---
title: "Lisa Su Firmly Bullish on AI Compute Power"
type: "News"
locale: "en"
url: "https://longbridge.com/en/news/293767909.md"
description: "At the Advancing AI 2026 event, AMD unveiled the Helios rack-scale AI platform and several new products. Lisa Su emphasized that the focus of compute power is shifting toward inference, predicting that 60% of compute power will be used for inference by 2026. AMD is entering the AI factory competition with a complete system strategy, focusing on computing leadership, an open platform, and making AI ubiquitous, aiming to address the continuous inference demands brought by AI Agents"
datetime: "2026-07-24T14:05:00.000Z"
locales:
  - [zh-CN](https://longbridge.com/zh-CN/news/293767909.md)
  - [en](https://longbridge.com/en/news/293767909.md)
  - [zh-HK](https://longbridge.com/zh-HK/news/293767909.md)
---

# Lisa Su Firmly Bullish on AI Compute Power

Advancing AI is AMD's annual launch event for its AI computing ecosystem. On July 24, 2026 (Beijing Time), AMD centrally released the Helios rack-scale AI platform, the sixth-generation EPYC "Venice" server CPU, Instinct MI430X and MI350P, ROCm.AI, as well as the "Gorgon Halo" and Kria AI Robotics developer platforms for local Agents and robotics. Helios, with the MI455X as its computational core, was the star of the entire launch, marking AMD's entry into the AI factory competition with a complete system composed of GPUs, CPUs, networking, and software.

In her Keynote, Lisa Su summarized AMD's strategy into three points: computing leadership, an open platform, and "making AI ubiquitous." Behind these three strategies lies AMD's clear judgment on changes in the AI industry: while training will continue to expand, the center of gravity for compute consumption has accelerated toward inference; Agents have expanded a single model call into continuous inference, tool invocation, code execution, and data access. Consequently, AI infrastructure has evolved from a GPU issue into a systems engineering challenge involving CPUs, networking, software, power supply, and data centers.

Lisa Su predicted that approximately 60% of global AI compute power will be used for inference in 2026, emphasizing that Agents require both "massive GPUs to complete inference" and "massive CPUs to orchestrate every step."

On July 24, at the AMD Advancing AI 2026 event, Dr. Lisa Su, Chair and CEO of AMD; Vamsi Boppana, Senior Vice President of AMD's Artificial Intelligence Group; and Dan McNamara, Senior Vice President and General Manager of Computing and Enterprise AI at AMD, engaged in a nearly half-hour discussion with global media, including Tencent Technology. Compared to the product launches in the Keynote, this discussion touched upon several core issues facing AMD: whether AI demand will continue to grow at a high speed, whether Helios can be delivered on schedule, how AMD plans to catch up with NVIDIA, and whether an open ecosystem is truly a viable business path.

Lisa Su further provided AMD's judgment on demand: "The more useful AI becomes, the more people want to use it." **In her view, corporate investment in AI has shifted from initial cost reduction to more direct value creation, such as product development and business growth. Therefore, even as companies begin to control Token budgets, this will not change the long-term upward trend in compute demand.**

When directly asked, "Is AMD still following NVIDIA?" Lisa Su did not shy away. She stated that AMD has chosen a different path: covering different workloads with Chiplet and product portfolio architectures, and connecting customers and partners through an open ecosystem. She emphasized that AMD is not satisfied with natural market growth but rather "expects to grow faster than the market average in each segment it participates in"; moving from MI450 to MI500 is also "not a minor upgrade, but another major leap."

This competitive approach is also reflected in the development method of Helios. Lisa Su stated that Helios was not developed independently by AMD before being handed over to customers, but was co-designed with leading companies such as OpenAI, Meta, and Anthropic. Customers entered the product definition process earlier, spanning chips, software, data centers, and supply chains. She referred to this collaboration as an important "secret weapon" for the on-schedule mass production of Helios.

