Posts in category: nVidia AI

CoreWeave today became one of the first cloud providers to bring NVIDIA GB200 NVL72 systems online for customers at scale, and AI frontier companies Cohere, IBM and Mistral AI are already using them to train and deploy next-generation AI models and applications.

CoreWeave, the first cloud provider to make NVIDIA Grace Blackwell generally available, has already shown incredible results in MLPerf benchmarks with NVIDIA GB200 NVL72 — a powerful rack-scale accelerated computing platform designed for reasoning and AI agents. Now, CoreWeave customers are gaining access to thousands of NVIDIA Blackwell GPUs.

“We work closely with NVIDIA to quickly deliver to customers the latest and most powerful solutions for training AI models and serving inference,” said Mike Intrator, CEO of CoreWeave. “With new Grace Blackwell rack-scale systems in hand, many of our customers will be the first to see the benefits and performance of AI innovators operating at scale.”

The ramp-up for customers of cloud providers like CoreWeave is underway. Systems built on NVIDIA Grace Blackwell are in full production, transforming cloud data centers into AI factories that manufacture intelligence at scale and convert raw data into real-time insights with speed, accuracy and efficiency.

Leading AI companies around the world are now putting GB200 NVL72’s capabilities to work for AI applications, agentic AI and cutting-edge model development.

Personalized AI Agents

Cohere is using its Grace Blackwell Superchips to help develop secure enterprise AI applications powered by leading-edge research and model development techniques. Its enterprise AI platform, North, enables teams to build personalized AI agents to securely automate enterprise workflows, surface real-time insights and more.

With NVIDIA GB200 NVL72 on CoreWeave, Cohere is already experiencing up to 3x more performance in training for 100 billion-parameter models compared with previous-generation NVIDIA Hopper GPUs — even without Blackwell-specific optimizations.

With further optimizations taking advantage of GB200 NVL72’s large unified memory, FP4 precision and a 72-GPU NVIDIA NVLink domain — where every GPU is connected to operate in concert — Cohere is getting dramatically higher throughput with shorter time to first and subsequent tokens for more performant, cost-effective inference.

“With access to some of the first NVIDIA GB200 NVL72 systems in the cloud, we are pleased with how easily our workloads port to the NVIDIA Grace Blackwell architecture,” said Autumn Moulder, vice president of engineering at Cohere. “This unlocks incredible performance efficiency across our stack — from our vertically integrated North application running on a single Blackwell GPU to scaling training jobs across thousands of them. We’re looking forward to achieving even greater performance with additional optimizations soon.”

AI Models for Enterprise 

IBM is using one of the first deployments of NVIDIA GB200 NVL72 systems, scaling to thousands of Blackwell GPUs on CoreWeave, to train its next-generation Granite models, a series of open-source, enterprise-ready AI models. Granite models deliver state-of-the-art performance while maximizing safety, speed and cost efficiency. The Granite model family is supported by a robust partner ecosystem that includes leading software companies embedding large language models into their technologies.

Granite models provide the foundation for solutions like IBM watsonx Orchestrate, which enables enterprises to build and deploy powerful AI agents that automate and accelerate workflows across the enterprise.

CoreWeave’s NVIDIA GB200 NVL72 deployment for IBM also harnesses the IBM Storage Scale System, which delivers exceptional high-performance storage for AI. CoreWeave customers can access the IBM Storage platform within CoreWeave’s dedicated environments and AI cloud platform.

“We are excited to see the acceleration that NVIDIA GB200 NVL72 can bring to training our Granite family of models,” said Sriram Raghavan, vice president of AI at IBM Research. “This collaboration with CoreWeave will augment IBM’s capabilities to help build advanced, high-performance and cost-efficient models for powering enterprise and agentic AI applications with IBM watsonx.”

Compute Resources at Scale

Mistral AI is now getting its first thousand Blackwell GPUs to build the next generation of open-source AI models.

