Search “NVIDIA”

NVIDIA announces expansion of AI infrastructure capacity with supply chain partners

On October 21, 2026, NVIDIA announced at GTC Berlin the expansion of global manufacturing capacity with supply chain partners including Amkor, IBIDEN, KYEC, Samsung Electro-Mechanics, SPIL and Unimicron. According to information released by NVIDIA Newsroom that day, it plans to increase substrate capacity 2.5 times over the next three years; Amkor plans to increase advanced packaging capacity by more than 300% by the end of 2028; NVIDIA is providing KYEC with a $1.1 billion capacity reservation prepayment to support its new test facility in the United States, targeting mass production in April 2027; SPIL plans to increase co-packaged optics optical engine capacity 15 times by mid-2027 and seeks to double it again by the end of 2028. The Elec reported on October 22 that the agreements with Ibiden, Samsung Electro-Mechanics and Unimicron will expand semiconductor packaging substrate capacity 2.5 times over the next three years, and that Ibiden's Ono plant in Japan will triple its capacity and reach full production in the first quarter of 2028; the report also said NVIDIA prepaid KYEC $1.1 billion, doubling its total test capacity in 2027 compared with 2026. There are currently no facts of subsequent progress.

NVIDIA announces expansion of AI infrastructure capacity with supply chain partners

On October 21, 2026, NVIDIA announced at GTC Berlin the expansion of global manufacturing capacity with supply chain partners including Amkor, IBIDEN, KYEC, Samsung Electro-Mechanics, SPIL and Unimicron. According to information released by NVIDIA Newsroom that day, it plans to increase substrate capacity 2.5 times over the next three years; Amkor plans to increase advanced packaging capacity by more than 300% by the end of 2028; NVIDIA is providing KYEC with a $1.1 billion capacity reservation prepayment to support its new test facility in the United States, targeting mass production in April 2027; SPIL plans to increase co-packaged optics optical engine capacity 15 times by mid-2027 and seeks to double it again by the end of 2028. The Elec reported on October 22 that the agreements with Ibiden, Samsung Electro-Mechanics and Unimicron will expand semiconductor packaging substrate capacity 2.5 times over the next three years, and that Ibiden's Ono plant in Japan will triple its capacity and reach full production in the first quarter of 2028; the report also said NVIDIA prepaid KYEC $1.1 billion, doubling its total test capacity in 2027 compared with 2026. There are currently no facts of subsequent progress.

View story:NVIDIA announces expansion of AI infrastructure capacity with supply chain partners

NVIDIA Expands Nemotron Retriever Series and Releases Multiple Open Models

On October 21, 2026, NVIDIA expanded the Nemotron Retriever series, releasing multiple open models covering embedding, reranking, and document parsing. The models introduced include Nemotron 3.5 Embed (8B BF16, 1B BF16, 1B NVFP4), Nemotron 3.5 Rerank 8B BF16, and Nemotron 3.5 Parse with 631 million parameters, aimed at multimodal retrieval and document processing for enterprise agent workflows. NVIDIA says that adding Rerank to the agent retrieval pipeline reduces the token cost per query by about 14% (with the 8B embedding model) and 13% (with the 1B model), respectively, while also improving retrieval accuracy. Parse scores 75.20 on ParseBench and processes 10.5 pages per second on a single H100. Open weights, data, and training recipes have been released on Hugging Face and can be deployed via NVIDIA NIM and through Baseten, DeepInfra, and FriendliAI.

View story:NVIDIA Expands Nemotron Retriever Series and Releases Multiple Open Models

NVIDIA and Google Cloud announce the launch of the first A5X instances

On October 21, 2026, NVIDIA and Google Cloud announced at GTC Berlin that the first A5X instances equipped with NVIDIA Vera Rubin NVL72 are already running at one customer. The two companies plan to make them generally available in November 2026 and to expand the instances to Europe. A5X combines ConnectX-9 SuperNICs, Google Cloud's Virgo network, and fourth-generation liquid cooling, targeting large-scale agentic inference and reinforcement learning workloads; Google Cloud says it plans to scale this infrastructure to as many as 960,000 Rubin GPUs. The first customer is Ineffable Intelligence, a London-based company founded by David Silver, which will use it for large-scale reinforcement learning; WPP also demonstrated a creative agent architecture based on NVIDIA Nemotron models and OpenShell on A5X.

