Dec 14, 2023 · 22m · big-technology
Intel's CEO Shares His Plan To Win The AI Chip War — With Pat Gelsinger
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this interview with Alex Kantrowitz, Intel CEO Pat Gelsinger outlines the company's comprehensive strategy to win the AI chip race through hybrid CPU-GPU architectures, dedicated Gaudi accelerators, local AI PCs, and a massive revitalization of its contract foundry business.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Alex holds 25.4% of the talking time here. How this is scored →
Alex as informed peer 4.3 Guest teaching 4.7 Guest disagreement 1.5 Alex pushing back 2.0
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 12:24 Gelsinger corrects Alex on foundry profitability
When Alex suggests foundry manufacturing is inherently the least profitable part of the chip business, Gelsinger directly counters, noting TSMC makes massive profits.
Hardest push from Alex ▶ 6:34 Alex presses on CPU vs GPU AI capabilitiesAlex challenges Gelsinger's claim that AI inferencing will run on CPUs, highlighting the apparent contradiction with the earlier architectural explanation.
Biggest teaching moment ▶ 3:19 Gelsinger clarifies training vs inferencing economicsGelsinger uses a weather model analogy to educate Alex on the distinct hardware requirements and market dynamics separating AI training and inferencing.
Alex holds their own ▶ 2:39 Alex cites custom silicon developments at Google and AmazonAlex demonstrates strong sector expertise by referencing Google's Gemini training on in-house TPUs and Amazon's proprietary accelerators.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| Understanding the Fundamental Differences Between CPUs and GPUs | 3 | 6 | 1 | 1 | Alex asks foundational questions comparing CPUs and GPUs and Nvidia's dominance. Gelsinger provides technical explanations and an illustrative sedan vs. F1 car analogy. | |
| The Evolving AI Accelerator Landscape and Intel's Dual Strategy | 5 | 4 | 2 | 2 | Alex demonstrates industry knowledge citing Google's Gemini training on TPUs and Amazon's accelerators. Gelsinger breaks down the difference between training and inferencing workloads and outlines Intel's foundry strategy. | |
| Integrating AI Capabilities Across Xeon CPUs and Gaudi Accelerators | 4 | 5 | 2 | 3 | Alex presses Gelsinger on how CPU chips can run AI inferencing effectively if GPUs are inherently better suited. Gelsinger clarifies that embedding matrix functions into standard CPUs handles mixed enterprise workloads efficiently. | |
| Evaluating Intel's Market Share Growth and Supply Crunch Dynamics | 5 | 4 | 2 | 3 | Alex asks for an honest assessment of market share against Nvidia and questions Intel's historical foundry failures. Gelsinger admits previous efforts were non-serious 'hobbies' and outlines how current standardization and geopolitical supply crunches change the equation. | |
| Geopolitical Vulnerabilities, Fab Construction, and the CHIPS Act | 5 | 5 | 1 | 2 | Alex asks why fab construction takes so long and cites historical Texas Instruments decisions regarding offshore profitability. Gelsinger explains the technical scale of $30B mega-projects and how the CHIPS Act offsets structural cost disadvantages. | |
| Assessing Apple Silicon and Foundry Opportunities | 4 | 4 | 1 | 1 | Alex asks about Apple Silicon and Intel's upcoming product lineup. Gelsinger discusses pitch meetings with major tech firms and envisions the next generation of 'AI PCs.' |
Statements from this episode (15)
Gelsinger: GPUs are uniquely suited for AI due to matrix processing
“A GPU, instead, is really built for a very specific class of workloads, and generally those have been called throughput workloads, so it does lots of floating point processing and matrix operations, You know, and so it's very dedicated for things like graphics…”
Gelsinger: CPUs are sedans while GPUs are Formula 1 racecars
“So it's designed basically, you can sort of think about it, you know, as your general purpose sedan, that's sort of the CPU and the GPU. All it does, it gets on the F one track and all it does is go fast on very specific workloads.”
Gelsinger: Nvidia got lucky on AI because AI workloads resemble graphics
“Yeah, and it very much is that way, and Jensen and I, you know, we've known each other for 35 years, you know, this general purpose workload, and we always are adding more capabilities to the CPU, but over here, it was always just go really fast for graphics, …”
Gelsinger: Intel aims to manufacture chips for Nvidia, AMD, and hyperscalers
“You know, we're also going to be a foundry. We're going to be the manufacturer for many of those chips as well. So we want to be the manufacturer for Nvidia, for AMD, for Google, for Amazon. We want to be their manufacturing partner, even if we're not using ou…”
Gelsinger: Intel, AMD, and Nvidia will be AI's three big merchant providers
“There's going to be those that build their own, you know, and that's what you see Amazon, Microsoft, Google are doing. They're going to say, hey, I'm going to own this and do this myself. And then there's going to be the general providers in the marketplace, a…”
Gelsinger: About a third of Intel data center CPU purchases are for AI
“And for my standard CPUs today in the data center, you know, we see about a third of the purchases are being based For AI workloads.”
Intel plans to merge Ponte Vecchio and Gaudi into one product line
“And Ponte Vecchio and Gaudi, we're bringing those together into a single product line, you know, going forward, because we're going to compete in that space as well.”
Gelsinger: Intel's AI business is approximately doubling growth rate quarter-over-quarter
“We're now seeing our growth rate and quarter to quarter, we approximately doubled the growth rate, you know, but we're still small market share, you know, today, but we're rising quickly because customers are looking for alternatives, you know, today because t…”
Gelsinger: Foundry manufacturing is almost as profitable as chip design
“So this is now a very profitable business. You know, almost as profitable as the chip business itself in many respects”
Gelsinger: Intel standardized its foundry processes to match the industry
“And Intel before was very proprietary. So if you wanted to use my foundry, you had to be proprietary on me. Well, now we have standardized our processes like the rest of the industry. So it's much easier to use us as a foundry.”
Gelsinger: Global chip production will reach 50/50 East-West balance by 2030
“By the end of the decade, so, you know, seven, eight years, I think we can get close to fifty-fifty by the end of the decade, and if we accomplish that, right, over a seven or eight year period, I think the world is going to sleep much better at night”
Gelsinger: Leading-edge chip fab costs $30B and takes five years
“You know, it takes us about five years to have one of these factories up up and running on a leading edge process technology. You know, the total project is about thirty billion dollars, right, to build one of these factory complexes.”
Gelsinger: Chip fab locations are more geopolitically vital than oil reserves
“Where oil reserves are has defined geopolitics for 50 years. Where technology and fabs are for the future is more important.”
Gelsinger: Apple built in-house silicon because Intel stumbled
“Well, you know, they used to use Intel chips and when Intel stumbled Apple stepped in and did their own chips.”
Gelsinger: AI PCs will replace typing with autonomous voice interactions
“We see this AI PC Having that same kind of shift where all of a sudden, maybe I don't type to my computer anymore. I just talk to it in the future. It knows when I'm there. It translate languages. You know, it has new insights and capabilities. It becomes my p…”