Foundry.

 

[delong.height]

I watched this video last night. I think that if Intel can't solve its design and manufacturing problems, having the rights to use CUDA would be meaningless.

On the other hand, if I can buy authentic Nvidia products, why would I buy an authorized Nvidia clone from Intel—unless it's significantly cheaper than Nvidia’s? But in that case, it wouldn’t provide higher margins to support Intel’s design and Intel Foundry.

This pushes Moore's Laws further. And technically, AMD can offer better GPU than Nvidia but failed on software/ecosystem. Whereas Cuda is only getting better with its strong ecosystem. But imagine if the world is divided into Nvidia, AMD, and Intel. Each having their own Cuda. With Nvidia being so strong, does anyone ever want to go to AMD or Intel and begin the ecosystem on their platform from scratch with little to none community?

He never really explains what's in this for nVidia - just why they should give up their IP and dominant market position for nothing in return ("nothing" being the theoretical right to produce x86 devices, which they almost certainly neither want nor need to do).

Did intel get any return for sharing x86 with AMD?

 

[hist78]

Did intel get any return for sharing x86 with AMD?

It’s a cross-licensing deal between Intel and AMD with no payment involved. The two important components are Intel’s x86-related technologies and AMD’s x64-related technologies.

 

[siliconbruh999]

Did intel get any return for sharing x86 with AMD?

Nope if anything AMD has benefited more from it than Intel being able to use the SW and ISA Extension without doing much work. The only Contribution AMD made was X64

 

[Daniel Nenni]

First off, why did he not reference discussions on SemiWiki? Second, he did not mention how TSMC got to where they are today. Third, he did not mention China's push into semiconductors.

I don't know what readership Reddit's semiconductor subgroups have but they are no SemiWiki. We had 1M active readers in 2024 and we have 1.4M thus far in 2025. I see traffic from Reddit when they post a SemiWiki article and I have never seen any meaningful conversations. Semiconductors are very complex and so are the people who make them.

If you want to compete with TSMC you have to follow TSMC's path to greatness. You have to work closely with the top tier fabless companies. Remember, TSMC has no design experience, that all comes from customer and partner collaboration. Nvidia and AMD are very close with TSMC, let's call it Taiwan close. Also, TSMC was partially funded by the Taiwan Government and still gets preferential treatment via tax incentives, subsidized loans, land for science parks, etc... TSMC also gets funding from foreign governments (US, Japan, Germany).

Now let's look at China. How much money has the Chinese Government invested in the semiconductor ecosystem over the last 20 years? It is hard to pinpoint exactly but I would guess it is in the 100s of billions of US dollars. I heard the number $200B when I was last there in early 2020. It must be closer to $300B now, if not more.

So, again, how is Intel going to compete with Taiwan and China? By licensing CUDA? Or by getting government assistance like TSMC and the China foundries?

 

[Paul2]

Now let's look at China. How much money has the Chinese Government invested in the semiconductor ecosystem over the last 20 years? It is hard to pinpoint exactly but I would guess it is in the 100s of billions of US dollars. I heard the number $200B when I was last there in early 2020. It must be closer to $300B now, if not more.

And most of that came to complete nothing. There is no other precedent of any state splurging so much money onto industry in history. In many ways, most of their entries never constituted a genuine attempts to begin with. A government company, endlessly unprofitable, endlessly subsidised is just a treasury siphoning apparatus for whomever gets the distributor license.

You can't talk about Chinese business without talking about Chinese politics. You will blind yourself to many things without which any credible analysis is impossible.

One of such things is that the recent "sudden improvement" in PRC's government's return on investments come directly from directors of state companies being forced to cough up money that they were previously freely appropriated through such schemes, and that comes directly from the fact that the PRC's government has much less free money now than it had during its 10% growth-per-year era.

So their "sudden improvement" is in effect because of their macro level picture is, in reality, getting worse.

I repeat again, if anyone has a free billion dollars, and can't effortlessly produce >10% ROE almost anywhere, and in any industry, there is something wrong not with the market, but the person, or organisation with that free billion dollars. And this is a reminder about that much of the world economy, and billion dollar deals in the world have nothing to do with genuine productive business, but with powerful entities/people moving money in between their vassals/dependent entities.

 

[samwilde]

Except China has already achieved that with its investment in solar. They own 80% of that market.

You ignore the externalities. In solar it is cheap electricity to end users. Who cares if the industry is in a perpetual bloodbath?

Cheap chips mean China will remain competitive in consumer electronics.

 

[Xebec]

Except China has already achieved that with its investment in solar. They own 80% of that market.

You ignore the externalities. In solar it is cheap electricity to end users. Who cares if the industry is in a perpetual bloodbath?

Cheap chips mean China will remain competitive in consumer electronics.

I'm not pro-China, but +1 for China for conquering green energy; they've also mastered LFP batteries better than anyone. They've got a good lock on cheap and scalable energy for the next 100 years.

 

[Paul2]

Except China has already achieved that with its investment in solar. They own 80% of that market.

You ignore the externalities. In solar it is cheap electricity to end users. Who cares if the industry is in a perpetual bloodbath?

Cheap chips mean China will remain competitive in consumer electronics.

And I also remember the time when Chinese government was dishing out three million dollars to any company with word "solar" in its title just for that, and of which most was likely fraud.

This however doesn't preclude that it wouldn't work even if 90% of that was the most plain fraud, and only there as a pretext for officially sanctioning siphoning of the treasury.

