NVIDIA, AMD and Intel Join SGLang Consortium
LMSYS Org launched the SGLang Consortium on Nov. 12, 2026, with NVIDIA, AMD, Intel, Google, Meta, Cloudflare, Modal and RadixArk as founding members. They join two roadmap meetings a year, but SGLang's maintainers keep the final say on code.
NVIDIA, AMD and Intel, three rival chipmakers, have joined Google, Meta, Cloudflare, Modal and RadixArk as founding members of the SGLang Consortium. LMSYS Org, the nonprofit behind the open-source SGLang inference engine, launched the group on Nov. 12 at the first SGLang Summit in San Francisco. Members will help shape the project's roadmap.
Key takeaways
- Eight companies are founding members of the consortium: NVIDIA, AMD, Intel, Google, Meta, Cloudflare, Modal and RadixArk.
- Each member names a technical representative and joins two roadmap meetings a year, but SGLang's maintainers keep the final say on technical decisions and on which code is accepted.
- The launch came with no new benchmark, price cut or savings figure. It sets up a way to work together, not a measured result.

What the SGLang Consortium is
SGLang is an open-source engine that serves large language and multimodal models in production. Its job starts after training: it schedules incoming requests, manages GPU memory and runs the model as fast and cheaply as the hardware allows.
In its announcement, LMSYS said the consortium gives members "a standing forum" to discuss inference trends with the maintainers and turn shared priorities into joint engineering work. That work will focus on four areas: support for frontier models, reasoning and multimodal workloads, distributed inference, and hardware optimization.
How the SGLang Consortium is governed
Members can contribute engineering expertise, deployment feedback, testing resources or joint development. Working groups may meet between the two yearly roadmap meetings, and LMSYS said it will post public updates on the roadmap and progress.
Control of the code does not move to the companies. "SGLang's maintainers keep responsibility for technical decisions and code acceptance, and the project stays open to the entire community," LMSYS wrote.
Why NVIDIA, AMD and Intel joined
LMSYS cited a Deloitte estimate that inference will make up roughly two-thirds of all AI compute in 2026, up from about half in 2025. The engine running that work sets how much performance each company's chips actually deliver to customers.
"Better software helps developers get more performance from NVIDIA compute," said Ian Buck, NVIDIA's vice president of hyperscale and high-performance computing. Ramine Roane, AMD's corporate vice president of AI applications, said SGLang "runs across hardware, including AMD Instinct GPUs."
Ying Sheng, a co-author of the original SGLang paper and co-founder and CEO of RadixArk, pointed to the limits on new data centers. "With the challenges in building new data centers, software approaches to improve AI efficiency will be incredibly important," she said.
How big SGLang already is
The SGLang documentation says the engine generates trillions of tokens a day across more than 400,000 GPUs. LMSYS lists Meta, Microsoft, Google, Amazon, Thinking Machines Lab, Cloudflare, Reflection and Modal among companies using it.
It already supports NVIDIA and AMD GPUs, Intel Xeon processors, Google TPUs, Huawei's Ascend NPUs and Moore Threads accelerators. Its latest tagged release, v0.5.21 on Oct. 2, bundled 779 pull requests from 227 contributors under the Apache 2.0 license.
Context: the open inference race
SGLang is not the only open engine. It competes with vLLM, and model makers are publishing their own serving code; DeepSeek, for instance, built its DeepInfer framework mainly on vLLM to run models such as DeepSeek-V4.1 Flash. A formal consortium gives SGLang a clear channel to chipmakers that rival engines may now try to match.
What's next
The first real signal will be the public roadmap updates LMSYS promised. Developers should watch whether hardware work from AMD, Intel and Google lands in SGLang releases at the same pace as NVIDIA support. For teams already running SGLang, nothing changes today.
FAQ
Do the member companies now control SGLang?
No. Members advise on the roadmap and contribute work, but the maintainers decide on technical matters and which code is merged.
Is SGLang still free to use?
Yes. It remains open source under the Apache 2.0 license and open to the entire community.
Did the consortium announce faster speeds or lower prices?
No. The launch announcement included no benchmark, price change or savings estimate.
References
- LMSYS Org Launches SGLang Consortium with NVIDIA, AMD, Intel, Google, Meta, Cloudflare, Modal and RadixArk to Advance Open-Source AI Inference — LMSYS Org (Nov. 12, 2026)
- Welcome to SGLang — SGLang documentation
- Releases · sgl-project/sglang — GitHub
- LMSYS Org Forms SGLang Consortium with NVIDIA, AMD, Google and Meta — AIwire (Nov. 13, 2026)
Editor responsible for AI Newsway. Writes about AI, developer tools, cloud and productivity technology.
Comments 0
Comments are currently disabled.