# NVIDIA GTC Berlin 2026

> Source: https://aiwiki.ai/wiki/nvidia_gtc_berlin_2026
> Updated: 2026-10-25
> Fact-checked: 2026-10-25
> Categories: AI Events, AI Infrastructure, NVIDIA, Physical AI
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "NVIDIA GTC Berlin 2026." aiwiki.ai, 25 Oct 2026. https://aiwiki.ai/wiki/nvidia_gtc_berlin_2026
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

**NVIDIA GTC Berlin 2026** was the European edition of [NVIDIA](https://aiwiki.ai/wiki/nvidia)'s [GPU Technology Conference](https://aiwiki.ai/wiki/gtc) (GTC), held in Berlin, Germany, from October 20 to 22, 2026.[1] Workshops opened the event on October 20, NVIDIA founder and CEO [Jensen Huang](https://aiwiki.ai/wiki/jensen_huang) gave the keynote at the Tempodrom on October 21, and more than 100 conference sessions on agentic AI, robotics, physical AI, open models and accelerated computing ran at STATION-Berlin.[2][29] NVIDIA published eight press releases datelined "GTC Berlin" on October 21, plus a series of blog posts, collected in an online press kit.[3] The headline items were the containerized [NVIDIA DSX POD](https://aiwiki.ai/wiki/nvidia_dsx_pod) AI factory design, the [NVIDIA OpenShell Reference Architecture](https://aiwiki.ai/wiki/nvidia_openshell_reference_architecture) for running [AI agents](https://aiwiki.ai/wiki/agentic_ai) under hardware-enforced controls, an expansion of the Isaac robotics platform built around the Agile ONE SN reference humanoid, and supply-chain capacity plans with six manufacturers.[3][4][6][10][11]

The conference closed a week-long tour of Europe in which NVIDIA announced planned startup investments in France, the United Kingdom and Germany and said it had committed "nearly $10 billion" to the European technology ecosystem.[2][28][30] Much of the Berlin material was framed around Europe: sovereign AI clouds, European industrial customers, open models that can be adapted to local languages and data, and Huang's argument that Europe should be "producing the AI" rather than only consuming it.[2][13][17] Protesters, holding a banner with a Greenpeace logo, interrupted the keynote about 20 minutes in.[36][38]

| Field | Detail |
|---|---|
| Organizer | NVIDIA[1] |
| Dates | October 20-22, 2026 (workshops October 20; keynote October 21; event wrapped up October 22)[1][29] |
| Keynote | Jensen Huang, October 21, 2026, 11 a.m. to 1 p.m. CEST, Tempodrom, Berlin[2][12] |
| Conference venue | STATION-Berlin, with more than 100 sessions[2] |
| Pregame show | Hosted by Deirdre Bosa and Gavin Baker from the Tempodrom, starting 9 a.m. CEST[29] |
| Press releases | 8, all dated October 21, 2026[3] |
| Main themes | AI factories, agent safety, open models, physical AI and robotics, telecom, healthcare, autonomous vehicles[2][3] |

## Background

GTC is NVIDIA's developer and product conference; the main GTC article describes its history and its regional editions. For 2026, NVIDIA billed its European edition as GTC Berlin and ran a rolling coverage page, "NVIDIA GTC Berlin: Live Updates on What's Next in AI," covering October 12 to 24, which grouped the pre-conference tour stops with the Berlin announcements.[1]

## Road to GTC Berlin

In the week before the conference, Huang and other NVIDIA executives visited Barcelona, Paris, London, Dublin and Stockholm before arriving in Berlin. NVIDIA's coverage page called the Barcelona stop "GTC Berlin's First Stop."[1] The keynote recap described this as "NVIDIA's weeklong road across Europe."[2]

| Date (2026) | Location | Event or announcement |
|---|---|---|
| October 12 | France | NVIDIA announced planned investments in four startups based in France: Doctolib, Entalpic, [Genesis AI](https://aiwiki.ai/wiki/genesis_ai) and Pigment. The same post said NVIDIA had committed nearly $10 billion across Europe's technology ecosystem.[30] |
| October 13 | Barcelona and Paris | Build-a-Claw developer events at the Palacio de Pedralbes in Barcelona and the Pavillon Vendôme in Paris, where attendees built robotic claw machines with open models. That evening Huang spoke at NVIDIA's France AI Ecosystem Reception and then met French President Emmanuel Macron at the Élysée Palace.[1] |
| October 13 | United Kingdom | NVIDIA announced planned investments in Boltz, Foresight, Hiverge, Prima Mente and Riverlane, and referred to its five-year, $250 million compute commitment for UK agentic safety and security research.[1][31] |
| October 14 | London and Dublin | Huang attended a Build-a-Claw event, the AI Safety and Security Research Forum and NVIDIA's U.K. AI Ecosystem Reception in London. In Dublin, NVIDIA's Deepu Talla spoke at the International AI Summit.[1] |
| October 15 | Stockholm | A Build-a-Claw event at the Norrsken Foundation, a discussion on the Sferical One shared supercomputer, and the Sweden AI Ecosystem Reception. NVIDIA's coverage noted that Volvo Cars announced an expanded automated-driving partnership with NVIDIA the same day.[1] |
| October 17 | Berlin | A Build-a-Claw event at NLND Berlin that NVIDIA said drew more than 500 people, the last leg of the European Build-a-Claw tour before the conference. The [RTX Spark](https://aiwiki.ai/wiki/nvidia_rtx_spark) Windows laptops had gone on sale on October 16.[1][19] |
| October 19 | Berlin and Erlangen | As part of the Invest in Germany Summit, Chancellor Friedrich Merz and two federal ministries signed a Joint Declaration of Intent with NVIDIA setting out a joint initiative on AI models for physical AI. Merz and Huang then held an onstage conversation at Kraftwerk, a former power station in Berlin, and Huang announced planned investments in [Agile Robots](https://aiwiki.ai/wiki/agile_robots), Avelios Medical, Mirelo and Reflex Aerospace. [Siemens](https://aiwiki.ai/wiki/siemens) and NVIDIA also showed industrial AI applications at Siemens' Erlangen electronics factory.[1][28] |
| October 20 | Berlin | Full-day workshops opened the conference. NVIDIA also published a post on European cloud providers building with its [Vera CPU](https://aiwiki.ai/wiki/nvidia_vera_cpu).[1][20] |