Some of that might have worked, and the rest was largely due to lack of anyone as serious on the market. As of now, out of hundreds of "solar" companies, there are only 4 cell makers of note left.

 

[Barnsley]

Except China has already achieved that with its investment in solar. They own 80% of that market.

You ignore the externalities. In solar it is cheap electricity to end users. Who cares if the industry is in a perpetual bloodbath?

Cheap chips mean China will remain competitive in consumer electronics.

So no truth in the story that all the solar companies want to stop the race to the bottom as none of them are making any money?

 

[Daniel Nenni]

And I also remember the time when Chinese government was dishing out three million dollars to any company with word "solar" in its title just for that, and of which most was likely fraud.

This however doesn't preclude that it wouldn't work even if 90% of that was the most plain fraud, and only there as a pretext for officially sanctioning siphoning of the treasury.

Some of that might have worked, and the rest was largely due to lack of anyone as serious on the market. As of now, out of hundreds of "solar" companies, there are only 4 cell makers of note left.

Solar reminded me of the DRAM race, not something the US was good at doing but clearly there was money in it if you waited the competition out. It took government funding of course.

I don't think you can compare the leading edge logic race to that but TSMC did win the war of attrition, with the help of government funding.

The China investment in the semiconductor ecosystem included a big bunch of fraud as well. I remember talking to China start-up companies who's business plan was to get China funding at each and every step and not necessarily to build the best products.

I did not hear fraud with Taiwan and the money they put into semiconductors. Hopefully the US can do it without fraud. Buying stock seems like a good enough path?

 

[MKWVentures]

I find it interesting how all these ideas are about how other companies should be regulated to save Intel. Those companies have looked at Intel foundry and chosen, repeatedly, not to use them .... for a variety of reasons. The tech world is not looking to save Intel. If Intel cannot be successful, they will move on and others will grow.

And the solution is simple. Intel products for all its woes makes money and is a leader. Intel foundry does not have a viable business plan to survive. I think LBT will make the correct decision.

 

[Daniel Nenni]

I find it interesting how all these ideas are about how other companies should be regulated to save Intel. Those companies have looked at Intel foundry and chosen, repeatedly, not to use them .... for a variety of reasons. The tech world is not looking to save Intel. If Intel cannot be successful, they will move on and others will grow.

And the solution is simple. Intel products for all its woes makes money and is a leader. Intel foundry does not have a viable business plan to survive. I think LBT will make the correct decision.

What is the correct decision?

Trust is key for the foundry business because fabless companies are literally trusting foundries with their livelihood. Pat Gelsinger's over optimistic rhetoric did not inspire confidence nor trust. Lip-Bu is a very different type of leader:

"I have always operated within the highest legal and ethical standards. My reputation has been built on trust – on doing what I say I’ll do, and doing it the right way. This is the same way I am leading Intel." Lip-Bu Tan.

 

[Paul2]

I did not hear fraud with Taiwan and the money they put into semiconductors. Hopefully the US can do it without fraud. Buying stock seems like a good enough path?

Because the government disowned most of the industry by the early nineties, and was no longer doing outright wealth transfers to Chiang family's favourites. And as life comforts went, Taiwan was already infinitely above PRC in income, and social progress. Most civil servants had comfy lives, while in PRC most bureaucrats had to steal to simply not to starve.

 

[Brady]

I find it interesting how all these ideas are about how other companies should be regulated to save Intel. Those companies have looked at Intel foundry and chosen, repeatedly, not to use them .... for a variety of reasons. The tech world is not looking to save Intel. If Intel cannot be successful, they will move on and others will grow.

No leading edge foundry will ever grow again without state backing. As you are well aware, the startup costs are astronomical (just ask Rapidus).

This is about saving America’s best bet at having a leading edge, American-R&D-driven foundry.

This video is an exploration of how forcing Nvidia to share their Tensor Core IP with Intel would be a (potentially! hypothetically!) effective and actually efficient way to fund Intel’s foundry by giving Foundry a new whale: Intel Products making Nvidia IP!

 

[Xebec]

If you want to compete with TSMC you have to follow TSMC's path to greatness.

Is there no disruptive way to challenge TSMC?

(Atomic Semi/peers, another novel way of building chips, building a world class packaging service without fabbing chips, etC)

 

[Daniel Nenni]

Is there no disruptive way to challenge TSMC?

(Atomic Semi/peers, another novel way of building chips, building a world class packaging service without fabbing chips, etC)

I think there is a disruptive way to challenge TSMC, in fact I am counting on it, but the first order of business is to get Intel and Samsung Foundry back on solid financial ground and engaged with the semiconductor ecosystem. Both companies have had missteps and trust needs to be reestablished moving forward. I have more confidence in Intel providing disruptive technology but Samsung is definitely the low price alternative which is also important.

Am I alone on this? Is TSMC untouchable?

 

[MKWVentures]

I think there is a disruptive way to challenge TSMC, in fact I am counting on it, but the first order of business is to get Intel and Samsung Foundry back on solid financial ground and engaged with the semiconductor ecosystem. Both companies have had missteps and trust needs to be reestablished moving forward. I have more confidence in Intel providing disruptive technology but Samsung is definitely the low price alternative which is also important.

Am I alone on this? Is TSMC untouchable?

We keep trying to find ways to make the Jets and Panthers competitive with the Chiefs.

Customers are free to bet on the Jets and Panthers to win the Superbowl. No one is stopping them. For some reason, they do not.

The Chiefs (Nvidia, Apple, TSMC) do not win because of government backing.