Mistral AI, a Paris-based leader in open-source AI, is using CoreWeave’s infrastructure, now equipped with GB200 NVL72, to speed up the development of its language models. With models like Mistral Large delivering strong reasoning capabilities, Mistral needs fast computing resources at scale.

To train and deploy these models effectively, Mistral AI requires a cloud provider that offers large, high-performance GPU clusters with NVIDIA Quantum InfiniBand networking and reliable infrastructure management. CoreWeave’s experience standing up NVIDIA GPUs at scale with industry-leading reliability and resiliency through tools such as CoreWeave Mission Control met these requirements.

“Right out of the box and without any further optimizations, we saw a 2x improvement in performance for dense model training,” said Thimothee Lacroix, cofounder and chief technology officer at Mistral AI. “What’s exciting about NVIDIA GB200 NVL72 is the new possibilities it opens up for model development and inference.”

A Growing Number of Blackwell Instances

In addition to long-term customer solutions, CoreWeave offers instances with rack-scale NVIDIA NVLink across 72 NVIDIA Blackwell GPUs and 36 NVIDIA Grace CPUs, scaling to up to 110,000 GPUs with NVIDIA Quantum-2 InfiniBand networking.

These instances, accelerated by the NVIDIA GB200 NVL72 rack-scale accelerated computing platform, provide the scale and performance needed to build and deploy the next generation of AI reasoning models and agents.

Read More

Every company and country wants to grow and create economic opportunity — but they need virtually limitless intelligence to do so. Working with its ecosystem partners, NVIDIA this week is underscoring its work advancing reasoning, AI models and compute infrastructure to manufacture intelligence in AI factories — driving the next phase of growth in the U.S. and around the world.

Yesterday, NVIDIA announced it will manufacture AI supercomputers in the U.S. for the first time. Within the next four years, the company plans with its partners to produce up to half a trillion dollars of AI infrastructure in the U.S.

Building NVIDIA AI supercomputers in the U.S. for American AI factories is expected to create opportunities for hundreds of thousands of people and drive trillions of dollars in growth over the coming decades. Some of the NVIDIA Blackwell compute engines at the heart of those AI supercomputers are already being produced at TSMC fabs in Arizona.

NVIDIA announced today that NVIDIA Blackwell GB200 NVL72 rack-scale systems are now available from CoreWeave for customers to train next-generation AI models and run applications at scale. CoreWeave has thousands of NVIDIA Grace Blackwell processors available now to train and deploy the next wave of AI.

Beyond hardware innovation, NVIDIA also pioneers AI software to create more efficient and intelligent models.

Marking the latest in those advances, the NVIDIA Llama Nemotron Ultra model was recognized today by Artificial Analysis as the world’s most accurate open-source reasoning model for scientific and complex coding tasks. It’s also now ranked among the top reasoning models in the world.

NVIDIA’s engineering feats serve as the foundation of it all. A team of NVIDIA engineers won first place in the AI Mathematical Olympiad, competing against 2,200 teams to solve complex mathematical reasoning problems, which are key to advancing scientific discovery, disciplines and domains. The same post-training techniques and open datasets from NVIDIA’s winning effort in the math reasoning competition were applied in training the Llama Nemotron Ultra model.

The world’s need for intelligence is virtually limitless, and NVIDIA’s AI platform is helping meet that need — everywhere, all at once.

Read More

The final days of the AI Mathematical Olympiad’s latest competition were a transcontinental relay for team NVIDIA.

Every evening, two team members on opposite ends of the U.S. would submit an AI reasoning model to Kaggle — the online Olympics of data science and machine learning. They’d wait a tense five hours before learning how well the model tackled a sample set of 50 complex math problems.

After seeing the results, the U.S. team would pass the baton to teammates waking up in Armenia, Finland, Germany and Northern Ireland, who would spend their day testing, modifying and optimizing different model versions.

“Every night I’d be so disappointed in our score, but then I’d wake up and see the messages that came in overnight from teammates in Europe,” said Igor Gitman, senior applied scientist. “My hopes would go up and we’d try again.”

