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

博客园 - 司徒正美
The GitHub Blog
The GitHub 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
Apple Machine Learning Research
Apple Machine Learning Research
L
LangChain Blog
GbyAI
GbyAI
博客园_首页
V
Visual Studio Blog
Martin Fowler
Martin Fowler
WordPress大学
WordPress大学
H
Hackread – Cybersecurity News, Data Breaches, AI and More
博客园 - 叶小钗
腾讯CDC
博客园 - Franky
IT之家
IT之家
Google DeepMind News
Google DeepMind News
Microsoft Azure Blog
Microsoft Azure Blog
D
Docker
大猫的无限游戏
大猫的无限游戏
Recent Announcements
Recent Announcements
小众软件
小众软件
博客园 - 三生石上(FineUI控件)
B
Blog
酷 壳 – CoolShell
酷 壳 – CoolShell

The Stack

How Lloyds has transformed its trusted critical data for AI Micron's revenue surges with no end to the memory shortage in sight Ubiquiti UniFi OS vulnerabilities exploited in the wild Mistral's new OCR 4 model shows size matters EU approves €76m in state aid for chip testing firm Even FedEx is a winner of the data centre boom Why AWS thinks AI coding assistants need a new place to run Why China is forging closer ties with open-source foundations China's CPU-only supercomputer tops power list Runtime: Why 'Bring-your-own-Cloud' is taking off; the TRAIT&R framework for agents, and... AWS adds yet another security tool to speed up patching Scattered Spider: UK teen pleads guilty to TfL attack OSS security finally gets a Magic Quadrant Places for People forking out £60m for new ERP system Google offers ATT&CK for bad AI, with surging cost forecast STACKUP: The Stack's weekly tech startups and funding wrap Bring Your Own Cloud: why more enterprises are buying in Runtime: Vercel sets up a new framework for agents, Databricks finds a new Genie, and… Borussia Dortmund's IT head has to sometimes think "outside of the box" Accenture stumbles after Middle East, AI, merger questions Cabinet Office offering £100k+ for AI "ambassador" Critical Splunk Enterprise vulnerability being exploited Accenture buys majority stake in OT security company Dragos npm's security rethink, Satya’s warning, a COBOL win, and... US spending billions to counter China's tech influence NatGeo Society CTO on migrating 15 years of video data to the cloud Alibaba opens French data centres in Europe cloud push SpaceX to take over Cursor in $60bn play for AI coding market Hosting firm Hetzner hikes prices sharply amid supply chain pain HSBC's CIO: AI is easy. True impact is hard.
AWS takes NVIDIA "G7" Blackwell instances GA
The Stack · 2026-06-22 · via The Stack

Unsatisfied with AWS’s choice of 850+ EC2 instances types?

Public cloud’s biggest hyperscaler just took its “G7” instances to general availability – albeit only in its US East (Ohio) and US West (Oregon) regions.

The new instances are AWS’s first offering of compute underpinned by NVIDIA’s Blackwell server GPUs – and intended for meaty AI inference work.

Networking is handled by AWS’s Nitro System which handles networking, storage, and other I/O functions, and can, it added, “deploy firmware updates, bug fixes, and optimizations while it remains operational…”

AWS said the G7 instance can be deployed as “on-demand instances, as part of savings plans, or spot instances” (on paper; inference demand is so rampant that The Stack cannot imagine much capacity is available) and can be provisioned with up to 8 GPUs per instance and up to 700 Gbps of networking throughput” – seven times more than its G6 instances. 

The general availability makes AWS, it boasted, the “first major cloud provider to support NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs.”

It will cost you, though. (AWS pricing, June 22, below.)

NVIDIA in May reported data centre revenue of $75 billion, up 92% year over year and 21% sequentially, driven by Blackwell architecture demand.

Executives boasted on May 20 that “Microsoft's Farweave, the world's most powerful AI data center, is now live ahead of schedule [and] powered by hundreds of thousands of Blackwell GPUs. Starting this year, AWS will add more than one million Blackwell and Rubin GPUs and are [sic] collaborating on spectrum networking. At Google, Blackwell will be offered to customers in the cloud including confidential computing capability, they added on the call.

AWS is making prepackaged GPU drivers for AI inference and graphics workloads available via its AWS Deep Learning AMIs (DLAMI) or NVIDIA Workstation AMIs; for EKS, users can build EKS AMIs with NVIDIA driver version R595 with EKS-provided automation. Users can deploy Amazon Linux, Ubuntu, RHEL, and Windows Server as operating systems with G7.