English translation is being prepared.

On October 21, 2026, NVIDIA and Google Cloud announced at GTC Berlin that the first A5X instances equipped with NVIDIA Vera Rubin NVL72 are already running at one customer. The two companies plan to make them generally available in November 2026 and to expand the instances to Europe. A5X combines ConnectX-9 SuperNICs, Google Cloud's Virgo network, and fourth-generation liquid cooling, targeting large-scale agentic inference and reinforcement learning workloads; Google Cloud says it plans to scale this infrastructure to as many as 960,000 Rubin GPUs. The first customer is Ineffable Intelligence, a London-based company founded by David Silver, which will use it for large-scale reinforcement learning; WPP also demonstrated a creative agent architecture based on NVIDIA Nemotron models and OpenShell on A5X.

View story:NVIDIA and Google Cloud announce the launch of the first A5X instances

Nvidia adds Agent Knowledge Foundation to Agent Toolkit

At GTC Berlin, Nvidia added the Agent Knowledge Foundation to its Agent Toolkit. The component packages the Nemotron Retriever models, a new Auto-Ontology library, the Kumo family of models for structured data, and CUDA-X data processing libraries; data remains in place and is governed by enterprises' existing permissions. Nvidia says Nemotron Retriever improved search relevance by 65% in tests in the open-source OpenSearch environment, SAP's fine-tuned reranker runs nearly three times faster, and ServiceNow's reranking offers better quality with nearly 25% lower latency. Auto-Ontology provides a REST API and an MCP server for agents to query. Nvidia says it does not compete with data platforms such as Snowflake, but partners with them. The news was reported by The New Stack; there are no subsequent developments yet.

View story:Nvidia adds Agent Knowledge Foundation to Agent Toolkit

Mistral and NVIDIA partner to develop AI-native engineering workflows

On October 21, 2026, Mistral and NVIDIA announced a partnership to develop AI-native engineering workflows, combining Mistral Large 4 with NVIDIA CUDA-X, Warp, and Omniverse to accelerate engineering simulation and close the loop between simulation, model development, and design. The two parties also plan to launch public benchmarks and open datasets for physical models, and to establish evaluations for agentic AI in CAE and EDA scenarios, examining tool calling, multi-step process completion, and execution time and compute cost. According to information released by Mistral AI, Mistral Large 4 has approximately 1 trillion total parameters and 49 billion activated parameters, was trained on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's data center in Europe, and the weights are planned for release at the end of October. The Decoder reported the same day that the partnership targets industries including aerospace, automotive, and semiconductors, and that Mistral plans to expand compute to 1 gigawatt by 2030. There is currently no further progress.

Mistral and NVIDIA partner to develop AI-native engineering workflows

On October 21, 2026, Mistral and NVIDIA announced a partnership to develop AI-native engineering workflows, combining Mistral Large 4 with NVIDIA CUDA-X, Warp, and Omniverse to accelerate engineering simulation and close the loop between simulation, model development, and design. The two parties also plan to launch public benchmarks and open datasets for physical models, and to establish evaluations for agentic AI in CAE and EDA scenarios, examining tool calling, multi-step process completion, and execution time and compute cost. According to information released by Mistral AI, Mistral Large 4 has approximately 1 trillion total parameters and 49 billion activated parameters, was trained on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's data center in Europe, and the weights are planned for release at the end of October. The Decoder reported the same day that the partnership targets industries including aerospace, automotive, and semiconductors, and that Mistral plans to expand compute to 1 gigawatt by 2030. There is currently no further progress.

View story:Mistral and NVIDIA partner to develop AI-native engineering workflows

English translation is being prepared.