Table dates follow the U.S. Pacific dates on NVIDIA's blog bylines; the France and UK investment posts went live in Europe on the mornings of October 13 and October 14, and Huang presented the UK investments at the London reception on October 14.[1][30][31] Further detail on the France, UK and Germany investment plans is in the main NVIDIA article.

## Keynote

Huang's keynote ran on October 21 at the Tempodrom, a Berlin event venue near Anhalter Bahnhof.[2] NVIDIA's recap blog by Brian Caulfield is the main published account; NVIDIA also posted the keynote deck in its press kit.[2][3]

### Europe and the AI factory

Huang opened by naming European companies working with NVIDIA on AI, among them L'Oréal, Revolut, Spotify, Merck, Vodafone, Zara, Porsche, Nestlé and HSBC, and said "we are now at the beginnings of a new industrial revolution." He described AI as a "five-layer cake": energy powers computers in an AI factory, which generates tokens from the models running on it, and those models are built into applications. "To be part of this AI economy, you don't want to just be consuming the AI; you want to be producing the AI," he said.[2]

He described [Vera Rubin](https://aiwiki.ai/wiki/nvidia_vera_rubin) as "in full production" and then announced DSX POD, a reference design for complete Vera Rubin AI factories that partners build inside modular containers. NVIDIA said onsite deployment could shrink from six months to 14 days once site utilities are ready.[2][4] According to The Next Platform, Huang argued that idle capital was the issue: he put the cost of a 1-gigawatt data center at about $50 billion, said operators expect about $40 billion in revenue from it, and said a three-month delay costs about $10 billion of that revenue.[36] He also claimed NVIDIA had improved AI performance "by 1 million times" over the previous ten years.[36]

### Agents and safety

Huang described an agent as five parts: a model, a harness that runs its work and memory, a sandbox that isolates the harness, knowledge, and tools. "The most important thing is to make sure the agent is secure," he said, adding that "every single time an agent is launched, the privileges are different."[2] He then introduced the OpenShell Reference Architecture, built on the open source [NVIDIA OpenShell](https://aiwiki.ai/wiki/nvidia_openshell) runtime.[2][6]

He also introduced the NVIDIA Agent Knowledge Foundation, a set of models and libraries for connecting agents to company data, including Nemotron Retriever tools and new NVIDIA Kumo models for predictions from structured data. "What turns these super agents, these agents into super agents, is that you give them access to your company's knowledge," he said.[2] He called CUDA-X libraries "the tools of the trade" for agents and said more than 11 million developers build on NVIDIA's accelerated computing platform.[2]

A video showed NVIDIA Agent Toolkit in three settings: [Lila Sciences](https://aiwiki.ai/wiki/lila_sciences) agents helping design a potential B-cell lymphoma therapy through virtual experiments, Siemens agents preparing new products for production at the Erlangen factory, and Dassault Systèmes engineering agents refining a vehicle design.[2]

### Open models

Huang listed NVIDIA's open model families: [Nemotron](https://aiwiki.ai/wiki/nemotron) for language, [Cosmos](https://aiwiki.ai/wiki/nvidia_cosmos) world foundation models, [GR00T](https://aiwiki.ai/wiki/isaac_gr00t) for humanoid robots, [Alpamayo](https://aiwiki.ai/wiki/nvidia_alpamayo_2_super) for autonomous vehicles, [Ising](https://aiwiki.ai/wiki/nvidia_ising) for quantum-classical computing and [Earth-2](https://aiwiki.ai/wiki/nvidia_earth_2) for physics AI. He said companies could post-train these models on their own data to create what he called "sovereign intelligence." "We are the frontier," he said.[2]

### Telecom, healthcare and vehicles

Huang said telecom networks could become distributed AI infrastructure and that NVIDIA had made its first commercial phone call on T-Mobile's network with Nokia "after nearly eight years of work." "Now CUDA makes phone calls," he said.[2] He presented NVIDIA's work with Siemens Healthineers on AI-enabled adaptive radiotherapy, which he described as a physical-AI application because the system has to track a tumor that changes shape during treatment.[2] On vehicles, he said the [NVIDIA Hyperion](https://aiwiki.ai/wiki/nvidia_drive_hyperion) platform was being designed into more fleets and announced that Volkswagen's van, JLR, Volvo and Mercedes-Benz's next-generation vehicles would be Hyperion-compatible: "Welcome to Hyperion."[2]