While the team was disheartened by their lack of improvement on the public dataset during the competition’s final days, the real test of an AI model is how well it can generalize to unseen data. That’s where their reasoning model leapt to the top of the leaderboard — correctly answering 34 out of 50 Olympiad questions within a five-hour time limit using a cluster of four NVIDIA L4 GPUs.

“We got the magic in the end,” said Northern Ireland-based team member Darragh Hanley, a Kaggle grandmaster and senior large language model (LLM) technologist.

Building a Winning Equation

The NVIDIA team competed under the name NemoSkills — a nod to their use of the NeMo-Skills collection of pipelines for accelerated LLM training, evaluation and inference. The seven members each contributed different areas of expertise, spanning LLM training, model distillation and inference optimization.

For the Kaggle challenge, over 2,200 participating teams submitted AI models tasked with solving 50 math questions — complex problems at the National Olympiad level, spanning algebra, geometry, combinatorics and number theory — within five hours.

The team’s winning model uses a combination of natural language reasoning and Python code execution.

To complete this inference challenge on the small cluster of NVIDIA L4 GPUs available via Kaggle, the NemoSkills team had to get creative.

Their winning model used Qwen2.5-14B-Base, a foundation model with chain-of-thought reasoning capabilities which the team fine-tuned on millions of synthetically generated solutions to math problems.

These synthetic solutions were primarily generated by two larger reasoning models — DeepSeek-R1 and QwQ-32B — and used to teach the team’s foundation model via a form of knowledge distillation. The end result was a smaller, faster, long-thinking model capable of tackling complex problems using a combination of natural language reasoning and Python code execution.

To further boost performance, the team’s solution reasons through multiple long-thinking responses in parallel before determining a final answer. To optimize this process and meet the competition’s time limit, the team also used an innovative early-stopping technique.

A reasoning model might, for example, be set to answer a math problem 12 different times before picking the most common response. Using the asynchronous processing capabilities of NeMo-Skills and NVIDIA TensorRT-LLM, the team was able to monitor and exit inference early if the model had already converged at the correct answer four or more times.

TensorRT-LLM also enabled the team to harness FP8 quantization, a compression method that resulted in a 1.5x speedup over using the more commonly used FP16 format. ReDrafter, a speculative decoding technique developed by Apple, was used for a further 1.8x speedup.

The final model performed even better on the competition’s unseen final dataset than it did on the public dataset — a sign that the team successfully built a generalizable model and avoided overfitting their LLM to the sample data.

“Even without the Kaggle competition, we’d still be working to improve AI reasoning models for math,” said Gitman. “But Kaggle gives us the opportunity to benchmark and discover how well our models generalize to a third-party dataset.”

Sharing the Wealth 

The team will soon release a technical report detailing the techniques used in their winning solution — and plans to share their dataset and a series of models on Hugging Face. The advancements and optimizations they made over the course of the competition have been integrated into NeMo-Skills pipelines available on GitHub.

Key data, technology, and insights from this pipeline were also used to train the just-released NVIDIA Llama Nemotron Ultra model.

“Throughout this collaboration, we used tools across the NVIDIA software stack,” said Christof Henkel, a member of the Kaggle Grandmasters of NVIDIA, known as KGMON. “By working closely with our LLM research and development teams, we’re able to take what we learn from the competition on a day-to-day basis and push those optimizations into NVIDIA’s open-source libraries.”

After the competition win, Henkel regained the title of Kaggle World Champion — ranking No. 1 among the platform’s over 23 million users. Another teammate, Finland-based Ivan Sorokin, earned the Kaggle Grandmaster title, held by just over 350 people around the world.

For their first-place win, the group also won a $262,144 prize that they’re directing to the NVIDIA Foundation to support charitable organizations.

