Former Intel CEO on Nvidia: You have to be 10x better to dethrone the king

[Daniel Nenni]

Now that Pat Gelsinger is no longer occupying the corner office at embattled chipmaker Intel, he can acknowledge one thing about the semiconductor industry.

Nvidia has a wide, wide lead over its rivals on the tech front.

"You see, they are executing well ... At the end of the day, [Nvidia CEO Jensen Huang] is on it, driving his teams to stay in front end," the former Intel CEO told me on Yahoo Finance's Opening Bid podcast.

"They have built meaningful moats around their franchise," he added.

Gelsinger led aggressive efforts to turn around Intel for more than three years. He slashed thousands of jobs, improved costs, secured CHIPS Act funding, built chip foundries, and promised fast AI chips that could compete with Nvidia and AMD.

He was fired in early December amid missed financial targets, lack of progress on the AI chip front, and a cash drain on the foundry business.

Intel's fourth quarter sales fell 7% year over year to $14.3 billion, and net earnings plunged 76%. The company forecasts it will only break even on the profit line this year.

Analysts think Nvidia may post upward of $5 per share in earnings in 2025. Nvidia's stock is up 1,220% over the past five years, while Intel's has dropped 69%.

Intel announced Lip-Bu Tan as its new CEO in March.

Tan was the CEO of Cadence Design Systems from 2009 to 2021. After serving on Intel's board for two years, he left in August 2024 following clashes with Gelsinger over how to position the business.

Tan frequently pushed for a better artificial intelligence strategy to take on Nvidia and faster decision making at the notoriously bureaucratic Intel.

It will be hard for someone to dethrone Nvidia, said Gelsinger.

"I've always viewed that there's sort of this 10x rule, where if you're not 10x better than the king, you're not going to displace that. Here you have to be at least sustainably 10x better, and then people will say, okay, yeah, I'm going to go invest in that," Gelsinger added.

Tan will get to outline his strategy publicly — and detail how he will navigate Trump tariffs — when the company announces first quarter results on April 24 after the market close.

https://www.yahoo.com/finance/news/former-intel-ceo-on-nvidia-you-have-to-be-10x-better-to-dethrone-the-king-171633066.html

 

[Daniel Nenni]

Back in the day the formula was 2x the performance at the same price or the same performance at half of the price. This was for CPUs and we rated them in MIPS (millions of instructions per second). I have no idea what Pat means by 10X better? Seems like unmeasurable executive level speak to me.

 

[hist78]

Now that Pat Gelsinger is no longer occupying the corner office at embattled chipmaker Intel, he can acknowledge one thing about the semiconductor industry.

Nvidia has a wide, wide lead over its rivals on the tech front.

"You see, they are executing well ... At the end of the day, [Nvidia CEO Jensen Huang] is on it, driving his teams to stay in front end," the former Intel CEO told me on Yahoo Finance's Opening Bid podcast.

"They have built meaningful moats around their franchise," he added.

Gelsinger led aggressive efforts to turn around Intel for more than three years. He slashed thousands of jobs, improved costs, secured CHIPS Act funding, built chip foundries, and promised fast AI chips that could compete with Nvidia and AMD.

He was fired in early December amid missed financial targets, lack of progress on the AI chip front, and a cash drain on the foundry business.

Intel's fourth quarter sales fell 7% year over year to $14.3 billion, and net earnings plunged 76%. The company forecasts it will only break even on the profit line this year.

Analysts think Nvidia may post upward of $5 per share in earnings in 2025. Nvidia's stock is up 1,220% over the past five years, while Intel's has dropped 69%.

Intel announced Lip-Bu Tan as its new CEO in March.

Tan was the CEO of Cadence Design Systems from 2009 to 2021. After serving on Intel's board for two years, he left in August 2024 following clashes with Gelsinger over how to position the business.

Tan frequently pushed for a better artificial intelligence strategy to take on Nvidia and faster decision making at the notoriously bureaucratic Intel.

It will be hard for someone to dethrone Nvidia, said Gelsinger.

"I've always viewed that there's sort of this 10x rule, where if you're not 10x better than the king, you're not going to displace that. Here you have to be at least sustainably 10x better, and then people will say, okay, yeah, I'm going to go invest in that," Gelsinger added.

Tan will get to outline his strategy publicly — and detail how he will navigate Trump tariffs — when the company announces first quarter results on April 24 after the market close.

https://www.yahoo.com/finance/news/former-intel-ceo-on-nvidia-you-have-to-be-10x-better-to-dethrone-the-king-171633066.html

Does this "10x better rule" mentioned by Pat Gelsinger also apply to Intel Foundry vs TSMC?

 

[tooLongInEDA]

Does this "10x better rule" mentioned by Pat Gelsinger also apply to Intel Foundry vs TSMC?

If it did, they have no chance.

But this is just such self-evident nonsense.

When - for example - was Intel ever even 2x better than AMD (or vice versa) ? I'd be surprised if the gap (on production cost or peformance) was ever much over 50% for more than a few months. 10x might be possible in exceptional circumstances in software. But in hardware ?