### Robotics and RTX Spark

Huang announced [Agile ONE SN](https://aiwiki.ai/wiki/agile_robots_se_agile_one), a reference humanoid built on Isaac GR00T with a 31-degree-of-freedom body, which NVIDIA said would cut researcher setup time "from weeks to days."[2] The Next Platform quoted him saying it "will be available in the first quarter of next year"; NVIDIA's press release says "early 2027."[10][36] NVIDIA said partners including [Agility Robotics](https://aiwiki.ai/wiki/agility_robotics), [Figure](https://aiwiki.ai/wiki/figure_ai), [KUKA](https://aiwiki.ai/wiki/kuka) and [NEURA Robotics](https://aiwiki.ai/wiki/neura_robotics) build on the same open physical AI stack.[2]

He closed with RTX Spark. Comparing the laptop's 1 petaflop and 128GB of memory with the [DGX-1](https://aiwiki.ai/wiki/nvidia_dgx) he delivered to OpenAI in 2016, he said "the only difference, of course, is that I can hold this like this, and DGX-1 took two people to carry upstairs." A video showed an agent training Hugging Face's [Microduck](https://aiwiki.ai/wiki/microduck) robot in [Isaac Sim](https://aiwiki.ai/wiki/nvidia_isaac_sim), and four Microducks then appeared on stage.[2] The keynote ended with a Eurovision-style musical recap with a "We are Europa" chorus, and Huang told the audience: "Let's work together to build Europe's AI economy and put AI to work."[2]

### Protest

About 20 minutes into the keynote, while Huang was showing Vera Rubin hardware, protesters stood on chairs and held up a banner with a Greenpeace logo, The Next Web reported, citing Tom's Hardware and NVIDIA's livestream.[38] According to the reports, the protesters raised "environmental destruction" and asked about reports that NVIDIA chips had been found in Russian drones used against Ukraine.[36][38] Huang told security staff to go easy on them and said, "All important issues, but let me get through my keynote," as they were led out.[36][38] The Next Platform described the activists as "thought to be from Greenpeace."[36] NVIDIA's own live blog of the keynote does not mention the interruption.[2][38]

## Announcements

The table lists the October 20-22 items that NVIDIA filed under GTC Berlin in its press kit and coverage page.[1][3] "PR" marks a press release; "Blog" marks an NVIDIA blog post.

| Date (2026) | Item | Area | Format | Source |
|---|---|---|---|---|
| Oct 20 | European cloud providers and server makers building with Vera CPUs | Infrastructure | Blog | [20] |
| Oct 21 | [DSX POD](https://aiwiki.ai/wiki/nvidia_dsx_pod) containerized Vera Rubin AI factory reference design | Infrastructure | PR | [4] |
| Oct 21 | European partners adopt the [NVIDIA DSX](https://aiwiki.ai/wiki/nvidia_dsx) AI Factory Platform; SimReady Central Library; new DSX OS components | Infrastructure | PR | [5] |
| Oct 21 | Supply-chain capacity plans with Amkor, IBIDEN, KYEC, Samsung Electro-Mechanics, SPIL and Unimicron | Manufacturing | PR | [11] |
| Oct 21 | [OpenShell Reference Architecture](https://aiwiki.ai/wiki/nvidia_openshell_reference_architecture) for safe AI agents | Agents | PR | [6] |
| Oct 21 | Agent Toolkit expansion; Agent Knowledge Foundation; Auto-Ontology library | Agents | Blog | [14] |
| Oct 21 | Confidential Bare Metal and expanded NVIDIA Confidential Computing | Security | Blog | [15] |
| Oct 21 | Open model portfolio: Nemotron, Nemotron Retriever, Kumo, Cosmos, Alpamayo 2 Edge, BioNeMo, Ising | Models | Blog | [13] |
| Oct 21 | Cosmos training tools, simulation checkpoints and Cosmos Data Factory blueprints | Physical AI | Blog | [21] |
| Oct 21 | Isaac platform expansion: Agile ONE SN, Isaac OS, Isaac Lab 3.0 general availability, Human 1M Dataset | Robotics | PR | [10] |
| Oct 21 | Hyperion adoption by European mobility companies; Halos SAFE; Alpamayo 2 Edge | Automotive | PR | [9] |
| Oct 21 | Telecom operators and Nokia build distributed AI infrastructure; new ARC configurations | Telecom | PR | [7] |
| Oct 21 | RFD4-Proteina open protein design model; Siemens Healthineers adaptive radiotherapy work | Healthcare | PR | [8] |
| Oct 21 | cuDSS sparse solver library reaches general availability | Simulation | Blog | [16] |
| Oct 21 | Factory Operations Blueprint (FOX) in early access; European industrial customer stories | Manufacturing | Blog | [12] |
| Oct 21 | Google Cloud A5X Vera Rubin instances | Cloud | Blog | [18] |
| Oct 21 | European NVIDIA Cloud Partners expand capacity and support OpenShell | Cloud | Blog | [17] |
| Oct 21 | RTX Spark, DGX Spark and DGX Station software and hardware updates | Local AI | Blog | [19] |
| Oct 21 | Companies using DSX to plan grid and AI factory together | Infrastructure | Blog | [27] |
| Oct 21 | Financial services, retail, public sector and Earth-2 weather customer posts | Industries | Blog | [23][24][25][26] |
| Oct 22 | [NVIDIA Minerva](https://aiwiki.ai/wiki/nvidia_minerva) AIOps platform | Infrastructure | Blog | [22] |

### AI factories and supply chain

**DSX POD.** NVIDIA described DSX POD as a reference design for complete Vera Rubin AI factories built by partners in modular containers and validated before delivery, cutting deployment time "from six months to as little as 14 days." It named 17 partners, including Dell Technologies, HPE, Lenovo, Schneider Electric and Vertiv, to deliver the units worldwide in early 2027.[4] CNBC reported that each unit holds seven Vera Rubin racks, weighs about 28,000 kilograms and is designed to travel by truck on normal roads; NVIDIA did not announce a price.[35] The design has its own article.