Meet the full team — Igor Gitman, Darragh Hanley, Christof Henkel, Ivan Moshkov, Benedikt Schifferer, Ivan Sorokin and Shubham Toshniwal — in the video below:

Sample math questions in the featured visual above are from the 2025 American Invitational Mathematics Examination. Find the full set of questions and solutions on the Art of Problem Solving wiki. 

Read More

NVIDIA is working with its manufacturing partners to design and build factories that, for the first time, will produce NVIDIA AI supercomputers entirely in the U.S.

Together with leading manufacturing partners, the company has commissioned more than a million square feet of manufacturing space to build and test NVIDIA Blackwell chips in Arizona and AI supercomputers in Texas.

NVIDIA Blackwell chips have started production at TSMC’s chip plants in Phoenix, Arizona. NVIDIA is building supercomputer manufacturing plants in Texas, with Foxconn in Houston and with Wistron in Dallas. Mass production at both plants is expected to ramp up in the next 12-15 months.

The AI chip and supercomputer supply chain is complex and demands the most advanced manufacturing, packaging, assembly and test technologies. NVIDIA is partnering with Amkor and SPIL for packaging and testing operations in Arizona.

Within the next four years, NVIDIA plans to produce up to half a trillion dollars of AI infrastructure in the United States through partnerships with TSMC, Foxconn, Wistron, Amkor and SPIL. These world-leading companies are deepening their partnership with NVIDIA, growing their businesses while expanding their global footprint and hardening supply chain resilience.

NVIDIA AI supercomputers are the engines of a new type of data center created for the sole purpose of processing artificial intelligence — AI factories that are the infrastructure powering a new AI industry. Tens of “gigawatt AI factories” are expected to be built in the coming years. Manufacturing NVIDIA AI chips and supercomputers for American AI factories is expected to create hundreds of thousands of jobs and drive trillions of dollars in economic security over the coming decades.

“The engines of the world’s AI infrastructure are being built in the United States for the first time,” said Jensen Huang, founder and CEO of NVIDIA. “Adding American manufacturing helps us better meet the incredible and growing demand for AI chips and supercomputers, strengthens our supply chain and boosts our resiliency.”

The company will utilize its advanced AI, robotics and digital twin technologies to design and operate the facilities, including NVIDIA Omniverse to create digital twins of factories and NVIDIA Isaac GR00T to build robots to automate manufacturing.

Read More

As a teenager, Bradley Rothenberg was obsessed with CAD: computer-aided design software.

Before he turned 30, Rothenberg channeled that interest into building a startup, nTop, which today offers product developers — across vastly different industries — fast, highly iterative tools that help them model and create innovative, often deeply unorthodox designs.

One of Rothenberg’s key insights has been how closely iteration at scale and innovation correlate — especially in the design space.

He also realized that by creating engineering software for GPUs, rather than CPUs — which powered (and still power) virtually every CAD tool — nTop could tap into parallel processing algorithms and AI to offer designers fast, virtually unlimited iteration for any design project. The result: almost limitless opportunities for innovation.

Product designers of all stripes took note.

A decade after its founding, nTop — a member of the NVIDIA Inception program for cutting-edge startups — now employs more than 100 people, primarily in New York City, where it’s headquartered, as well as in Germany, France and the U.K. — with plans to grow another 10% by year’s end.

Its computation design tools autonomously iterate alongside designers, spitballing different virtual shapes and potential materials to arrive at products, or parts of a product, that are highly performant. It’s design trial and error at scale.

“As a designer, you frequently have all these competing goals and questions: If I make this change, will my design be too heavy? Will it be too thick?” Rothenberg said. “When making a change to the design, you want to see how that impacts performance, and nTop helps evaluate those performance changes in real time.”

U.K.-based supermarket chain Ocado, which builds and deploys autonomous robots, is one of nTop’s biggest customers.

Ocado differentiates itself from other large European grocery chains through its deep integration of autonomous robots and grocery picking. Its office-chair-sized robots speed around massive warehouses — approaching the size of eight American football fields — at around 20 mph, passing within a millimeter of one another as they pick and sort groceries in hive-like structures.

