惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
MyScale Blog
MyScale Blog
Jina AI
Jina AI
爱范儿
爱范儿
cs.CL updates on arXiv.org
cs.CL updates on arXiv.org
I
Intezer
The Cloudflare Blog
T
Threat Research - Cisco Blogs
G
Google Developers Blog
Stack Overflow Blog
Stack Overflow Blog
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
D
Docker
AI
AI
Scott Helme
Scott Helme
Attack and Defense Labs
Attack and Defense Labs
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
L
LangChain Blog
Recent Announcements
Recent Announcements
Security Latest
Security Latest
Hugging Face - Blog
Hugging Face - Blog
W
WeLiveSecurity
Last Week in AI
Last Week in AI
Security Archives - TechRepublic
Security Archives - TechRepublic
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
P
Proofpoint News Feed
S
Securelist
S
Security Affairs
Project Zero
Project Zero
博客园 - 叶小钗
Google DeepMind News
Google DeepMind News
T
Tor Project blog
A
About on SuperTechFans
V2EX - 技术
V2EX - 技术
宝玉的分享
宝玉的分享
T
Tenable Blog
博客园 - 聂微东
人人都是产品经理
人人都是产品经理
Simon Willison's Weblog
Simon Willison's Weblog
Forbes - Security
Forbes - Security
K
KPMG report finds enterprise disconnect between AI and its ROI | CIO
V
V2EX
AWS News Blog
AWS News Blog
The GitHub Blog
The GitHub Blog
酷 壳 – CoolShell
酷 壳 – CoolShell
P
Privacy & Cybersecurity Law Blog
阮一峰的网络日志
阮一峰的网络日志
I
InfoQ
C
CXSECURITY Database RSS Feed - CXSecurity.com
H
Hacker News: Front Page
美团技术团队

The Register - Special Features: The Future of the Datacenter

AI is rewriting how power flows through the datacenter All aglow about DCs, investors launch $300M at microreactor startup Why do bit barns keep bumping up our bills, Senators ask DC operators Delays? What delays? Oracle insists its $300B cloud contract with OpenAI is on track Galactic Brain space datacenter coming in 2027, pledges startup Aetherflux Activist groups urge Congress to pause datacenter buildouts Bezos-backed Unconventional AI addresses datacenter power Meta and Google tap NextEra to feed their hungry datacenters Datacenters accused of hoarding grid capacity Amazon’s Trainium3 is the latest to conform to Nvidia’s mold Palantir aims to help energy companies meet AI power crunch Datacenters planned for Scotland could drain a loch of power Datacenters must generate their own power or fail HPE to ship rack-scale AI system using AMD's Helios in 2026 London grid crunch delays new housing amid datacenter boom Britain plots atomic reboot as datacenter demand surges OCP learning how to get quantum computers into existing DCs Nvidia's AI factory dream gets the Omniverse treatment Qualcomm announces AI accelerators and racks they'll run in
AMD taking AI fight to Nvidia with Helios rack-scale system
Dan Robinson Dan Robinson · 2025-11-06 · via The Register - Special Features: The Future of the Datacenter

The Future of the Datacenter

CEO Lisa Su says next-gen MI400 GPUs and architecture gaining traction with hyperscalers

AMD plans to launch its Helios rack-scale architecture in 2026 as a direct challenge to Nvidia in the AI infrastructure market, pending successful integration of its next-gen GPUs and processors.

The chip firm's President & CEO Dr Lisa Su talked up its progress in the datacenter, AI, and server businesses during an investor conference call for AMD's third quarter 2025 financial results.

She claimed server CPU revenue reached an all-time high, driven by rapid adoption of AMD's 5th Gen Epyc Turin processors, which accounted for nearly half of overall Epyc revenue during the quarter.

Looking ahead, Su said AMD's datacenter AI business is to enter "its next phase of growth" and customer interest is building ahead of the launch of the next-gen MI400 series GPU accelerators and the Helios rack-scale solution next year.

AMD showed off Helios in June, at an event in San Jose, California. It is designed to operate a rack full of accelerators as if they were one single large GPU, like Nvidia's DGX GB200 NVL72 system.

"Helios integrates our Instinct MI400 Series GPUs, Epyc Venice CPUs and Pensando NICs in a double-wide rack solution optimized for the performance, power, cooling and serviceability required for the next generation of AI infrastructure and supports Meta's new open rack wide standard," Su said on an earning conference call last night.

Venice is the codename for AMD's 6th-gen Epyc CPUs, set to be produced using TSMC's 2nm manufacturing process, also due next year.

"Development of both our MI400 Series GPUs and Helios rack is progressing rapidly, supported by deep technical engagements across a growing set of hyperscalers, AI companies and OEM and ODM partners to enable large-scale deployments next year."

The ZT Systems team, a maker of high-performance servers for cloud operators that AMD acquired last year, is playing a critical role in Helios development, Su added.

In response to an analyst question, she said customers for its MI450 GPU will buy them as part of the rack-scale solutions. "We will have other form factors as well for the MI450 Series, but there's a lot of interest in the full rack-scale solution."

However, AMD has also seen signs of a pickup in CPU demand in the midst of all this AI frenzy.

"A number of our large hyperscale clients are now forecasting significant CPU build into 2026. And so from that standpoint, I think it's a positive demand environment, and it is because AI is requiring quite a bit of general-purpose compute," she explained.

AMD was cagey about when it might start to see the benefits from its rack scale push in its bottom line.

"We're not guiding 2026, but our priority in datacenter GPU business is to really expand the top line revenue growth and the gross margin dollars," said chief finance officer Jean Hu.

Another analyst, Timothy Arcuri of UBS Investment Bank, inquired about AMD's agreement with OpenAI, announced last month.

"It looks like they could be something like half of your datacenter GPU revenue in the 2027, 2028 time frame. So how much risk, in your mind, is there around that single customer for you?" he asked.

Su brushed this aside, saying the deal had been in the works for quite some time, and AMD is happy with the planned scale of the deployments.

"I think we're dimensioning the supply chain in such a way that we would have ample supply to have multiple customers at similar scale as we go into the '27, '28 time frame, and that's certainly the goal," she stated.

Overall AMD's revenue grew 36 percent to $9.2 billion during the third quarter ended September 27, compared with the same period last year. Its datacenter business saw revenue up 22 percent to $4.3 billion, attributed to sales of its Instinct MI350 series GPUs and server share gains.

In client and gaming systems, revenue was up 73 percent to $4 billion, with Su saying that "Our PC processor business is performing exceptionally well with record quarterly sales as breadth of our leadership Ryzen portfolio accelerates growth."

Revenue declined 8 percent year-on-year to $857 million for AMD's embedded portfolio, although the company says it is on track for a second straight year of record design wins, reflecting the growing adoption of its products across a broad range of markets and expanding set of applications.

For the fourth quarter of 2025, AMD forecasts revenue of $9.6 billion, plus or minus $300 million, which would represent year-on-year growth of roughly 25 percent. The firm said that its current outlook does not include any revenue from AMD Instinct MI308 shipments to China. ®