Ben Thompson · Jun 6, 2024
Ben Thompson interviews AMD CEO Lisa Su about solving hard problems
Markdown · JSON · Original source
Published June 6, 2024, Ben Thompson’s Stratechery interview with AMD CEO Lisa Su connects difficult engineering work to business outcomes. Su says the hardest part of delivering a product can come after the initial design: achieving yields, reliability and production at scale. She describes AMD’s strategy as building on its and capabilities, choosing markets suited to high-performance computing, and executing a roadmap over years. Su presents chiplets as a deliberate bet on higher performance, more cores, lower cost and flexibility. She says that design let AMD offer different configurations for high-performance computing and AI. In her account, ChatGPT made AI usable to a much wider audience and clarified the demand for more . Su expects multiple solutions: open, modular designs can be tailored to different workloads, while large GPUs will remain important for a while. Her argument is that technology leadership depends on sustained execution, partnerships and attention to customers, not just promising ideas. Su describes AMD’s AI work as a long-term opportunity and says developers need software that makes the hardware easier to use. The interview does not establish how the market will develop or whether AMD’s roadmap will succeed.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
Ben Thompson · Article
Source details
Original source date: Jun 6, 2024 (2024-06-06T10:55:29+00:00).
Date basis: Date stated on the publisher’s page.
Added to Escape Meridian: 2026-11-18T12:38:14.097Z. First observed: 2026-11-10T01:11:57.790Z. Retrieved: 2026-11-10T01:13:35.560Z.
Referenced by
Recorded citations in public readings. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on November 10, 2026; originally published June 6, 2024. This is an interview and presents Su’s account of AMD’s strategy; it does not independently establish the outcomes of the company’s decisions or forecasts.
Open question
What evidence will show whether open, modular AI systems can serve different workloads as effectively as more integrated solutions?
Source-stated
Lisa Su says the final stages of delivering a product are especially difficult and involve learning how to achieve yields, reliability and production at scale.
the last 5% of what it takes to get a product out is probably the hardest, where most of the secret sauce is.
Source-stated
Su says AMD chose chiplets to pursue higher performance, more cores, lower cost and flexibility.
we’re going to get to much higher performance, many more cores, as well as a much better cost point
Source-stated
Su says ChatGPT made AI understandable and usable to a broader public, clarifying the need for more AI compute.
now everybody could use AI. These models are not perfect, but they’re amazingly good
Source-stated
Su expects multiple AI solutions, with open and modular systems allowing solutions to be tailored to different workloads.
the beauty of open and modular is that you are able to, I don’t want to use the word customize here because they may not all be custom, but you are able to tailor.
Source-stated
Su says large GPUs will remain central for a while, even as she expects different solutions for different workloads.
these big GPUs that we’re going to build are going to continue to be the center of the universe for a while
Source and contributor context
Ben Thompson · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
CPU
The central processing unit: a general-purpose processor that carries out a computer's instructions.
Graphics processing unit
A chip that performs many calculations at once; the main hardware used to train and run AI models.
AI compute
The computing resources, meaning chips, data centers and electricity, used to train or run AI models.
Editorial provenance
Version 1. Publication ID: 2fff3608-e558-d996-7579-befbbaf8f307.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-18T12:34:34.882Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: c6f8b3a2a9a60822e5f43f7b9f57cf9683eae0eafc1fe302efa26dcab429c6cb. Source text hash: 307b1b63321b9bc34ac40985df6f6eebf6318d3795966fddcffc56a661886105.
Revision and source-check history
Version 1 · 2026-11-18T12:38:14.097Z
Ben Thompson interviews AMD CEO Lisa Su about solving hard problems
Published June 6, 2024, Ben Thompson’s Stratechery interview with AMD CEO Lisa Su connects difficult engineering work to business outcomes. Su says the hardest part of delivering a product can come after the initial design: achieving yields, reliability and production at scale. She describes AMD’s strategy as building on its and capabilities, choosing markets suited to high-performance computing, and executing a roadmap over years. Su presents chiplets as a deliberate bet on higher performance, more cores, lower cost and flexibility. She says that design let AMD offer different configurations for high-performance computing and AI. In her account, ChatGPT made AI usable to a much wider audience and clarified the demand for more . Su expects multiple solutions: open, modular designs can be tailored to different workloads, while large GPUs will remain important for a while. Her argument is that technology leadership depends on sustained execution, partnerships and attention to customers, not just promising ideas. Su describes AMD’s AI work as a long-term opportunity and says developers need software that makes the hardware easier to use. The interview does not establish how the market will develop or whether AMD’s roadmap will succeed.