**Q: Companies are beginning to control Token budgets. Why does AMD still believe the AI market will grow rapidly?**

Lisa Su: That is a very good question.

First, stepping back, predicting the market is perhaps one of the most difficult things because forecasts are unlikely to be completely accurate.

But what we can be sure of is that we have spent a lot of time communicating with customers, trying to understand the fundamental changes happening at the workload level. What we see now is very strong demand for compute power. This holds true in the GPU and accelerator fields, but the pace of change in the CPU field is actually even more aggressive.

As for why we believe this trend will continue, you also heard the views of OpenAI, Meta, and other large customers today. We must plan years in advance to prepare chips, electricity, data centers, and the entire supply chain. Therefore, our collaboration with customers is closer now than ever before.

AI is a typical example: the more useful it is, the more people want to use it. Of course, everyone is constantly optimizing.

Dan previously discussed what we have seen in our cooperation with enterprise customers. Regardless, the overall trend of AI adoption continues to rise. We see AI playing an increasingly significant role in product development and other key areas where enterprises create value.

The value brought by AI has far exceeded cost reduction. Cost reduction may have been the starting point for enterprises' initial deployment of AI, but now it is increasingly used to create real enterprise value.

Therefore, we are very confident in the demand curve.

At the same time, people are gradually realizing that this is no longer just about a single chip. It requires a complete family of chips, product portfolios, and hardware systems. This has always been AMD's unique judgment.

**Q: How complex is Helios? The Helios plan is scheduled for deployment in the second half of this year. How does AMD define this deployment? Why are you confident in completing it on schedule?**

Lisa Su: First, I want to acknowledge this very clearly. From the hardware displayed on stage, you should have already seen that Helios is indeed impressive.

It is an incredible engineering achievement, involving powerful computing capabilities, power supply, system integration, density, and many other aspects.

But you need to remember that we have been preparing for this moment for several years.

This includes the capabilities we have built internally, such as ROCm and platform software capabilities; the capabilities built by our networking team; and our acquisitions of Pensando and ZT Systems, along with the integration of these capabilities.

One "secret weapon" I hope you can feel today is that we are keeping fully in sync with our partners. In other words, AMD is not developing independently first and then handing the product over to partners. We are co-developing.

OpenAI is co-developing with us, Meta is co-developing with us. You also heard relevant shares from Tom Brown, co-founder of Anthropic. Because of this, we are very confident in our product ramp-up.

"Full production" means we are ready to ship. We will begin shipping at the end of the third quarter and continue to scale up in the fourth quarter and the first half of next year.

We have completed clear planning for which data centers will receive the first batch of products. We are also working closely with ODM partners. One difference between Helios and previous products is that every link in the supply chain has been aligned. This is also an important reason for our confidence.

I feel very good about the progress of Helios.

**Q: Today AMD released a new data center rack and announced cooperation with Cerebras as well as progress in software. However, looking at the timeline, these products still seem to be following NVIDIA. Or is AMD's strategy not to fight for first place, but rather to believe that the AI market is large enough that holding a certain market share, even in second place, is acceptable?**

Lisa Su: We did release many products today. But seriously, we believe we are paving the way for the direction truly needed for future AI computing. GPUs are certainly very important, and we are very confident in our GPU roadmap.

As I previewed today, moving from MI450 to MI500 is not a minor upgrade, but another major leap.

We believe AMD will take the lead in Scale-up computing expansion. This is a very important judgment, but it is the direction we currently see.

We have chosen a different path. Over the past decade or so, especially the past five years, we have been investing in Chiplet architecture and product portfolio architecture.

Take EPYC as an example. Now everyone is starting to find CPUs interesting, but we have been preparing for this moment for many years.

Mark Papermaster and his team have been thinking about how to build an ecosystem that allows different workloads to use the most suitable chips. This philosophy also runs through our overall roadmap.

Therefore, we are very confident in our current position. This is a huge market. AMD believes that by 2030, the total addressable market opportunity we face will reach $2 trillion, which is a very large number.