On October 21, 2026, NVIDIA released the OpenShell reference architecture for building, isolating, observing, and enforcing secure agents on enterprise infrastructure and edge systems; the architecture is now available. The architecture uses Vera Rubin to run model inference, Vera CPU to host the agent harness, and BlueField-4 DPU to run Sentry for chip-level policy enforcement; Sentry can isolate out-of-bounds agents at millisecond latency. On the same day, NVIDIA released Sentry, a security software framework that monitors AI agents and enforces access policies within a hardware-isolated trust domain on the BlueField-4 DPU, without relying on host software; the framework is part of the OpenShell reference architecture announced at GTC Berlin and is planned to be opened early next year. Subsequently, Red Hat and Nvidia expanded their collaboration, planning to integrate the OpenShell reference architecture as a native capability into its AI platform, and to provide runtime security controls, model access governance, and hardware support for autonomous AI agents through Red Hat AI Factory with Nvidia.

View story:NVIDIA releases OpenShell reference architecture

Nvidia releases new DLSS 5 model with 24% fewer parameters

Nvidia released a new version of the DLSS 5 model, with 24% fewer parameters than before. Testing published by TechSpot on October 21, 2026 shows that this October update model improves 4K rendering speed on the RTX 50 series by an average of 83% and 1440p by 94%, falling short of the official claim of two times. Taking NBA 2K27 as an example, the RTX 5080's frame rate at 4K rises from 53 FPS to 78 FPS, but there is still a 58% performance loss compared with turning off DLSS 5. The test also noted that the model's image quality is basically on par with the initial version, and issues such as lighting flicker, abnormal skin tones at long distances, and halos around character edges still exist. The model is currently available with the October update, and there are no further evidence-backed developments on its subsequent status.

View story:Nvidia releases new DLSS 5 model with 24% fewer parameters

TestingCatalog to livestream coverage of NVIDIA GTC Berlin keynote

TestingCatalog previews that it will provide live text coverage of the NVIDIA GTC Berlin keynote. A report published on October 21, 2026, said the keynote is about to begin, and TestingCatalog said it will provide live text coverage here, covering updates on GPUs, Nemotron, robotics, and more; the same report also mentioned that the official NVIDIA AI account said entry is now open and a pre-show livestream is underway. There are currently no subsequent developments; the event status remains at the preview stage and the keynote about to begin.

View story:TestingCatalog previews live coverage of NVIDIA GTC Berlin keynote

English translation is being prepared.

NVIDIA releases surgical generative simulation model Cosmos3-H-Dreams

NVIDIA has released the surgical generative simulation model Cosmos3-H-Dreams. The model is specialized from the 16-billion-parameter Cosmos3-H-Embodied-Nano for action-conditioned video generation of cholecystectomy. Official information shows that the model was trained with 28 action dimensions, 832 x 480 video, and 10 Hz action alignment, and was distilled into two student checkpoints, four-step and two-step; the two-step version was trained separately and offers faster inference but lower image quality. NVIDIA also provides the FlashDreams streaming runtime, as well as methods for adapting the model to proprietary data and controllers, and states that the model is a research simulator, not a validated clinical system. There are currently no subsequent development facts.

View story:NVIDIA releases surgical generative simulation model Cosmos3-H-Dreams

NVIDIA Expands Nemotron Retriever Series and Releases Multiple Open Models

On October 21, 2026, NVIDIA expanded the Nemotron Retriever series, releasing multiple open models covering embedding, reranking, and document parsing. The models introduced include Nemotron 3.5 Embed (8B BF16, 1B BF16, 1B NVFP4), Nemotron 3.5 Rerank 8B BF16, and Nemotron 3.5 Parse with 631 million parameters, aimed at multimodal retrieval and document processing for enterprise agent workflows. NVIDIA says that adding Rerank to the agent retrieval pipeline reduces the token cost per query by about 14% (with the 8B embedding model) and 13% (with the 1B model), respectively, while also improving retrieval accuracy. Parse scores 75.20 on ParseBench and processes 10.5 pages per second on a single H100. Open weights, data, and training recipes have been released on Hugging Face and can be deployed via NVIDIA NIM and through Baseten, DeepInfra, and FriendliAI.