And Gelsinger's implication seems to be that you provide a 10x better product for much the same price as the existing one - i.e. giving most of the value away for nothing.

I hesitate to mention it, but the idea just crossed my mind that Gelsinger's public utterances seem to divide into a set that make quite a lot of sense and another that don't seem to at all ... and that this all reminds me a little of Trump ...

 

[Xebec]

Back in the day the formula was 2x the performance at the same price or the same performance at half of the price. This was for CPUs and we rated them in MIPS (millions of instructions per second). I have no idea what Pat means by 10X better? Seems like unmeasurable executive level speak to me.

10X sounds a high to me, but I think the 2X rule only holds if there is no significant moat other than "trust the leader". Also note this is a combined hardware/software market.

AMD has => 2X price/perf over Nvidia in at least some segments of the AI market and they don't seem to be grabbing market share from Nvidia. (MI300/X is typically <1/2 retail cost of H100, and on paper MI300 is stronger). AMD's software (perception/reality?) "ruins" the >2X hardware advantage.

As another analogy, Microsoft has been able to fend off all kinds of 2X better alternatives; some of which are "free".

 

[Xebec]

When - for example - was Intel ever even 2x better than AMD (or vice versa) ? I'd be surprised if the gap (on production cost or peformance) was ever much over 50% for more than a few months. 10x might be possible in exceptional circumstances in software. But in hardware ?

I think that's exactly it. What's the value of CUDA? It's far beyond the cost of the hardware.

 

[MAB]

10X sounds a high to me, but I think the 2X rule only holds if there is no significant moat other than "trust the leader". Also note this is a combined hardware/software market.

AMD has => 2X price/perf over Nvidia in at least some segments of the AI market and they don't seem to be grabbing market share from Nvidia. (MI300/X is typically <1/2 retail cost of H100, and on paper MI300 is stronger). AMD's software (perception/reality?) "ruins" the >2X hardware advantage.

As another analogy, Microsoft has been able to fend off all kinds of 2X better alternatives; some of which are "free".

I've heard this 10x rule in my schooling, but it was for dedicated acclerators. So, for example, if you were proposing a hardened accelerator for a use case and it required significant changes to the programming model, the only way you would even have a chance at adoption or people listening to you is if you were offering 10x better performance on a relevant vector.

 

[Daniel Nenni]

As another analogy, Microsoft has been able to fend off all kinds of 2X better alternatives; some of which are "free".

Microsoft did not fend well with LINUX and iOS. I still use Microsoft on my laptop but am more often on my iPhone or iPad and SemiWiki is in the cloud. To me it is the market wanting choices and not being boxed into one vendor.

Nvidia is in the same boat. People want alternatives but Jensen and company make that very difficult with both superior hardware and ecosystem. Not to worry though, even the mighty can fall (Intel) and become the alternative.

 

[BruceA]

what did it take Airbus to make a viable run at Boeing years ago, wasn’t 10x.

For anyone architected to displace full infrastructure one could argue you need significant advantage, nobody directly best x86. ARM ate them from behind and with volume and scale slowly ate the bottoms out, same play x86 put on mini computers etc.

For a second competitive foundry I would say within 5% and on par pricing should be enough.

Intel is misguided to be so obsessed with leadership. Focused should be good enough PPA and with complete and stable PDK. World wants and needs more leading edge capacity.

 

[hist78]

I

what did it take Airbus to make a viable run at Boeing years ago, wasn’t 10x.

For anyone architected to displace full infrastructure one could argue you need significant advantage, nobody directly best x86. ARM ate them from behind and with volume and scale slowly ate the bottoms out, same play x86 put on mini computers etc.

For a second competitive foundry I would say within 5% and on par pricing should be enough.

Intel is misguided to be so obsessed with leadership. Focused should be good enough PPA and with complete and stable PDK. World wants and needs more leading edge capacity.

For all the added cost and complexity in development and production with multiple foundries, 5% cheaper won't go too far.

 

[BruceA]

I

For all the added cost and complexity in development and production with multiple foundries, 5% cheaper won't go too far.

Supply continuity if possible is a no brained for the superscalers.

A magnitude 8 earthquake or super typhoon is bad news for Apple, Nvidia, AMD, Qualcomm, Broadcom, Intel, MediaTek, Facebook, Google, Microsoft, Tesla. It was surprising that it took covid for people to realize the risk in the supply chain.

 

[siliconbruh999]

Supply continuity if possible is a no brained for the superscalers.

A magnitude 8 earthquake or super typhoon is bad news for Apple, Nvidia, AMD, Qualcomm, Broadcom, Intel, MediaTek, Facebook, Google, Microsoft, Tesla. It was surprising that it took covid for people to realize the risk in the supply chain.

Intel is still the most risk averse of these

 

[Xebec]

Microsoft did not fend well with LINUX and iOS. I still use Microsoft on my laptop but am more often on my iPhone or iPad and SemiWiki is in the cloud. To me it is the market wanting choices and not being boxed into one vendor.

Nvidia is in the same boat. People want alternatives but Jensen and company make that very difficult with both superior hardware and ecosystem. Not to worry though, even the mighty can fall (Intel) and become the alternative.