In early designs, Ocado’s robots often broke down or even caught fire. Their weight also meant Ocado had to build more robust — and more expensive — warehouses.

Using nTop’s software, Ocado’s robotics team quickly redesigned 16 critical parts in its robots, cutting the robot’s overall weight by two-thirds. Critically, the redesign took around a week. Earlier redesigns that didn’t use nTop’s tools took about four months.

“Ocado created a more robust version of its robot that was an order of magnitude cheaper and faster,” Rothenberg said. “Its designers went through these rapid design cycles where they could press a button and the entire robot’s structure would be redesigned overnight using nTop, prepping it for testing the next day.”

The Ocado use case is typical of how designers use nTop’s tools.

nTop software runs hundreds of simulations analyzing how different conditions might impact a design’s performance. Insights from those simulations are then fed back into the design algorithm, and the entire process restarts. Designers can easily tweak their designs based on the results, until the iterations land on an optimal result.

nTop has begun integrating AI models into its simulation workloads, along with an nTop customer’s bespoke design data into its iteration process. nTop uses the NVIDIA Modulus framework, NVIDIA Omniverse platform and NVIDIA CUDA-X libraries to train and infer its accelerated computing workloads and AI models.

“We have neural networks that can be trained on the geometry and physics of a company’s data,” Rothenberg said. “If a company has a specific way of engineering the structure of a car, it can construct that car in nTop, train up an AI in nTop and very quickly iterate through different versions of the car’s structure or any future car designs by accessing all the data the model is already trained on.”

nTop’s tools have wide applicability across industries.

A Formula 1 design team used nTop to virtually model countless versions of heat sinks before choosing an unorthodox but highly performant sink for its car.

Traditionally, heat sinks are made of small, uniform pieces of metal aligned side by side to maximize metal-air interaction and, therefore, heat exchange and cooling.

The engineers iterated with nTop on an undulating multilevel sink that maximized air-metal interaction even as it optimized aerodynamics, which is crucial for racing.

The new heat sink achieved 3x the surface area for heat transfer than earlier models, while cutting weight by 25%, delivering superior cooling performance and enhanced efficiency.

Going forward, nTop anticipates its implicit modeling tools will drive greater adoption from product designers who want to work with an iterative “partner” trained on their company’s proprietary data.

“We work with many different partners who develop designs, run a bunch of simulations using models and then optimize for the best results,” said Rothenberg. “The advances they’re making really speak for themselves.”

Learn more about nTop’s product design workflow and work with partners.

Read More

Get ready to explore the Deep South. South of Midnight, the action-adventure game from Compulsion Games, launches today on GeForce NOW. Following last week’s launch of Advanced Access in the cloud, GeForce NOW members can now pick up where they left off or dive into the adventure for the first time.

It leads six games available in the cloud this week.

Plus, catch the newest update from Epic Games’ Fortnite and miHoYo’s Honkai: Star Rail v3.2, just in time to celebrate the game’s second anniversary.

Unleash the Magic

South of Midnight, a captivating third-person action-adventure game, is set against the backdrop of a Gothic fantasy version of the American Deep South. Step into the shoes of Hazel, a young woman who embarks on a journey to find her missing mother and discovers her unique abilities as a Weaver — a magical individual who can mend broken bonds and spirits. Throughout her quest, Hazel encounters intriguing mythical creatures inspired by Southern folklore, explores surreal landscapes and unravels deep family secrets.

Blending mesmerizing visuals with innovative gameplay mechanics, South of Midnight combines traditional action-adventure elements with Souls-like combat and strategic spellcasting. Players use Hazel’s Weaver powers — such as Push, Pull and Weave — to tactically engage formidable enemies known as Haints, navigate challenging scenarios and solve intricate puzzles.