**DSX platform.** A separate release said European partners were adopting the DSX AI Factory Platform to cut water used for cooling and to deliver the same AI output with less power.[5] NVIDIA said dry-cooler designs with Vera Rubin's 100% liquid cooling at coolant temperatures up to 45 degrees Celsius could cut facility cooling water to near zero, and that at 2-gigawatt scale DSX could save 54 million liters of water a day compared with cooling towers. It said the [DSX MaxLPS](https://aiwiki.ai/wiki/nvidia_dsx_maxlps) power-allocation software could generate the same number of tokens with up to 30% less power. In NVIDIA's account, Nscale measured 49% higher inference throughput and 37% more managed GPU capacity within a 264-kilowatt power budget across four GB300 NVL72 racks in Iceland, and Lambda reported 24% more cluster token throughput in a B200 proof of concept.[5] These are vendor and partner figures. The release also launched the SimReady Central Library of simulation-ready equipment assets and added three DSX OS components in developer preview: Config Manager, Rack Management Service and Cluster Readiness Engine.[5]

**Supply chain.** NVIDIA announced capacity plans with Amkor Technology, IBIDEN, King Yuan Electronics (KYEC), Samsung Electro-Mechanics, Siliconware Precision Industries (SPIL) and Unimicron. It said it plans to increase substrate capacity by 2.5x over three years; Amkor is targeting a more than 300% increase in interposer wafer capacity in South Korea by the end of 2028; a $1.1 billion capacity reservation advance to KYEC will support new U.S. testing facilities, with high-volume production targeted for April 2027; and SPIL is targeting a 15x increase in optical-engine capacity for co-packaged optics by mid-2027. "Demand is extraordinary. Capacity is the constraint," Huang said in the release.[11]

**Minerva.** On October 22, NVIDIA launched Minerva, an AIOps platform that combines telemetry from GPUs, networking, storage, schedulers and workloads and uses AI models to find root causes in AI factories.[22][33] See [NVIDIA Minerva](https://aiwiki.ai/wiki/nvidia_minerva).

**Cloud and CPUs.** NVIDIA's cloud-partner post named Nscale, Crusoe, Nebius, Mistral, Verda and others as expanding NVIDIA-based capacity across Europe, the Nordics, Africa and Central Asia, and described model-service work by Fireworks (with Era4) and Baseten. It said Mistral was expanding its infrastructure with Vera Rubin systems planned for delivery in 2027, and that Nebius had adopted NVIDIA Groq 3 LPX alongside Vera Rubin NVL72 on Nebius Token Factory.[17] Google Cloud and NVIDIA said A5X instances built on Vera Rubin NVL72 would reach general availability the following month, with planned expansion to Europe, and that Google Cloud plans to scale its Vera Rubin infrastructure to as many as 960,000 Rubin GPUs. [Ineffable Intelligence](https://aiwiki.ai/wiki/ineffable_intelligence), the London company founded by [David Silver](https://aiwiki.ai/wiki/david_silver), was named as the first A5X customer.[18] A post published on October 20 said Nebius, Nscale and Verda were building with Vera CPUs.[20]

**Confidential computing.** NVIDIA added Confidential Bare Metal as a third deployment option for NVIDIA Confidential Computing, alongside confidential containers and confidential virtual machines, so that model builders can run sensitive workloads on dedicated third-party infrastructure. Cohere and Google Cloud described plans to use it.[15]

### Agents and enterprise software

**OpenShell Reference Architecture.** NVIDIA's design runs models on Vera Rubin, agent harnesses on Vera CPUs inside OpenShell, and NVIDIA Sentry security software on [BlueField](https://aiwiki.ai/wiki/bluefield)-4 DPUs. Huang summarized it as "agents can think on Rubin, act on Vera and be governed on BlueField." NVIDIA said the architecture was available immediately and listed model builders, cybersecurity vendors, banks, enterprise software companies and cloud providers working with it.[6][32] Its dedicated article covers the design in detail.

**Agent Toolkit and Agent Knowledge Foundation.** NVIDIA said its open source [Agent Toolkit](https://aiwiki.ai/wiki/nvidia_agent_toolkit) now comprises the new Agent Knowledge Foundation with CUDA-X libraries, Nemotron Retriever and Kumo models, the OpenShell runtime, [NemoClaw](https://aiwiki.ai/wiki/nvidia_nemoclaw) secure agent blueprints, and tools such as NeMo Relay and NeMo Compass. The foundation includes a new Auto-Ontology library that maps raw company data to business concepts such as customers, products and orders. NVIDIA said its own supply chain organization uses the foundation through Palantir Foundry.[14]

