Afternoon edition — Wednesday 21 October 2026

OpenAI and Google Cloud announced on 21 October that OpenAI's frontier models are being added to Model Garden, Google Cloud's catalogue of AI models. Google says GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are rolling out now in private preview to selected enterprise customers, and that developers can already connect their local Codex CLI and Desktop workflows to Google's Gemini platform for inference. OpenAI says general availability is planned in the coming weeks.

Customers will buy the models through Google Cloud Marketplace, with usage on a single Google Cloud bill and counted against existing Google Cloud spending commitments. Both companies say the models will also come to Google Antigravity, Google's coding tool, and to Gemini for business. Google says OpenAI models run on dedicated Google Cloud infrastructure, that prompts and outputs are not used to train foundation models, and that Zero Data Retention and Confidential Compute options are coming soon.

Google Cloud now calls itself the only hyperscale cloud provider offering all three leading closed models. OpenAI's announcement says Microsoft remains its primary cloud partner.

At GTC Berlin on 21 October, Nvidia announced the OpenShell Reference Architecture, a hardware and software design for building, isolating, monitoring and enforcing limits on AI agents in companies and on edge devices. It builds on Nvidia's open-source OpenShell runtime, which gives an autonomous agent a secure boundary to work within.

In Nvidia's design, models run on Vera Rubin systems, agent harnesses and tool calls run on Vera CPUs, and BlueField-4 network processors run Nvidia Sentry, software that watches agent behaviour from outside the agent. Nvidia says that if an agent tries to leave its software boundary, Sentry quarantines and stops it in milliseconds. The architecture also extends to RTX Spark, DGX Spark and DGX Station machines.

Nvidia lists Anthropic, AWS, Cisco, CrowdStrike, Google Cloud, IBM, Meta, Microsoft, OpenAI, Palo Alto Networks, SAP, ServiceNow and others among companies working with the architecture, and says the banks BBVA and Santander are exploring it. It runs on Linux and Kubernetes from Canonical and Red Hat, and on Windows through Microsoft Execution Containers.

OpenAI said on 21 October that it worked with PerturbAI and Scripps Research to split the analysis of an existing experiment across thousands of AI agents powered by GPT-Rosalind, its life-sciences model family. The dataset shows how disrupting individual genes affects different cell types across millions of mouse brain cells. OpenAI calls the approach distributed biological reasoning.

Each agent took one gene disruption, analysed the cellular responses, searched the scientific literature and proposed explanations. Their reports were combined into a reasoning atlas linking genes, cell effects, publications and possible mechanisms, which OpenAI says is open for researchers to explore. The run took 32 minutes; the team estimated comparable human analysis would take months to a year.

OpenAI says the agents reproduced findings from earlier human analyses, that their ratings of novelty matched those of blinded human experts, and that they raised new leads, including a difference between how the genes Tsc1 and Tsc2 affect cerebellar cells and a possible link between cholesterol regulation and neurodegeneration.

Nvidia introduced DSX POD at GTC Berlin on 21 October: a reference design for complete Vera Rubin AI data centres that partners build inside modular containers and test before delivery. Rack-to-rack cabling, power, piping and liquid cooling are done at the factory, so a unit only needs to be connected to a prepared site's power, cooling and network. Nvidia says this cuts deployment at a prepared site from six months to as little as 14 days.

Partners named include Bull, Dell Technologies, HPE, Lenovo, Schneider Electric, Vertiv, Foxconn, Anduril and others. Nvidia says partners will deliver the modular units worldwide in early 2027.

At the same event Nvidia and Google Cloud announced A5X instances built on Vera Rubin NVL72 systems. Nvidia's blog says general availability comes next month, with expansion to Europe planned, and that Google Cloud plans to scale its Vera Rubin infrastructure to as many as 960,000 Rubin GPUs across several sites. Ineffable Intelligence, the London company founded by AlphaGo creator David Silver, was the first customer.

OpenAI said on 21 October that it is extending ChatGPT ads to 26 more countries, bringing the total to 89. The expansion adds 14 Latin American countries, including Argentina, Chile, Colombia and Peru, and markets such as Nigeria, South Africa, Serbia, Ukraine, Cambodia, Bangladesh and Pakistan.

OpenAI says more than 100,000 advertisers have run campaigns since ads launched just over eight months ago, and that more than 1.2 billion people use ChatGPT each week. Ads appear only on the Free and Go plans; Plus, Pro and Enterprise remain ad-free. OpenAI says it will next add more ad formats, objectives, buying options and measurement tools.

In a Truth Social post on Wednesday, 21 October, US President Donald Trump urged data centre companies to pay "the community, and/or individuals within the community, a 'dividend'" to make their projects popular. "I have watched the process closely, and Data Centers that were highly unpopular in a certain area, have become extremely popular when a financial deal is struck. MAKE DEALS, and get them built!!!" he wrote, as quoted by CNBC.

According to Anadolu Agency, Trump also said that new and large data centres are required to "make their own electricity, and give or sell any excess back to the local community, or grid," which he said could lead to "a substantial decrease" in electricity rates. The post did not say how large a dividend should be or how the electricity requirement would be enforced. Nebius's chief revenue officer told CNBC the dividend idea was "not crazy".