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

F
Fox-IT International blog
Recent Announcements
Recent Announcements
D
Docker
IT之家
IT之家
B
Blog
Jina AI
Jina AI
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
博客园 - 【当耐特】
Google DeepMind News
Google DeepMind News
F
Fortinet All Blogs
量子位
C
Check Point Blog
Microsoft Azure Blog
Microsoft Azure Blog
罗磊的独立博客
博客园 - 司徒正美
李成银的技术随笔
美团技术团队
Blog — PlanetScale
Blog — PlanetScale
雷峰网
雷峰网
The GitHub Blog
The GitHub Blog
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
J
Java Code Geeks
T
The Blog of Author Tim Ferriss
酷 壳 – CoolShell
酷 壳 – CoolShell
MongoDB | Blog
MongoDB | Blog
P
Proofpoint News Feed
L
LangChain Blog
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
Y
Y Combinator Blog
大猫的无限游戏
大猫的无限游戏
有赞技术团队
有赞技术团队
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
V
Visual Studio Blog
T
Tailwind CSS Blog
H
Help Net Security
Engineering at Meta
Engineering at Meta
小众软件
小众软件
B
Blog RSS Feed
Stack Overflow Blog
Stack Overflow Blog
月光博客
月光博客
M
Microsoft Research Blog - Microsoft Research
宝玉的分享
宝玉的分享
人人都是产品经理
人人都是产品经理
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
GbyAI
GbyAI
H
Hackread – Cybersecurity News, Data Breaches, AI and More
Last Week in AI
Last Week in AI
Martin Fowler
Martin Fowler
Stack Overflow Blog
Stack Overflow Blog