### Open models

NVIDIA's October 21 open-models post said its portfolio includes the new [Nemotron 3.5 Super VL](https://aiwiki.ai/wiki/nemotron_3_5_super_vl) and [Nemotron 3.5 Bolt](https://aiwiki.ai/wiki/nemotron_3_5_bolt) models, new Nemotron Retriever models for embedding, reranking and document parsing, the new Kumo family of open-weight foundation models for structured business data, Cosmos 3.5 world models, Alpamayo 2 Edge, BioNeMo and Ising.[13] The post said the smaller [Nemotron 3.5 Bolt E4B 4GB](https://aiwiki.ai/wiki/nemotron_3_5_bolt_e4b) was expected to be available on October 23, and NVIDIA released [Nemotron 3.5 Bolt E4B](https://aiwiki.ai/wiki/nemotron_3_5_bolt_e4b) that day.[13] Kumo consists of Kumo Structured models for relational and tabular prediction, forecasting and anomaly detection, plus the Kumo Engine library stack; NVIDIA named Cognite, Coinbase, Databricks, DoorDash, Revolut and Tubi among its first users.[13] NVIDIA said RFD4-Proteina, developed with the Baker Lab at the University of Washington's Institute for Protein Design, generates designs about 13x faster than RFdiffusion3 in a benchmark of 200-amino-acid proteins.[13]

The Cosmos post added five Cosmos Data Factory blueprints (autonomous-vehicle dataset generation, image attribute augmentation, event video generation, defect image generation and video data augmentation), extended Cosmos Framework training and optimization tools to other open models, and introduced new Cosmos 3 Nano variants for simulation.[21]

### Robotics

The Isaac release introduced Agile ONE SN, a reference humanoid from Agile Robots preconfigured with Isaac GR00T and powered by [Jetson Thor](https://aiwiki.ai/wiki/jetson_thor). NVIDIA specified an Agile ONE S body 1.40 meters tall and weighing 40 kilograms with 31 degrees of freedom and a 3-kilogram payload per arm, dual WUJI Hand 2 hands with 20 degrees of freedom each, and Jetson Thor compute rated at 2,070 FP4 teraflops with 128GB of memory, with availability scheduled for early 2027.[10] The same release made [Isaac Lab](https://aiwiki.ai/wiki/isaac_lab) 3.0 generally available, introduced Isaac OS as an open source runtime framework built on Holoscan, and released the SimReady Central Library of more than 2,000 assets and the Human 1M Dataset on Hugging Face, with the full Human 1M Dataset limited to Cosmos Coalition partners.[10] It also announced three open source reference workflows for [DGX Station](https://aiwiki.ai/wiki/nvidia_dgx_station), including one for refining GR00T 1.7 models with reinforcement learning.[10]

European robotics companies named in the release included ABB Robotics, KUKA, Sereact, Hexagon Robotics, Humanoid, Flexion, Franka Robotics and NEURA Robotics, which NVIDIA said uses Isaac Sim and Isaac Lab to generate synthetic data and train skills for its Neuron AI foundation model.[10] NVIDIA's coverage page later ran a feature, "Robots Steal the Show at GTC Berlin," on the humanoids, Microducks and manipulators at the event.[1]

### Automotive

NVIDIA said Bolt with Lucid, Mercedes-Benz, Volkswagen Commercial Vehicles and Einride were showcasing autonomous-mobility use cases on Hyperion, and that Volvo Cars was adopting Hyperion and the Alpamayo Stack software.[9] Bolt selected Alpamayo Stack for its European robotaxi program and plans, subject to regulatory approval, to begin gathering driving data in Madrid in 2027, as part of a planned European rollout of at least 25,000 level 4-capable future Lucid vehicles built on Hyperion.[9] NVIDIA also introduced Halos SAFE (Safety Assurance Framework for Evaluation) within the [NVIDIA Halos](https://aiwiki.ai/wiki/nvidia_halos) safety system, and Alpamayo 2 Edge, a 2-billion-parameter reasoning model built on Cosmos and optimized for DRIVE AGX Thor.[9]

### Telecom

NVIDIA said T-Mobile had completed the world's first AI-RAN call on a live network in Bellevue, Washington, with NVIDIA and Nokia, and that nine operators across Europe, the Middle East and Africa were building or expanding NVIDIA-powered infrastructure.[7] Deutsche Telekom plans to expand its NVIDIA-powered Industrial AI Cloud, which serves robotics companies such as Agile Robots, and a GTC Berlin demonstration with Agile Robots split robot intelligence across the robot, a distributed edge model and central infrastructure.[7] NVIDIA also introduced two Aerial RAN Computer configurations, MGX 2U and MGX 4U, using RTX PRO 6000 Blackwell Server Edition GPUs.[7]

### Healthcare and life sciences

NVIDIA and the UW Medicine Institute for Protein Design released RFD4-Proteina as a BioNeMo open model for protein design. The release said it achieved higher computational design success rates than evaluated baselines across four benchmark suites, according to a research draft, and that it was available through partners including Hugging Face, Google Cloud, Anthropic's Claude Science and OpenAI.[8] Siemens Healthineers is working with NVIDIA on adaptive radiotherapy; the release said Nemotron models were being explored for agentic co-planning, which could be combined with CUDA-accelerated dose calculation and the Holoscan platform.[8] David Baker, who won the 2024 Nobel Prize in Chemistry, discussed the model in a video NVIDIA posted on October 22.[1]

