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

V
Visual Studio Blog
Y
Y Combinator Blog
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
Hugging Face - Blog
Hugging Face - Blog
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
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
The Cloudflare Blog
L
LangChain Blog
美团技术团队
N
Netflix TechBlog - Medium
量子位
酷 壳 – CoolShell
酷 壳 – CoolShell
B
Blog
博客园 - 司徒正美
爱范儿
爱范儿
D
DataBreaches.Net
月光博客
月光博客
U
Unit 42
B
Blog RSS Feed
Engineering at Meta
Engineering at Meta
Apple Machine Learning Research
Apple Machine Learning Research
Jina AI
Jina AI
MongoDB | Blog
MongoDB | Blog
腾讯CDC

The Register - Off-Prem

Enterprise cloud infrastructure uptake shows no sign of slowing The majority of corporate IT is now off premises for the first time Microsoft fiber foul-up cut off Azure California for almost five hours Web app turns your old phone into a new smart display Anyone with a shed, an extension cord, a couple of GPUs and an overdraft is building datacenters. Fujitsu just offloaded five Iran says it Google Cloud outage shows it’s still hard to understand hyperscalers’ real resilience regimes AWS customer learns the hard way how even the smallest oversight can be mission-critical Billing software error sends billion-dollar AWS estimates Top EU court clips YouTube AWS CloudFront outage serves errors instead of websites India’s tech services giant HCL is getting into the AI datacenter business Britain Microsoft shifts to annual exchange rate price revision for cloudy products Amazon’s Mechanical Turk to stop accepting new customers – and not even AI can save it Fire burns Google Cloud India’s network, which remains slow a week later EU sovereignty push gives tech buyers a new alphabet soup to swallow Google, Canonical team up to certify Ubuntu images for TPU VMs Arm moves into the heart of the cloud stack Snowflake to burn $6B on AWS Graviton CPUs and AI accelerators Big Tech extracts retirement-scale wealth from UK internet users, research shows Open Compute urges local government to bask in the warm glow of excess datacenter heat Google Cloud suspended major customer Railway.com without cause, causing outage Broadcom finds a VMware customer willing to stick around: London Stock Exchange Baidu says the quiet part out loud – you can’t build AI infrastructure, so clouds can cash in AWS racks M3 Ultra Macs that boast specs you can’t currently buy Red Hat blasts RHEL 10.1 into orbit aboard Voyager's micro datacenter Sovereign cloud is only possible if you’re Chinese or American: Gartner Cloudflare to fire 1,100 staff whose jobs just aren’t AI enough AWS warns of EC2 'impairment' as power loss hits notorious US-EAST-1 region
Tencent admits GPUs only pay for themselves when powering...
Simon Sharwood · 2026-05-14 · via The Register - Off-Prem

Off-Prem

Chinese web giant says accelerator shortage is over as local hardware arrives in volume

Chinese web giant Tencent struggles to earn a return on investment from GPUs – unless it uses them to power its advertising business.

“If we buy GPUs and we deploy them into our ad tech, then that's a relatively short-cycle investment,” said Chief Strategy Officer James Mitchell during the company’s Q1 2026 earnings call. “The GPUs yield better targeting, higher click-through rates and higher revenue and profit on a pretty accelerated basis,” he said.

But the company views GPUs powering work on its Hunyuan foundation model as “important for our franchise.”

Mitchell said Tencent is comfortable with this situation.

“There's been many products within Tencent … that went through lengthy incubation periods where they had no return on investment, but we were confident in the franchise value creation,” he said. “And then over time, they had more lengthy harvesting periods where we've been able to drive very healthy returns on that sunk investment.”

He predicted that AI will go through the same cycle

But Tencent is struggling to make the wheel turn because it’s only had enough GPUs to power its own services, leaving its public cloud without enough accelerators to rent to customers.

Mitchell said Chinese manufacturers will soon fill the gap.

“As the supply of China design GPUs progressively ramps up, then we'll be remedying that situation,” he said.

Chief financial officer Shek Hon Lo weighed in with an observation that two factors made it hard for Tencent to get all the GPUs it wants: US sanctions, and “limited fab capacity within China.”

“That's now being addressed because the China designed ASICs are seeing more supply from fabs within China as well as more supply from fabs in neighboring countries,” he said.

But Tencent still expects GPU procurement to be harder than buying CPUs, as Lo said the company has “very long-term” deals with CPU vendors.

“We've been a big customer for Intel and AMD for many years,” he said. “We've been progressively growing our volume with them for many years, and they believe it will continue to progressively grow our volume for many years to come.”

That remark will be cause for celebration at the US companies, which have watched other hyperscalers invest heavily in custom Arm silicon.

Tencent posted another strong quarter, with revenue of RMB196.5 billion ($28.9 billion) representing 12 percent growth. The company’s Weixin and QQ messaging apps have 1.95 billion monthly combined users. Tencent has tweaked their mobile apps “to act as communication interfaces for controlling AI agents, allowing users to orchestrate agents from mobile for complicated task execution on PC and cloud.”

Tencent’s Western rivals Google and Meta haven’t yet built similar apps. And they don’t experience the same hardware acquisition problems Tencent faces. ®