The Register - On-Prem: Systems

Moving to mainframe can be cheaper than sticking with VMware: Gartner Inference is giving AI chip startups a second chance to make their mark Qualcomm teases ‘dedicated CPU for agentic experiences’ and ‘agentic smartphones’ Qualcomm teases ‘dedicated CPU for agentic experiences’ and ‘agentic smartphones’ Fujitsu confirms mainframe biz to die in 2035, in time for quantum AI supercomputers to take over Fujitsu confirms mainframe biz to die in 2035, in time for quantum AI supercomputers to take over Microsoft levels up Azure Local to make it fit for large-scale sovereign clouds ZTE partners with China's National Clinical Research Center for Interventional Medicine to build a new paradigm of smart interventional medicine ZTE partners with China's National Clinical Research Center for Interventional Medicine to build a new paradigm of smart interventional medicine Tenstorrent’s Galaxy Blackhole AI servers escape the event horizon The crypto-to-AI bandwagon jumpers' club just landed another member: Core Scientific Core Scientific accelerates crypto-to-AI pivot Meta Arms itself to the teeth by signing for 'tens of millions' of AWS Graviton cores Meta Arms itself to the teeth by signing for 'tens of millions' of AWS Graviton cores AI now gobbling up power and management chips for servers AI now gobbling up power and management chips for servers Musk bets Tesla's AI future on Intel node that isn't finished yet Tesla stakes AI dreams on Intel's unfinished AI chip SK Hynix’s aspirations for ’Merica-made HBM inch closer to reality SK Hynix breaks ground on Indiana advanced packaging plant Datacenter boom keeps dirty coal plants alive in the US Datacenter boom keeps dirty coal plants alive in the US Forget one chip to rule them all: With TPU 8, Google has an AI arms race to win Oil crisis? What oil crisis? IT spending de-coupled from wider war shock AMD's Ryzen 9 9950X3D2 Dual Edition tested World's blandest man steps down from CEO job to spend more time in tastefully appointed home World's blandest man steps down from CEO job Growing AI power slurpage prompts MPs to examine low-energy computing Intel eases reliance on TSMC with 'Merica-made Core Series 3 processors Intel eases reliance on TSMC with Core Series 3 CPUs A beginner's guide to GPU virtualization: passthrough, vGPU, and MIG Guide to GPU virtualization: passthrough, vGPU, and MIG Orbital datacenter startup admits launch economics don't fly Orbital datacenter startup admits launch economics don't fly AI-powered mainframe exits are a bubble set to pop AI-powered mainframe exits are a bubble set to pop Cloud-smart strategy helps Interactive meet GenAI demands Cloud-smart strategy helps Interactive meet GenAI demands Oracle taps Bloom for 2.8 GW of fuel cells to keep datacenter binge going Oracle taps Bloom for fuel cells to support datacenter binge Britain gives Rolls-Royce the nod to sketch out its mini reactor future Britain gives Rolls-Royce the nod to sketch out its mini reactor future When the IBM PC and shoulder pads were big, Japan led the chip industry. It's trying to get back there now Japan going back to the future by reviving its chip industry AWS ponders selling its home-grown chips by the rack-load, has almost sold out AI capacity AWS ponders selling its home-grown chips by the rack-load, has almost sold out AI capacity Supply chain challenges risk delaying Nvidia's Rubin GPUs Nvidia's Rubin GPU is likely to be late thanks to memory shortage and technical challenges Supermicro investigating alleged China chip smuggling Intel gets trapped in Elon’s reality distortion field as it joins in megafab delusions Intel gets trapped in Elon’s reality distortion field as it joins in megafab delusions No-Nvidia interconnect club delivers 2.0 spec before v1.0 silicon ships OpenInfra General Manager talks sovereignty, governments deploying tech 'kill switches' Anthropic reveals $30bn run rate, plan to use new Google TPU Anthropic reveals $30bn run rate, plan to use new Google TPU How Nvidia learned to embrace the light in its quest for scale Nvidia embraces optical scale-up as copper reaches limits IBM wants Arm software on its mainframes to better support AI IBM wants Arm software on its mainframes to better support AI AI server farms heat up the neighborhood for miles around, paper finds AI datacenters create heat islands around them, paper finds Arm says agentic AI needs a new kind of CPU. Intel's DC chief isn't buying it Arm says AI agents need a new CPU. Intel doesn't buy it Memory-makers' shares are down. Some RAM prices have eased. Blaming Google is not a good idea Memory-makers' shares are down. Don't blame Google US PC shipments to fall 13% as memory and storage crunch hits budget systems US PC shipments to fall 13% as memory costs surge ZTE showcases end-to-end intelligent computing at CloudFest 2026 in Germany, empowering the digital future South Korean AI chip startup Rebellions eyes new shores for rack-scale invasion South Korean AI chip startup Rebellions eyes new shores for rack-scale invasion Enterprise infrastructure is entering an economic reset Enterprise infrastructure is entering an economic reset AMD's new desktop CPU oozes cache out of all 16 cores Apple signs meaningless deal to make some less-important parts in America Three more charged over alleged Nvidia GPU smuggling scheme to China Three more charged with trying to smuggle GPUs to China Dell slims down business laptops, fattens up cooling and battery life Alibaba delivers RISC-V server chip optimized to run China’s top AI models Alibaba delivers RISC-V server chip optimized to run China’s top AI models AI-pilled Arm CEO teases mystery products that will turn it into a money machine AI-pilled Arm CEO teases mystery products that will turn it into a money machine Arm rolls its own 136-core AGI CPU to chase AI hype train Arm rolls its own 136-core AGI CPU to chase AI hype train SoftBank to build massive AI datacenter on former US nuclear weapons site SoftBank to build massive AI datacenter on former US nuclear weapons site US chip testing firm shrugged off ransomware hit as minor – then came the data leak Explainer: AI-ready servers Explainer: AI-ready servers Elon Musk wants to build 50 times more chips than the world currently produces, using 'new physics' Elon Musk wants to build 50 times more chips than the world currently produces, using 'new physics' Australia to datacenter operators: BYO energy, pay your way, build green, or stay home Australia to datacenter operators: BYO energy or stay home Supermicro co-founder arrested, charged over $2.5B Nvidia GPU sales to China Supermicro co-founder charged over $2.5B GPU sales to China Jeff Bezos' rocket company Blue Origin applies to launch 51,000 datacenter satellites Blue Origin applies to launch 51,000 datacenter satellites Alibaba has made 470,000 AI chips, admits they’re inferior Decoding Nvidia's Groq-powered LPX and the rest of its new rack systems Your next car might need 300 GB of RAM, and so will autonomous robots Your next car might need 300 GB of RAM, and so will robots
Alibaba has made 470,000 AI chips, admits they’re inferior
2026-03-20 · via The Register - On-Prem: Systems

