Oct 17, 2024 · 34m · a16z
How to Build a Thriving AI Ecosystem with Lisa Su, CEO of AMD
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 episode of a16z's 'AI Revolution', AMD CEO Lisa Su joins Operating Partner Bob Swan to discuss the rapid expansion of generative AI compute, AMD's long-term high-performance chip strategy, open industry standards, and global supply chain resilience.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
The host as informed peer 3.1 Guest teaching 1.8 Guest disagreement 0.0 The host pushing back 0.0
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 7:33 Rejecting Board's Tablet Strategy Push
In the closest moment to friction in an otherwise harmonious chat, Su recalls refusing her board's recommendation in 2014 to build tablet chips, insisting AMD focus solely on high-performance computing.
Hardest push from the host ▶ 9:22 Questioning Single-Chip Differentiator DominanceSwan challenges the assumption that raw top-line chip performance will remain the sole market driver, forcing Su to explain how workload multimodality reshapes compute needs.
Biggest teaching moment ▶ 6:15 Demystifying the Pre- vs Post-ChatGPT AI LandscapeSu educates the host on how pre-generative AI was restricted to technical experts, whereas ChatGPT marked a paradigm shift by making natural language interaction universally accessible.
The host holds their own ▶ 12:49 Citing Industry Standards UALink and UECSwan demonstrates deep domain expertise as a former semiconductor CEO by introducing key emerging open networking consortiums like UALink and Ultra Ethernet.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| AI Revolution Series Title Card | 2 | 1 | 0 | 0 | Host Bob Swan opens with standard interview prompts regarding Su's career and the expanding role of compute. Su offers a high-level overview of tech evolution from PCs to cloud to AI, maintaining a completely collaborative tone. | |
| The ChatGPT Moment and Accessibility of AI | 2 | 3 | 0 | 0 | Su explains the shift from expert-only pre-generative AI to the accessible post-ChatGPT era. She shares a strategic historical anecdote where she rejected board pressure to make tablets in 2014 in favor of high-performance compute. | |
| Compute Across Form Factors and Workloads | 3 | 2 | 0 | 0 | Swan asks whether peak performance chips will remain the primary differentiator or if workload diversification will dominate. Su details the need for tailored compute engines across edge, client, and data center form factors. | |
| Industry Collaboration and Open Standards in Networking | 4 | 2 | 0 | 0 | Swan demonstrates semiconductor industry familiarity by referencing specific standards like UALink and the Ultra Ethernet Consortium. Su outlines semiconductor supply chain dynamics, 18-to-24 month lead times, and packaging bottlenecks. | |
| The Impact of the CHIPS Act and Government Policy | 4 | 1 | 0 | 0 | Swan lays out the specifics of the CHIPS Act authorization figures. Su strongly endorses the policy for prioritizing domestic supply resiliency and R&D, while Swan voices complete agreement. | |
| Balancing Long Cycle Hardware Bets with Fast Innovation | 3 | 2 | 0 | 0 | Swan prompts on managing long hardware development cycles against rapid software innovation. Su reflects on internal debates surrounding Moore's Law and taking multi-year risks on chiplets and 3D packaging. | |
| Customer Collaboration and Problem-Solving | 3 | 2 | 0 | 0 | Swan notes AMD's dramatic rise from under 1% server share to major hyperscaler adoption. Su outlines her customer-centric product strategy and the criticality of consistent roadmap execution. | |
| The Fabless Model and Foundry Partnerships | 4 | 2 | 0 | 0 | Swan brings up the historical GlobalFoundries spinout and fabless operations. Su details how AMD maintains 5-year strategic forecasting with primary foundry TSMC for advanced nodes and packaging. | |
| Strategic M&A, Xilinx, and System-Level Solutions | 3 | 2 | 0 | 0 | Swan highlights AMD's strategic acquisitions including Xilinx and ZT Systems. Su explains that ZT Systems provides system-level cluster solutions, forming the third pillar alongside hardware and software. | |
| The Role of Startups in the AI Semiconductor Ecosystem | 3 | 1 | 0 | 0 | Swan asks how large companies partner with AI startups. Su praises startup agility and jokingly pitches AMD GPU availability, leading to a lighthearted closing exchange. |
Statements from this episode (15)
Lisa Su: AI is the most important technology shift in 50 years
“I think AI is probably the most important one I'd like to say, you know, over the last 30 years, 40 years, 50 years because it It's something more than just technology. I mean, it really is, you know, AI becomes the ability for us all to become smarter, more p…”
Lisa Su: AI will be integrated into every product AMD builds
“And, you know, for us at AMD you know, my belief is AI is gonna be everywhere in every product that we build”
Lisa Su: AI hardware will diversify into specialized engines
“If you look, whether you're at the edge you know, with embedded applications industrial applications, automotive applications medical applications, or you're even at the client level in your PC or your phone, you're gonna need different types of AI, and so you…”
Lisa Su: Closed tech ecosystems usually become a problem over time
“Closed walls usually end up being a problem.”
Lisa Su: UALink and UEC prove AI competitors can unite around open standards
“The Ultra Accelerator Link and Ultra Ethernet Consortium are great examples where competitors and peers can come together and say, you know what, we're gonna adopt open standards and we're each gonna innovate on top of that, and you know, those are, you know, …”
Lisa Su: Adding semiconductor manufacturing capacity takes 18 to 24 months
“It takes 18 months, 24 months to really put that all on board.”
Lisa Su: Semiconductor industry prioritizes supply chain resilience over cost
“This idea of hey, let's try to just eke out every last penny in the supply chain is, has kind of gone a little bit of ways to let's build resilience into the supply chain.”
Lisa Su: CHIPS Act has made a huge difference for US semiconductors
“Some people say, hey it's not enough or, you know, it doesn't make a difference. I think it's made a huge difference. It's made a huge difference because what it's What it's really done is it's put at the top of the priority list resiliency and semiconductor b…”
Lisa Su: Moore's Law continues because people find new ways to extend it
“And by the way, I'm a believer in, in, you know Moore's law has been extended so many times because people are super smart and able to come up with different ways of extending the same principle of, you know, more transistors, more capability every couple of y…”
Lisa Su: Hardware bets take 3 to 5 years to prove out
“It, we didn't know. Like, we didn't know at the time when we were making that decision was it gonna be the right bet, but we knew that we had to make that bet, and you really don't figure that out until three to five years out, so your question about how do yo…”
Lisa Su: Solve 3-to-4-year customer problems rather than verbatim feature requests
“And that's where it's most beneficial is, you know, talking to our customers about, hey, like, what problems are you having? Like, what are you trying to solve? You know, two, three, four years out. And then, you know, our technologists can really come up with…”
Lisa Su: AMD held 1% server market share at Zen launch
“When we started in the hyperscaler market with our, you know, first generation products that you know, our Zen product portfolio, I think we were about maybe one percent share of the server market”
Lisa Su: Chip complexity has moved to advanced multi-chip packaging
“It's even more complicated because it's not just about silicon. It's about packaging and, you know, really how do we put these you know, these chips together for you know, very complex, multi-node you know, multi-chip, you know, type things.”
Lisa Su: ZT Systems acquisition adds system-level solutions to AMD
“That's what ZT Systems is, so it's kind of the third leg of our stool, if you think hardware, software, now solutions.”
Lisa Su: Large enterprises are increasingly open to working with AI startups
“I think even large, large enterprises who typically used to be, let's call it much more conservative in working with startups are also, becoming much more open because again, this is back to the disruption I talked about. Nobody wants to be behind in AI.”