The Platform Layer Became the Prize This Week, OpenAI Lands on Every Cloud While DeepSeek Opens Its Engine and NVIDIA Opens Robotics

By Crypt0 · Oct 21, 2026

OpenAI said on October 21 that its frontier models and Codex agentic workflows are coming to Google Cloud's Model Garden, putting its models on all three of the top United States cloud providers. DeepSeek open sourced DeepInfer, the inference framework behind its production services, under the permissive Apache 2.0 license. NVIDIA expanded its Isaac open robotics platform at GTC Berlin with the Agile ONE SN reference humanoid, open source Isaac Lab 3.0, and the OpenShell agent sandbox. Three different labs, one shared move, opening the layer they used to guard.

This was the week the AI business stopped being only about the smartest model and became about the widest platform. Distribution, serving, and physical deployment all opened up at once. Whoever makes it easiest to build is collecting the builders.

The OpenAI and Google Cloud partnership is the headline. OpenAI said its frontier models and Codex agentic workflows are coming to Google Cloud's Model Garden, with general availability planned in the coming weeks and access requests open now. Google Cloud customers will be able to buy the models through the Google Cloud Marketplace and draw the spend against existing cloud commitments. The models will also appear inside Google Antigravity and the Gemini Agent. Amazon began hosting OpenAI models in April after Microsoft's exclusive license on OpenAI's intellectual property ended, so this move puts OpenAI's models on all three of the top United States cloud providers. Microsoft remains OpenAI's primary cloud partner. OpenAI president Greg Brockman called it an important step in making the most capable models and Codex available where customers are already building, and Google Cloud CEO Thomas Kurian said the company is focused on giving customers a broad selection of models, including those from OpenAI.

DeepSeek answered with openness of its own. The Chinese lab published DeepInfer on GitHub on October 20, describing it as the high performance inference framework that powers its large scale production services, released under the Apache 2.0 license so anyone can use, change, and sell it. The first release runs DeepSeek V4.1 Flash in offline mode and supports Nvidia SM100 GPUs alongside Huawei Ascend 950 chips, a first for a frontier lab. DeepSeek's own production figures show one cluster serving 155 trillion tokens per day with a 96 percent input cache hit rate, about 400 milliseconds to first token, and 230 tokens per second. The lab also open sourced its token encoding layer under the MIT license. In April 2025 DeepSeek had said its serving engine was too tied to internal systems to release. Eighteen months later, the whole engine is public.

NVIDIA took the opening to the physical world. At GTC Berlin on October 21, the company expanded its Isaac open robotics development platform with a reference humanoid, new datasets, a runtime, and updates to its robot learning framework. Agile Robots unveiled the Agile ONE SN, a full stack reference humanoid preconfigured with NVIDIA Isaac GR00T and powered by the NVIDIA Jetson Thor edge AI platform. The robot stands 1.40 meters tall, weighs 40 kilograms, offers 31 degrees of freedom with a 3 kilogram payload per arm, and carries dual WUJI Hand 2 five finger hands with 20 degrees of freedom each, 1kHz force control, and tactile sensing. Jetson Thor delivers 2,070 FP4 teraflops with 128GB of unified memory. Agile Robots plans to build 1,000 to 2,000 of them next year, with availability in early 2027. NVIDIA also made Isaac Lab 3.0 generally available as an open source learning framework, introduced Isaac OS as an open source runtime built on Holoscan, released the SimReady Central Library with more than 2,000 simulation ready assets on Hugging Face, and added the OpenShell agent sandbox plus the Agent Knowledge Foundation with Kumo models for enterprise data. Jensen Huang said the ChatGPT moment for robotics is fast approaching, and NVIDIA is handing developers the full kit to meet it.

For builders, the takeaway is generous. Frontier models are now available for purchase on every major cloud, the serving stack behind one of the most affordable frontier labs is open for study, and a reference humanoid exists so robotics teams can start from a working body instead of a blank page. The leverage in AI is moving to whoever owns the platform, and this week the platforms opened their doors wider than ever.

Quick answers

What is this story about?

Why does this story matter?

Sources

New to crypto? Read the crypto glossary, browse frequent questions, read our story, or explore the story archive.