Systems

Sees optimizing its entire cloud around homebrew silicon as the way to compete

Chinese web giant Alibaba has revealed its T-Head chipmaking business has shipped 470,000 AI chips, and admitted they are currently inferior to rival products, but believes it can build a mutually optimized stack that makes performance gaps moot.

CEO Yongming Wu revealed the 470,000 figure during the company’s Q3 2026 earnings call, without specifying the model he referred to. Alibaba has created several chips at least three chips designed for AI workloads, including the infamously buggy XuanTie C908, the TH1520 for edge AI, and the Pingtouge Zhenwu 810E. The latter debuted in January and is thought to be competitive with Nvidia’s H20, a throttled version of a 2023-vintage Hopper architecture accelerator.

If Wu was referring solely to the 810E, delivering 470,000 since January is impressive as it puts Alibaba’s output at the same pace as Nvidia’s: CEO Jensen Huang last year said the company shipped six million Blackwell GPUs in a year.

REG AD

However the Zhenwu 810E cannot compete with accelerators from Nvidia or AMD, a fact Wu acknowledged.

REG AD

“Given that our chips still lag behind foreign counterparts and performance in various respects, we aspire to engage in more profound co-design with Alibaba's cloud infrastructure and the Qwen model to provide improved cost effectiveness,” the CEO said. “This is one key differentiator and how we approach chip design at T-Head that sets us apart from other chip companies. Our primary goal is to create AI capabilities that offer superior value for money.”

He also said developing its own chips gives Alibaba “guaranteed supply of AI computing power,” which he said is necessary given “the unique circumstances currently facing the AI industry in China” – likely a reference to US export bans on advanced accelerators.

Wu also thinks developing its entire stack will mean Alibaba can lower inferencing costs, and that doing so will mean the company’s cloud can rake in plenty of profits as demand for AI services increases.

Alibaba Cloud is already doing well, growing quarterly revenue 36 percent year-over-year to $6.2 billion. The company predicted it can reach $100 billion of annual cloud and AI revenue within five years.

Wu also commented on speculation that Alibaba plans to spin out T-Head, saying the company doesn’t rule out a future float but has no “definitive time line” for making it happen.

Alibaba’s revenue for the quarter landed at $40.7 billion, representing two percent growth. Execs pointed out that had it not sold some businesses, growth would have been nine percent.

Domestic e-commerce and logistics services remain the company’s main businesses, together accounting for almost half of revenue. International e-commerce services brought in $5.6 billion, a six percent year-over-year increase.

That’s an unwelcome result for Alibaba, given Beijing is looking to its e-commerce giants to help it achieve a goal of increasing export sales.

REG AD

That leaves Alibaba’s cloud as its only fast-growing business, and with a lot riding on its ability to achieve the swift surge to from around $25 billion to $100 billion annual revenue.

CEO Wu thinks that growth is achievable because Alibaba Cloud saw a 6x increase in token consumption on its model studio platform in the last six months alone.

He thinks that kind of growth will recur.

“Cloud and software budgets for enterprise IT services have traditionally represented only around 5 percent of corporate revenue,” he said. “As model-driven agents begin to handle mainstream work tasks across industries, our total addressable market will expand by several multiples.” ®