Ironically - I wrote my initial reply on my iPad too as that's how I usually access Semiwiki . I fully agree with your strategy, I used to host an annual meeting with the software/infrastructure architects at the last company I worked at to brainstorm "How would we replace our top 3 major [software] suppliers".

However, Nvidia is like Microsoft in the respect that by the time someone is able to take significant market share in a valued segment, they've moved on and what's left is scraps by comparison.

The 1990s brought successful strong DOS alternatives, and then along came Windows 95. In 2018, Windows was demoted in the Microsoft leadership hierarchy* due to reducing importance of the OS. Microsoft's primary focus was now productivity, with the M365 ecosystem currently dominant in the corporate workplace. (Alternatives largely relegated to smaller businesses and consumers with shallow pockets.) Microsoft is executing for another play at this game with "OpenAI".

The reason you need more than 2X to defeat Nvidia, is because Nvidia will move the goalposts much further than expected when challenged. (GPUs --> GP GPUs --> CUDA --> Data Center --> AI Accelerators --> Next).


*https://stratechery.com/2018/the-end-of-windows/ -- "What is more interesting, though, is the story of Windows’ decline in Redmond, culminating with last week’s reorganization that, for the first time since 1980, left the company without a division devoted to personal computer operating systems "

 

[Xebec]

Supply continuity if possible is a no brained for the superscalers.

A magnitude 8 earthquake or super typhoon is bad news for Apple, Nvidia, AMD, Qualcomm, Broadcom, Intel, MediaTek, Facebook, Google, Microsoft, Tesla. It was surprising that it took covid for people to realize the risk in the supply chain.

Because this became religion in many industries: https://en.wikipedia.org/wiki/Lean_manufacturing

(Not that it's a bad concept, but resiliency = a true cost that competitors may not bear, and it's hard to convince shareholders/boards on the 'true cost' when supply chain fails).

 

[siliconbruh999]

Ironically - I wrote my initial reply on my iPad too as that's how I usually access Semiwiki . I fully agree with your strategy, I used to host an annual meeting with the software/infrastructure architects at the last company I worked at to brainstorm "How would we replace our top 3 major [software] suppliers".

However, Nvidia is like Microsoft in the respect that by the time someone is able to take significant market share in a valued segment, they've moved on and what's left is scraps by comparison.

The 1990s brought successful strong DOS alternatives, and then along came Windows 95. In 2018, Windows was demoted in the Microsoft leadership hierarchy* due to reducing importance of the OS. Microsoft's primary focus was now productivity, with the M365 ecosystem currently dominant in the corporate workplace. (Alternatives largely relegated to smaller businesses and consumers with shallow pockets.) Microsoft is executing for another play at this game with "OpenAI".

The reason you need more than 2X to defeat Nvidia, is because Nvidia will move the goalposts much further than expected when challenged. (GPUs --> GP GPUs --> CUDA --> Data Center --> AI Accelerators --> Next).


*https://stratechery.com/2018/the-end-of-windows/ -- "What is more interesting, though, is the story of Windows’ decline in Redmond, culminating with last week’s reorganization that, for the first time since 1980, left the company without a division devoted to personal computer operating systems "

Nvidia's Moat is mostly software AMD has better HW spec for the price but the incomplete software ecosystem is causing them issues.

 

[DougDD]

what did it take Airbus to make a viable run at Boeing years ago, wasn’t 10x.

For anyone architected to displace full infrastructure one could argue you need significant advantage, nobody directly best x86. ARM ate them from behind and with volume and scale slowly ate the bottoms out, same play x86 put on mini computers etc.

For a second competitive foundry I would say within 5% and on par pricing should be enough.

Intel is misguided to be so obsessed with leadership. Focused should be good enough PPA and with complete and stable PDK. World wants and needs more leading edge capacity.

BruceA, good points on how Airbus caught up (and eclipsed temporarily?) Boeing.
But, right now, BA still has a larger market cap - but barely.

And this nonsense about Intel needing being 10X better than NVIDIA could be one
of the reasons why Gelsinger got removed. I remember when Pat said in a quarterly
conference call last fall that Intel wasn't intent on competing with NVIDIA head-to-
head anymore - they would focus on enterprise AI, not rack scale. As an Intel shareholder,
I was taken aback with that comment. The market is so damn lucrative, why try to beat NVDA
by 10X? Just fully commit to being a great second source (or maybe third in the beginning
and then competing for second later) because the market is still big enough for that - and
growing which is most important. And I bet that's Lip-Bu Tan's strategy.

That's what Pat said was Intel's strategy in foundry - be #2 by the end of the decade in size
but #1 in process technology.

 

[XtorTourist]

Back in the day the formula was 2x the performance at the same price or the same performance at half of the price. This was for CPUs and we rated them in MIPS (millions of instructions per second). I have no idea what Pat means by 10X better? Seems like unmeasurable executive level speak to me.

It's an exaggerated business heuristic made popular by Peter Thiel in Zero to One. Pat's trying to be one of the cool kids (from 10 years ago).