But we do not just want to share in market growth. We expect that AMD's growth rate will exceed the market average in each segment we participate in.

AMD will compete in its own way. Our approach includes leading technology, an open ecosystem, and selecting the right compute for different workloads. Therefore, this requires both breadth and depth.

Vamsi Boppana: Lisa has covered it quite comprehensively.

New segments will constantly emerge in the market. We need to stay focused and execute rigorously on the most important tasks at hand, otherwise, it is easy to get distracted by various directions.

But at some point, we will choose a segment and clearly state that we believe we will achieve undisputed leadership there. We are very confident about this.

Lisa Su: At least today we can clearly say that AMD has an undisputed leading position in the CPU field.

**Q: In 2025, AMD reached a chip procurement and deployment agreement with OpenAI and granted warrants corresponding to up to approximately 10% of its shares. Why was a different transaction structure adopted with Anthropic this time? What changes have occurred in the market over the past year?**

**Additionally, what is the current progress of the OpenAI deployment? According to previous plans, related deployments were supposed to begin in the second half of 2026.**

Lisa Su: First, we are very happy, honored, and excited to cooperate with the world's most important frontier AI companies, including OpenAI, Anthropic, Meta, and others.

I would explain it this way: every deal is different.

Our relationship with OpenAI is very special. You should be able to feel this from the shares by Sachin and Philippe today.

We started cooperating with OpenAI early on, beginning with the MI300. Both sides jointly studied the roadmap and promoted software development. Frankly speaking, OpenAI has made a significant commitment to AMD, including deployments of up to 6GW.

We previously stated that the first 1GW would begin deployment in the second half of 2026 and continue into 2027. I can confirm that this plan is proceeding entirely as expected.

If we look at the overall situation without targeting a specific customer, the total demand for the MI450 series has exceeded initial expectations.

Customers have seen the actual performance of the hardware and like what they see. Therefore, we are increasing our production capacity and supply scale for the coming quarters.

As for Anthropic, I mentioned it briefly on stage. AMD has been paying attention to Anthropic for several years. We have been closely watching this company since Dario and Tom just founded Anthropic.

But such collaborations always need to find the right timing. For any company, introducing a new chip platform requires substantial engineering investment. Therefore, we understand that Anthropic's decision to invest engineering capabilities, resources, capital, and time into the AMD platform is a very significant one.

Every deal is different. Anthropic's agreement mainly revolves around the MI450, with a scale of up to 2GW. The first 1GW is planned to begin deployment in the first half of 2027. We are also discussing longer-term plans with Anthropic.

Therefore, the two deals simply adopted different structures.

**Q: How exactly are Anthropic and AMD co-developing? Does it involve data center resources?**

Lisa Su: Certainly. The relationship between AMD and Anthropic is first built on providing them with massive computing power to support Anthropic's continuous scaling.

You can expect that Anthropic will use AMD's GPUs and CPUs through multiple channels, including cloud services, and potentially including their own data centers.

We are working closely with Anthropic to support these different forms of deployment. The scope of this cooperation goes far beyond simply buying chips. Both sides will jointly complete system deployment and startup, and will also collaborate on running Claude on the AMD platform.

From a software perspective, Claude and Codex are currently among the most powerful and popular platforms.

We want developers to be able to develop on AMD hardware in a very efficient way through these platforms. This is also an important part of the cooperation.

Regarding data centers, we are also actively ensuring that there is sufficient deployment capacity in the market. This is indeed one of AMD's key areas of focus, but it is not just for Anthropic; it involves the overall needs of the entire system and ecosystem.

**Q: This cooperation is not necessarily surprising, but considering AMD's own chip design resources and experience, will it continue to cooperate with Cerebras in the long term? Or is this just a phased arrangement, with AMD eventually developing its own similar products?**

Lisa Su: I do not view any cooperation as a temporary transitional solution. We do not engage in cooperation from the perspective of "just a stopgap measure at the beginning."