### Engineering simulation

NVIDIA released cuDSS, a [CUDA](https://aiwiki.ai/wiki/cuda)-X library that moves the solving of large sparse systems in engineering and scientific simulations onto GPUs, in general availability. Its post listed reported results from Applied Materials (up to 10x faster sparse solves in Ginestra), ZEISS SMT using COMSOL (a measured 4x speedup on an EUV lithography plasma benchmark) and Dassault Systèmes (an 8-10x speedup in Abaqus and its electromagnetic solvers), plus integrations at Cadence, Siemens and Synopsys. These are company-reported figures.[16]

## Industrial and robotics customer stories

NVIDIA's October 21 post "Europe's Industrial Leaders Reinvent Engineering and Manufacturing With NVIDIA," by James McKenna, described how European industrial companies use CUDA-X and [Omniverse](https://aiwiki.ai/wiki/nvidia_omniverse) libraries and the Cosmos and Nemotron open models.[12] NVIDIA's @NVIDIAAI account promoted it in a post naming Agile Robots, Audi, Dassault Systèmes, Schaeffler and Siemens.[34] The same post said the NVIDIA Factory Operations Blueprint (FOX), a reference for building factory manager agents and specialist agents for inspection, root-cause analysis, report generation and fleet management, was available in early access on GitHub.[12]

The table restates what NVIDIA's post says about each company. The figures come from NVIDIA and the companies, not independent measurement.

| Company | What NVIDIA says the company is doing |
|---|---|
| Siemens | Integrating agents built with FOX into Intelligence Center X, its agentic enterprise system, to help find root causes and coordinate factory operations, for estimated cost savings of 20%. Its Process Preparation X agents use Cosmos Data Factory tools to generate defect images for circuit-board inspection; at the Erlangen electronics factory this cut calibration time of inspection processes by 40%.[12] |
| Mercedes-Benz | Uses Siemens' Simcenter STAR-CCM+ on NVIDIA GPUs for faster aerodynamic simulations, and applies Cosmos 3 and the Metropolis VSS blueprint to build agents for vehicle inspection and windshield assembly verification.[12] |
| Cadence | Benchmarking automotive fluid-dynamics workloads on NVIDIA GPUs.[12] |
| Autodesk | Bringing the PhysicsNeMo and Warp frameworks into NavPack, integrated with Autodesk Alias, to help designers such as those at Bentley predict vehicle performance.[12] |
| imec and Infineon | Accelerating semiconductor simulation with Synopsys software; S-Litho benchmarks on imec's model showed up to 40x speedup, and Infineon reached a 12x speedup in 3D breakdown simulation with Sentaurus Device.[12] |
| Leonardo | Working with Neural Concept on AI-driven aircraft design using Omniverse libraries.[12] |
| Dassault Systèmes | Car manufacturers use its LEO and AURA Virtual Companions, powered by Nemotron, orchestrating CATIA and SIMULIA AI agents with SIMULIA solvers accelerated by CUDA-X to design and virtually test vehicle doors.[12] |
| UniversalAGI and Saab | Using UniversalAGI's physics AI foundation models with NVIDIA Warp to evaluate thousands of wave conditions in real time, so vessels can anticipate dangerous motion and select safer routes.[12] |
| PhysicsX and GB1 | AI-powered yacht design for Britain's America's Cup team, with Omniverse libraries to visualize hull geometry and pressure fields.[12] |
| Fermioniq, Deltares | Fermioniq's Sofia agent helps Deltares and Dutch ministries model river flow by translating natural-language requests into solvers built on Warp.[12] |
| [Mistral AI](https://aiwiki.ai/wiki/mistral_ai) | Plans to use CUDA-X, Warp and Omniverse libraries to accelerate physics simulation and generate training data for engineering AI models.[12] |
| Audi | Uses a battery-production [digital twin](https://aiwiki.ai/wiki/digital_twin) built with Omniverse libraries to test production concepts before equipment exists, and is working with Dassault Systèmes and Agile Robots to simulate and validate mobile humanoid robots and logistics processes in DELMIA virtual twins built with Omniverse libraries.[12] |
| Agile Robots | Named in the Audi project above.[12] Separately, the Isaac release named it as the maker of Agile ONE SN.[10] |
| Schaeffler | Integrating digital twins of its humanoid actuators with Isaac simulation frameworks so robotics partners can evaluate components in simulation.[12] The Isaac release added that Schaeffler is contributing its actuator technology as an exclusive foundational component for [Newton](https://aiwiki.ai/wiki/nvidia_newton).[10] |
| Wandelbots | At Volkswagen Sachsen's Transparent Factory, uses NOVA with Isaac Sim to validate assembly workflows digitally.[12] |
| Falkor (KONGSBERG) | Integrating agents and Omniverse libraries into its platform, running on Microsoft Azure NCv6 virtual machines with RTX PRO 6000 Blackwell Server Edition GPUs, for inspection and maintenance in process and energy industries.[12] |
| Schneider Electric and AVEVA | Developing an AI factory digital twin for [Nscale](https://aiwiki.ai/wiki/nscale) with Omniverse libraries to compare power, cooling and equipment configurations.[12] |

Other October 21 posts covered European financial institutions such as Revolut, Starling Bank, Lloyds Banking Group and Santander;[23] retailers such as Shopify, Zooplus, L'Oréal and Otto Group;[24] public-sector projects including deepset's work on the European Commission's AI@EC platform and Polish government systems;[26] and Earth-2 weather models, including a new StormScope EU model trained on European data and shown in a live demo at the event.[25]