GeForce NOW members can jump straight into South of Midnight and be among the first to play by skipping the hassle of downloads or updates. Experience Hazel’s journey facing cryptid encounters with buttery-smooth frame rates and ultra-low latency — all without needing the latest hardware — using an Ultimate or Performance membership. Stream the magic, dodge the Haints and weave through this mythical adventure instantly in the cloud.

Everybody Jump

GeForce NOW members can jump right into games without waiting for patches or downloads, whether for a seasonal update or the latest in-game event. This week, Fortnite and Honkai: Star Rail fans can dive into plenty of new content from the cloud.

Global superstar and two-time GRAMMY award winner Sabrina Carpenter takes center stage in Fortnite Festival season eight with her eccentric brand of pop. Unlock her outfit, along with her “Juno” and “Nonsense” Jam Tracks, through the Music Pass or Shop. The Music Pass offers a mix of free and premium rewards for just 1,400 V-Bucks or through Fortnite Crew, with XP earned across all Fortnite modes.

The latest update in the cloud is Honkai: Star Rail Version 3.2, which celebrates the game’s second anniversary. There are two new five-star characters — Castorice and Anaxa  — and players can experience anniversary-themed events like “Star Rail WORLD” and a Light Cone giveaway. The update also introduces a revamped banner system for better character pull customization, as well as reruns of fan-favorite characters and challenging new events that bring fresh bosses and gameplay mechanics.

Time to Celebrate 

Commandos: Origins is launching in the cloud this week for members to stream. Head on a mission that’ll shape the fate of the entire world, and witness the beginning of a legendary elite World War II force. The long-awaited sequel to the Commandos series brings players back to the foundations of the real-time tactics genre and the days of Jack O’Hara, aka the Green Beret, and his five companions, completing missions no others would dare to accept.

From the icy plains of the Arctic and the vast deserts of Africa to the western coastlines of Europe and the Eastern front, lead commandos to success in high-risk missions, guiding them in their fight for a free world.

Look for the following games available to stream in the cloud this week:

• South of Midnight (New release on Steam and Xbox, available on PC Game Pass, April 8)
• Commandos: Origins (New release on Steam and Xbox, available on PC Game Pass, April 9)
• The Talos Principle: Reawakened (New release on Steam, April 10)
• Backrooms: Escape Together (Steam)
• Diablo III (Xbox, available on PC Game Pass)
• Sultan’s Game (Steam)

To stream supported Battle.net games from PC Game Pass on GeForce NOW, including this week’s addition of Diablo III, read the knowledge article.

What are you planning to play this weekend? Let us know on X or in the comments below.

What’s the latest you’ve stayed up while gaming?

— NVIDIA GeForce NOW (@NVIDIAGFN) April 9, 2025

Read More

As AI-powered tools continue to evolve, NVIDIA GeForce RTX 50 Series and NVIDIA RTX PRO GPUs, based on the NVIDIA Blackwell architecture, are driving faster, smarter creative workflows.

Blackmagic Design’s public beta release of DaVinci Resolve Studio 20 software puts new capabilities and a suite of RTX-accelerated AI-powered features into the hands of video editors, helping automate tedious tasks and accelerate creative workflows. Backed by the latest version of the NVIDIA TensorRT software development kit, these features run faster and more efficiently on RTX GPUs. Additional updates include support for 4:2:2 color formats and AV1 ultra-high-quality encoding.

In addition, the April NVIDIA Studio Driver, designed to optimize creative apps, will be available for download next week. For automatic Studio Driver notifications, as well as easy access to applications like NVIDIA Broadcast, download the NVIDIA App.

For artists and creators looking to experiment with generative AI workflows, Black Forest Labs’ FLUX.1-dev is now available as an NVIDIA NIM microservice — making it easy to deploy and scale across RTX-powered systems. FLUX.1-dev is a text-to-image generation model suite that enables users to generate high-quality, photorealistic visuals using text prompts and image inputs — optimized for peak performance on NVIDIA RTX and RTX PRO GPUs.