Nvidia adds Agent Knowledge Foundation to Agent Toolkit

At GTC Berlin, Nvidia added the Agent Knowledge Foundation to its Agent Toolkit. The component packages the Nemotron Retriever models, a new Auto-Ontology library, the Kumo family of models for structured data, and CUDA-X data processing libraries; data remains in place and is governed by enterprises' existing permissions. Nvidia says Nemotron Retriever improved search relevance by 65% in tests in the open-source OpenSearch environment, SAP's fine-tuned reranker runs nearly three times faster, and ServiceNow's reranking offers better quality with nearly 25% lower latency. Auto-Ontology provides a REST API and an MCP server for agents to query. Nvidia says it does not compete with data platforms such as Snowflake, but partners with them. The news was reported by The New Stack; there are no subsequent developments yet.

TestingCatalog previews live coverage of NVIDIA GTC Berlin keynote

TestingCatalog previews that it will provide live text coverage of the NVIDIA GTC Berlin keynote. A report published on October 21, 2026, said the keynote is about to begin, and TestingCatalog said it will provide live text coverage here, covering updates on GPUs, Nemotron, robotics, and more; the same report also mentioned that the official NVIDIA AI account said entry is now open and a pre-show livestream is underway. There are currently no subsequent developments; the event status remains at the preview stage and the keynote about to begin.

Evan Conrad says NVIDIA cannot build its own compute trading market

On October 21, 2026, SF Compute CEO Evan Conrad argued on the Latent Space podcast that NVIDIA cannot build its own compute trading market. His reasoning is that companies with large hardware or inference businesses find it difficult to operate a neutral exchange. In the same appearance, Conrad also said that the compute market must first achieve physical delivery—that is, actually delivering usable GPU compute—before discussing financialization; otherwise, market manipulation and basis risk are likely to arise. He explained that SF Compute is expanding from subleasing GPU clusters to building its own cloud operations capability in order to achieve physical delivery of GPU contracts, so that buyers receive usable compute rather than cash settled against an index; the company does not hold hardware in order to remain neutral, and it already supports customers in partitioning InfiniBand fabric online themselves. He also said that SF Compute has placed 5% of its capacity on the order book, that customers who spend $100 to buy capacity can on average recover $25 by reselling it, and that hundreds of millions of dollars in volume has already been traded. There are currently no subsequent developments.

Nvidia releases new DLSS 5 model with 24% fewer parameters

Nvidia released a new version of the DLSS 5 model, with 24% fewer parameters than before. Testing published by TechSpot on October 21, 2026 shows that this October update model improves 4K rendering speed on the RTX 50 series by an average of 83% and 1440p by 94%, falling short of the official claim of two times. Taking NBA 2K27 as an example, the RTX 5080's frame rate at 4K rises from 53 FPS to 78 FPS, but there is still a 58% performance loss compared with turning off DLSS 5. The test also noted that the model's image quality is basically on par with the initial version, and issues such as lighting flicker, abnormal skin tones at long distances, and halos around character edges still exist. The model is currently available with the October update, and there are no further evidence-backed developments on its subsequent status.

NVIDIA releases real-time surgical world model Cosmos3-H-Dreams

View story:English translation is being prepared.

NVIDIA releases surgical generative simulation model Cosmos3-H-Dreams

NVIDIA has released the surgical generative simulation model Cosmos3-H-Dreams. The model is specialized from the 16-billion-parameter Cosmos3-H-Embodied-Nano for action-conditioned video generation of cholecystectomy. Official information shows that the model was trained with 28 action dimensions, 832 x 480 video, and 10 Hz action alignment, and was distilled into two student checkpoints, four-step and two-step; the two-step version was trained separately and offers faster inference but lower image quality. NVIDIA also provides the FlashDreams streaming runtime, as well as methods for adapting the model to proprietary data and controllers, and states that the model is a research simulator, not a validated clinical system. There are currently no subsequent development facts.