The reason we cooperate with Cerebras is that Cerebras has very interesting technology. There are many different paths to achieve acceleration for specific workloads. Cerebras' technology is one of the very interesting ones, and it works well with Helios.

AMD's so-called open ecosystem means that we will cooperate with multiple companies possessing valuable technologies, as long as these technologies meet our judgments for long-term development.

In the long run, I believe there will be more decoupling and separation of workloads. The speed of this trend's advancement partly depends on the cost structure and price levels of different solutions.

But from AMD's own perspective, we have a very strong foundational roadmap, including MI400, MI500, and MI600. At the same time, we also have the ability to mix and match different computing technologies.

We are very excited about our cooperation with Cerebras. You can expect that AMD will continue to promote the splitting and targeted acceleration of more workloads in the future.

Vamsi Boppana: Let me quickly add a point. Actually, you don't need to look too far ahead. In AMD's existing product portfolio, there are already many small AI engines capable of handling ultra-low latency, ultra-low power consumption, and real-time response tasks.

You can look at our embedded product portfolio, as well as applications in automotive, medical devices, and other fields. The core point is that AMD believes AI will enter various computing forms in a tightly integrated manner, with different scenarios having different characteristics. We will provide corresponding solutions for these different scenarios.

Risk Warning and Disclaimer

The market carries risks; investment requires caution. This article does not constitute personal investment advice, nor does it consider the specific investment goals, financial status, or needs of individual users. Users should consider whether any opinions, viewpoints, or conclusions in this article align with their specific circumstances. Responsibility for investments made based on this lies solely with the investor.

### Related Stocks

- [AMD.US](https://longbridge.com/en/quote/AMD.US.md)
- [DAMD.US](https://longbridge.com/en/quote/DAMD.US.md)
- [AMDL.US](https://longbridge.com/en/quote/AMDL.US.md)
- [AMDG.US](https://longbridge.com/en/quote/AMDG.US.md)
- [AMUU.US](https://longbridge.com/en/quote/AMUU.US.md)
- [AMDD.US](https://longbridge.com/en/quote/AMDD.US.md)
- [AMDY.US](https://longbridge.com/en/quote/AMDY.US.md)
- [AMDW.US](https://longbridge.com/en/quote/AMDW.US.md)
- [SMH.US](https://longbridge.com/en/quote/SMH.US.md)
- [SOXX.US](https://longbridge.com/en/quote/SOXX.US.md)
- [SOXL.US](https://longbridge.com/en/quote/SOXL.US.md)
- [SOXQ.US](https://longbridge.com/en/quote/SOXQ.US.md)
- [XSD.US](https://longbridge.com/en/quote/XSD.US.md)
- [PSI.US](https://longbridge.com/en/quote/PSI.US.md)
- [FTXL.US](https://longbridge.com/en/quote/FTXL.US.md)
- [META.US](https://longbridge.com/en/quote/META.US.md)
- [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md)
- [CBRS.US](https://longbridge.com/en/quote/CBRS.US.md)
- [NVD.DE](https://longbridge.com/en/quote/NVD.DE.md)

## Related News & Research

- [Corsair collaborates with AMD to showcase AI compute at Advancing AI 2026](https://longbridge.com/en/news/293622257.md)
- [CBRS Stock Alert: What to Know as Cerebras Teams Up With AMD](https://longbridge.com/en/news/293768646.md)
- [The Next Evolution of AI Infrastructure: Inside the Architecture Powering the AI Factory Era](https://longbridge.com/en/news/293657379.md)
- [EXCLUSIVE-Intel, AMD sign long-term server CPU deals with Chinese clients as prices surge, sources say](https://longbridge.com/en/news/293552691.md)
- [Primis AI Becomes ChipNexus and Launches NEX for Agentic Chip-Design Automation](https://longbridge.com/en/news/293657601.md)