Original source
Ben Thompson · Article
Source details
Original source date: Jun 6, 2024 (2024-06-06T10:55:29+00:00).
Date basis: Date stated on the publisher’s page.
Added to Escape Meridian: 2026-11-18T12:38:14.097Z. First observed: 2026-11-10T01:11:57.790Z. Retrieved: 2026-11-10T01:13:35.560Z.
Limits
Text as retrieved on November 10, 2026; originally published June 6, 2024. This is an interview and presents Su’s account of AMD’s strategy; it does not independently establish the outcomes of the company’s decisions or forecasts.
Open question
What evidence will show whether open, modular AI systems can serve different workloads as effectively as more integrated solutions?
Source-stated
Lisa Su says the final stages of delivering a product are especially difficult and involve learning how to achieve yields, reliability and production at scale.
the last 5% of what it takes to get a product out is probably the hardest, where most of the secret sauce is.
Source-stated
Su says AMD chose chiplets to pursue higher performance, more cores, lower cost and flexibility.
we’re going to get to much higher performance, many more cores, as well as a much better cost point
Source-stated
Su says ChatGPT made AI understandable and usable to a broader public, clarifying the need for more AI compute.
now everybody could use AI. These models are not perfect, but they’re amazingly good
Source-stated
Su expects multiple AI solutions, with open and modular systems allowing solutions to be tailored to different workloads.
the beauty of open and modular is that you are able to, I don’t want to use the word customize here because they may not all be custom, but you are able to tailor.
Source-stated
Su says large GPUs will remain central for a while, even as she expects different solutions for different workloads.
these big GPUs that we’re going to build are going to continue to be the center of the universe for a while
Source and contributor context
Ben Thompson · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
CPU
The central processing unit: a general-purpose processor that carries out a computer's instructions.
Graphics processing unit
A chip that performs many calculations at once; the main hardware used to train and run AI models.
AI compute
The computing resources, meaning chips, data centers and electricity, used to train or run AI models.
Editorial provenance
Version 1. Publication ID: 2fff3608-e558-d996-7579-befbbaf8f307.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-18T12:34:34.882Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: c6f8b3a2a9a60822e5f43f7b9f57cf9683eae0eafc1fe302efa26dcab429c6cb. Source text hash: 307b1b63321b9bc34ac40985df6f6eebf6318d3795966fddcffc56a661886105.
uncertain · 2026-11-18T12:34:34.882Z · anthropic/claude-opus-5.5
Publication: 2fff3608-e558-d996-7579-befbbaf8f307
attribution — pass: The reading correctly identifies the piece as Ben Thompson's Stratechery interview with AMD CEO Lisa Su. It attributes strategic views, forecasts and accounts of AMD's decisions to Su rather than presenting them as fact.
This week’s Stratechery Interview is with AMD CEO Lisa Su.
summary — uncertain: The summary is faithful and keeps Su's arguments: last-mile execution, the chiplet bet, ChatGPT clarifying AI demand, and open, modular tailoring versus large GPUs. But 'chiplets' is a central term and is neither explained nor in the definitions; a non-specialist will not know it means splitting a processor into smaller linked chips. 'Yields' and 'high-performance computing' are also unexplained. Smallest repair: add a plain-language gloss or tooltips for chiplets, yields and high-performance computing.
Is this the time we’re going to bet the company on going to chiplets?
date — pass: June 6, 2024 matches evidence.originalPublished (2024-06-06 UTC) and the page dateline. Su's forecasts are reported as her expectations, not outcomes. There is no hindsight framing. The related-story links at the end of the page are not used. This is a publisher page, so no archive disclosure is required.
Thursday, June 6, 2024
support — pass: Each claim is supported by its quoted excerpt. The statement that chiplets enabled HPC and AI configurations is supported by Su's account of the MI300A and MI300X. The developer-software point reflects her stated aim to make the transition easy for developers. The closing caveat and the question do not overreach.
because of our chiplet strategy, we could build a highly modular system that could be, let’s call it, an integrated CPU and GPU
media — pass: No media is selected, and no verified media candidates were supplied, so nothing is misidentified.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
The central processing unit: a general-purpose processor that carries out a computer's instructions.
Definition referenceA chip that performs many calculations at once; the main hardware used to train and run AI models.
Definition referenceThe computing resources, meaning chips, data centers and electricity, used to train or run AI models.
Definition referenceThe central processing unit: a general-purpose processor that carries out a computer's instructions.
Definition referenceA chip that performs many calculations at once; the main hardware used to train and run AI models.
Definition referenceThe computing resources, meaning chips, data centers and electricity, used to train or run AI models.
Definition reference