Podcast Summary: Bloomberg Talks — AMD CEO Lisa Su on New Data Center Chip and AI Demand

Date: January 6, 2026
Host: Ed Ludlow (Bloomberg)
Guest: Dr. Lisa Su, CEO, AMD
Location: CES, Las Vegas

Episode Overview

In this episode taped live at CES 2026, Bloomberg’s Ed Ludlow talks with AMD CEO Dr. Lisa Su about the company’s groundbreaking new data center chip—MI455X—which is the world’s first 2-nanometer chip for rack-scale systems, the accelerating global demand for AI compute, AMD’s enterprise and edge strategies, expanding markets like China, and the role of AI across industries and the broader economy.

Key Discussion Points and Insights

1. Launch of the Helios System & MI455X: First 2nm Data Center Chip

• Significance:
  • Helios is described as AMD’s “first rack scale system solution,” powered by the MI455X chip—the world’s first 2nm chip of its type.
  • MI455X offers a leap forward with 320 billion transistors, constructed from 2 and 3nm chips.
  • Su contextualizes this launch as meeting the surging demand for AI compute, predicting usage will expand from 1B active users today to over 5B within five years (01:01).

“The demand for AI compute is just continuing to increase... MI455 is a significant leap forward in terms of technology capability.” — Dr. Lisa Su (01:01)

2. Roadmap & Partnership Highlights

• Deployment Timeline: MI455X will begin shipping in 2H26 and ramp up from there (02:17).
• Partners: AMD is working closely with OpenAI (Greg Brockman joined the keynote), Oracle, and other industry leaders (02:17).

3. Enterprise & Edge Computing: The MI440X

• Su introduces the MI440X, aimed at small data centers and enterprises preferring on-premise or private cloud AI deployments using existing infrastructure (02:58-04:10).
• Key industries include financial services and healthcare, where data sovereignty is crucial.

“Enterprises are starting to increase their adoption of AI... In some cases they want their own control of their data centers.” — Dr. Lisa Su (03:40)

4. AMD vs. Nvidia: The Differentiator

• AMD differentiates itself from Nvidia by focusing on:
  • Performance per dollar (TCO advantage)
  • Open ecosystem (not closed vertical integration)
  • Strong ecosystem partnerships (05:19)

“Our claim to fame is really outstanding performance at very advantaged total cost of ownership… and an open ecosystem.” — Dr. Lisa Su (05:42)

5. AI Compute Bottleneck & The “Yadaflop” Era

• Su shares the severity of the compute deficit:
  • About 100 zettaflops of compute exist globally today, but needs will increase 100x to a yadaflop (10^24 flops) in next 4–5 years (06:42).
• Software progress is stalling due to hardware limitations; compute and memory scaling is crucial.

“The models we have today are great... but it can get so much better... we’re still in the early innings of unlocking the power of AI.” — Dr. Lisa Su (07:45)

6. Barriers to Scaling AI (Energy, Memory, Supply Chain)

• Ludlow asks about the biggest constraints: memory, energy, supply chain (08:28).
• Su answers the issues aren’t isolated—it’s all of the above, requiring industry-wide alignment and scaling.

“It's not any one thing… we’re all looking at how do we build faster?... that's why partnership is just so important.” — Dr. Lisa Su (09:49)

7. System Dominance and the Open Ecosystem Strategy

• While competitors (e.g., Nvidia) seek to own all components inside the server, AMD prioritizes seamless turnkey solutions that are optimized but maintain openness and standards-based integration (10:41).

“We are very focused on an open ecosystem... it’s all about ensuring that we get the best of all worlds.” — Dr. Lisa Su (11:10)

8. Future Performance Leaps: MI500 and Beyond

• Su reveals MI455 offers 10x the performance of its predecessor (MI355, released just six months ago), and MI500 (coming in 2027) will offer an additional 10x, totaling a mind-boggling 1000x from MI300 to MI500 (12:15).

“It is just incredible engineering at every level.” — Dr. Lisa Su (12:16)

9. China: Demand, Export Controls, and Licenses

• China remains a vital market for AMD, especially in data center chips.
• They have received licenses for older models (MI308), are applying for new ones (MI325), and remain optimistic about further approvals (12:54-14:39).
• Su notes China’s big drive for open-source AI and strong demand for advanced compute.

10. Market Metrics: Taking Share & Growth Outlook

• Su points to the MI455 launch and deep industry partnerships as evidence of AMD’s momentum (15:09).
• AMD projects “tens of billions in AI revenue into 2027.”

11. PC Market Position

• AMD saw strong PC segment growth in 2025 due to a robust product portfolio, early AI bets, and Windows 11 refresh cycle (16:13-17:06).
• Su calls out untapped potential in enterprise laptops, contrasting with strength in gaming and consumer PCs.

12. AI’s Economic Impact

• Ludlow presses Su for signs of AI’s broader economic effect (17:47).
• Su says productivity gains are already visible internally and for customers, expects clear GDP-level impact in coming years.

“…every CEO that I talk to is talking about AI. It is front and center in terms of how to build a better company.” — Dr. Lisa Su (18:48)

13. Robotics & Physical AI: Partnership with Generative Bionics

• AMD powers humanoid robots through partnerships like Generative Bionics, both on real-time (“brain inside the robot”) and infrastructure sides (19:52-20:55).
• Su hints at end-to-end solutions—from chip to deployed intelligence—suggesting future expansion.

14. Public Policy, US Leadership, and Infrastructure

• Su discusses the US government’s infrastructure and deregulation push to accelerate domestic chip leadership and deployment (21:17).
• Applauds progress, especially around public-private partnerships (Genesis Mission), export controls, and building US-based capacity.

15. Looking Ahead: AI in 2026 & Beyond

• Predicts accelerating, tangible adoption of AI in real-world business and daily life.
• Closes with a sense of excitement and momentum.

“You ain’t seen nothing yet... AI is not just hype out there... it's things that people are using every day, real time and feeling like, hey, my life is better because I have this technology.” — Dr. Lisa Su (23:23)

Notable Quotes & Memorable Moments

• Predicting the “Yadaflop” Revolution:

  “A yada flop is 10 to the 24th in terms of flops. To give you a view... that's another 100 times more compute than we have today.” — Dr. Lisa Su (07:12)

• AI’s Perceived Limits and Potential:

  “The models... today are great...but it can get so much better... we’re still in the early innings.” — Dr. Lisa Su (07:45)

• Market Share & AI Revenue Outlook:

  “We’ve talked about tens of billions of dollars in AI revenue as we get into 2027.” — Dr. Lisa Su (15:27)

• On Accelerated Engineering:

  “MI455 is 10 times better than the chip that we just launched six months ago... MI500 is another 10X on top of that.” — Dr. Lisa Su (12:15)

Recommended Listening

• Hear Lisa Su’s explanation of scaling AI infrastructure (08:54).
• For China export controls and market outlook, see (12:54–14:39).
• For insights into AMD’s robotics and “physical AI” push, see (19:42–20:55).

Tone & Style

This discussion is energetic, technically rich, and optimistic throughout, with Su expressing excitement for both the state of AI adoption in 2026 and the explosive hardware advances driven by AMD. The conversation balances technical details with big-picture analysis of AI’s economic and societal impact, always with a candid and forward-looking slant.