## Partner announcements

Several partners timed their own announcements to the event. CoreWeave announced the [CoreWeave Zero Trust AI Framework](https://aiwiki.ai/wiki/coreweave_zero_trust_ai_framework) in a release datelined Berlin, October 21, saying it was "unveiled at NVIDIA GTC Berlin."[39] Dell Technologies issued a release datelined NVIDIA GTC in Berlin, October 21, on Dell AI Factory with NVIDIA updates, including a security architecture that uses the OpenShell Reference Architecture.[40]

## Coverage

Independent reporting focused on DSX POD, OpenShell and Agile ONE SN. CNBC's Kif Leswing described DSX POD as NVIDIA's "latest effort to move up the stack" and reported analyst estimates that a single Vera Rubin rack can cost more than $7 million.[35] Yahoo Finance's Daniel Howley led with OpenShell and the Agile Robots humanoid.[37] The Next Platform's Joe Fay called DSX POD "the hardware highlight" of the keynote and noted that dry cooling and 800-volt DC power were already part of the standard DSX architecture.[36]

## References

1. NVIDIA Writers. "NVIDIA GTC Berlin: Live Updates on What's Next in AI." NVIDIA Blog, October 12-24, 2026. https://blogs.nvidia.com/gtc-berlin-2026-news/
2. Brian Caulfield. "GTC Berlin 2026: Live Updates From NVIDIA CEO Jensen Huang's Keynote." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/gtc-berlin-2026-keynote/
3. NVIDIA Newsroom. "GTC Berlin 2026 News" (online press kit). Accessed October 25, 2026. https://nvidianews.nvidia.com/online-press-kit/gtc-berlin-2026-news
4. NVIDIA. "NVIDIA Introduces DSX POD AI Factories to Accelerate Europe's Infrastructure Deployment." Press release, October 21, 2026. https://nvidianews.nvidia.com/news/nvidia-introduces-dsx-pod-ai-factories-to-accelerate-europes-infrastructure-deployment
5. NVIDIA. "European Ecosystem Builds World's Most Efficient AI Factories With NVIDIA DSX." Press release, October 21, 2026. https://nvidianews.nvidia.com/news/european-ecosystem-builds-worlds-most-efficient-ai-factories-with-nvidia-dsx
6. NVIDIA. "NVIDIA Launches OpenShell Reference Architecture for Safe AI Agents." Press release, October 21, 2026. https://nvidianews.nvidia.com/news/openshell-reference-architecture
7. NVIDIA. "NVIDIA, Nokia and Leading Telecom Operators Turn Networks Into Distributed Intelligence Platforms." Press release, October 21, 2026. https://nvidianews.nvidia.com/news/nvidia-nokia-and-leading-telecom-operators-turn-networks-into-distributed-intelligence-platforms
8. NVIDIA. "Healthcare and Life Sciences Leaders Build on NVIDIA to Advance Drug Discovery and Expand Patient Care." Press release, October 21, 2026. https://nvidianews.nvidia.com/news/healthcare-and-life-sciences-leaders-build-on-nvidia-to-advance-drug-discovery-and-expand-patient-care
9. NVIDIA. "European Mobility Leaders Build Safe Vehicles and Robotaxi Fleets on NVIDIA Hyperion." Press release, October 21, 2026. https://nvidianews.nvidia.com/news/european-mobility-leaders-build-safe-vehicles-and-robotaxi-fleets-on-nvidia-hyperion
10. NVIDIA. "NVIDIA Advances Isaac Platform for General-Purpose Robotics With New Reference Humanoid, Frameworks, Simulation Pipelines and Data." Press release, October 21, 2026. https://nvidianews.nvidia.com/news/nvidia-advances-isaac-platform-for-general-purpose-robotics-with-new-reference-humanoid-frameworks-simulation-pipelines-and-data
11. NVIDIA. "NVIDIA and Supply Chain Partners Expand Capacity to Accelerate AI Infrastructure Buildout." Press release, October 21, 2026. https://nvidianews.nvidia.com/news/nvidia-and-supply-chain-partners-expand-capacity-to-accelerate-ai-infrastructure-buildout
12. James McKenna. "Europe's Industrial Leaders Reinvent Engineering and Manufacturing With NVIDIA." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/europe-industrial-leaders-engineering-manufacturing/
13. Kari Briski. "NVIDIA Open Models Advance Europe's Sovereign Intelligence Across Industries." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/open-models-europe-industry/
14. Erik Pounds. "NVIDIA Agent Toolkit Expands, Grounding Enterprise Agents in Company Data." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/enterprise-platforms-agent-toolkit/
15. John Fanelli. "NVIDIA Confidential Computing Expands, Enabling Frontier Model Builders to Securely Scale Across Global Infrastructure." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/confidential-computing-model-builders-infrastructure/
16. Timothy Costa. "NVIDIA cuDSS Now Generally Available to Accelerate Engineering and Scientific Simulations." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/cudss-generally-available-engineering-scientific-simulation/
17. Stuart Pitts. "NVIDIA Cloud Partners Scale Across Europe: Sovereign, Open and Built for Safe AI Agents." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/cloud-partners-europe-sovereign-safety-open-agents/
18. Uttara Kumar. "Google Cloud and NVIDIA Announce General Availability of NVIDIA Vera Rubin Infrastructure." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/google-cloud-vera-rubin-infrastructure/
19. Gerardo Delgado. "Bringing the Spark to Europe: Local Intelligence Powers Next Wave of Agents." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/local-ai-gtc-rtx-dgx-spark-station-sync/
20. Hannah Coutand. "European Leaders Turn to NVIDIA Vera CPUs to Build Local Infrastructure." NVIDIA Blog, October 20, 2026. https://blogs.nvidia.com/blog/european-leaders-vera-cpu-infrastructure/
21. Pranjali Joshi. "NVIDIA Cosmos Opens New Tools to Train, Evaluate and Deploy World Models." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/cosmos-open-train-evaluate-deploy-world-models/
22. Brian Sparks. "NVIDIA Unveils Minerva to Bring End-to-End Observability for AI Factories." NVIDIA Blog, October 22, 2026. https://blogs.nvidia.com/blog/minerva-platform-infrastructure/
23. Jason Renders. "Leading European Financial Services Organizations Embrace Open Models to Build Proprietary Intelligence." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/european-financial-services-organizations-open-models-proprietary-intelligence/
24. Mavi Nafria. "NVIDIA Transforms European Retail, From Digital Discovery to Physical Operations." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/european-retail-open-models-robotics/
25. Niall Robinson. "Pioneers Across Europe Deploy NVIDIA Earth-2 Open Models to Power Real-World Weather and Climate Prediction." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/europe-earth-2-open-models-weather-climate-prediction/
26. Eva-Maria Hempe. "NVIDIA Open Models Power a New Wave of European Public Sector Projects." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/european-public-sector-open-models/
27. Vladimir Troy. "How Companies Are Using NVIDIA DSX to Break the Power Bottleneck." NVIDIA Blog, October 21, 2026. https://blogs.nvidia.com/blog/dsx-power/
28. NVIDIA Writers. "Germany Wants to Lead the World in Industrial AI: How NVIDIA Will Help." NVIDIA Blog, October 19, 2026. https://blogs.nvidia.com/blog/germany-wants-to-lead-the-world-in-industrial-ai-how-nvidia-will-help/
29. NVIDIA Writers. "GTC Kicks Off With Workshops: Gear Up for the Keynote" (October 20, 2026), "Now Live: Pregame for NVIDIA CEO Jensen Huang's GTC Berlin Keynote" (October 21, 2026) and "At GTC Berlin, Attendees Swarm Demos and Displays" (October 22, 2026). Entries on the NVIDIA GTC Berlin live updates page. https://blogs.nvidia.com/gtc-berlin-2026-news/
30. Nat Ives. "NVIDIA Announces New Investments in France's Startups." NVIDIA Blog, October 12, 2026. https://blogs.nvidia.com/blog/france-ecosystem-agentic-computing/
31. Anthony Hills. "NVIDIA Announces New Investments in the UK Startup Ecosystem." NVIDIA Blog, October 13, 2026. https://blogs.nvidia.com/blog/uk-ecosystem-deployment/
32. NVIDIA Newsroom (@nvidianewsroom). Post on X announcing the OpenShell Reference Architecture, "Announced at GTC Berlin," October 21, 2026. https://x.com/nvidianewsroom/status/2112846545905230041
33. NVIDIA AI Infrastructure (@NVIDIAAIInfra). Post on X launching NVIDIA Minerva, October 22, 2026. https://x.com/NVIDIAAIInfra/status/2113313944966177176
34. NVIDIA AI (@NVIDIAAI). "Europe's industrial future is taking shape at GTC Berlin." Post on X, October 21, 2026. https://x.com/NVIDIAAI/status/2112859682415243776
35. Kif Leswing. "Nvidia announces plug and play Vera Rubin racks that can be shipped on a truck." CNBC, October 21, 2026. https://www.cnbc.com/2026/10/21/nvidia-announces-vera-rubin-racks-that-can-be-shipped-on-a-truck-.html
36. Joe Fay. "Huang Shrugs Off Protestors To Kick Off GTC Berlin With DSX Pods, Robots." The Next Platform, October 21, 2026. https://nextplatform.com/control/2026/10/21/huang-shrugs-off-protestors-to-kick-off-gtc-berlin-with-dsx-pods-robots/5304447
37. Daniel Howley. "Nvidia debuts platform for keeping AI agents from going rogue and humanoid robot design at GTC Berlin." Yahoo Finance, October 21, 2026. https://finance.yahoo.com/technology/article/nvidia-debuts-platform-for-keeping-ai-agents-from-going-rogue-and-humanoid-robot-design-at-gtc-berlin-113000342.html
38. Ana-Maria Stanciuc. "Protesters heckle Huang in Berlin over Nvidia chips in Russian drones." The Next Web, October 21, 2026. https://thenextweb.com/news/huang-gtc-berlin-keynote-protest-nvidia-chips-russian-drones
39. CoreWeave. "CoreWeave Launches the CoreWeave Zero Trust AI Framework, Keeping Customers in Control of Access to Their Models, Data and Runtime Environments." Press release, October 21, 2026. https://www.coreweave.com/news/coreweave-launches-the-coreweave-zero-trust-ai-framework-keeping-customers-in-control-of-access-to-their-models-data-and-runtime-environments
40. Dell Technologies. "Dell Technologies Advances Enterprise Agentic AI from the Deskside to Data Center with NVIDIA." Press release, October 21, 2026. https://www.dell.com/en-us/dt/corporate/newsroom/announcements/detailpage.press-releases~usa~2026~10~dell-technologies-and-nvidia-advance-enterprise-agentic-ai-from-the-deskside-to-data-center.htm