With support for FP4 compute, the FLUX.1-dev NIM microservice delivers up to 2x performance, over 4x faster generation on high-end GPUs and a 60% reduction in memory requirements compared with FP16 models. This means faster outputs, more creative iterations and broader accessibility across a wider range of RTX AI PCs and RTX PRO AI workstations.

AI-Assisted Video Editing

DaVinci Resolve Studio 20 beta delivers over 100 new professional-grade video features, including an array of AI tools and effects to speed up productivity. It also now integrates the latest version of NVIDIA TensorRT, optimizing AI performance for creative workloads.

GeForce RTX 50 Series and RTX PRO GPUs — powered by the NVIDIA Blackwell architecture and fifth-generation Tensor Cores — accelerate AI features that enhance creative workflows in video editing and post-production.

Take UltraNR Noise Reduction, an AI-driven noise reduction mode that intelligently targets and reduces digital noise in video footage to maintain image clarity while minimizing softening, especially in images with motion. UltraNR Noise Reduction runs up to 75% faster on the GeForce RTX 5090 GPU than the previous generation.

As part of the beta update, Magic Mask is an AI-powered feature that enables users to quickly and accurately select and track objects, people or features within a scene, simplifying the process of creating masks and effects. Magic Mask v2 adds a paint brush to further adjust masking selections for more accurate and faster workflows.

Among the latest RTX-accelerated AI features are AI IntelliScript, which auto-generates timelines based on original project scripts; AI Animated Subtitles, which syncs animated text to spoken words; and AI Multicam Smartswitch, which intelligently selects camera angles by detecting speakers.

Explore more AI effects in the beta update, which can be downloaded from Blackmagic Design support and is available at no charge for existing customers with DaVinci Resolve Studio 20.

Prompt, Set, Render With the FLUX.1-dev NIM Microservice

Image-generation models — a standout application of generative AI — can translate natural language into high-quality visuals across a wide range of styles. These models are transforming creative workflows, from storyboarding to concept art, by turning ideas into stunning imagery in just seconds.

The challenge is that these models are large and can be slow to run. Users often have to use optimized models or plug-ins to achieve faster inferencing — but getting all the bits and pieces needed to run the models at the best speed possible can be hard.

NVIDIA NIM microservices make it easier for creators and AI enthusiasts to get the full performance of their GPU in these models. NIM packages optimized generative AI models with everything needed to run them as fast as possible on RTX AI PCs and RTX PRO AI workstations.

One of the latest NIM microservices, Black Forest Labs’ FLUX.1-dev, demonstrates what’s possible with RTX-accelerated image generation. FLUX.1-dev includes a collection of models:

• FLUX.1-dev, which generates images from text prompts
• FLUX.1-Depth-dev, which adds depth map guidance for more structure and spatial control
• FLUX.1-Canny-dev, which uses canny edge detection to define shapes and composition more precisely

Optimized for NVIDIA hardware, the FLUX.1-dev NIM microservice delivers up to 2x performance with TensorRT and support for Blackwell FP4 and NVIDIA Ada FP8 precision — making it ideal for everything from concept art to pre-visualization for post-production.

Edit Like a Pro With 4:2:2

The 4:2:2 color format delivers huge quality benefits over creating in 4:2:0 and is highly sought after by professional video editors, presenting 2x the color information while only increasing raw file sizes by 30%.

With the additional color information, video editors gain improved color grading accuracy, increased flexibility during color correction and enhanced chroma keying while letting creators work with smaller files, maximizing efficiency and quality.

DaVinci Resolve Studio 20 beta has added support for hardware-accelerated 4:2:2 encoding and decoding on GeForce RTX 50 Series and RTX PRO Blackwell GPUs — and comes at a time where 4:2:2 is becoming more available in consumer cameras. Now, creators can shoot footage, import it quickly into their DaVinci Resolve projects and export their finished project in 4:2:2 color.

DaVinci Resolve 20 beta takes full advantage of the additional hardware decoders found in GeForce RTX 5080 and 5090 GPUs, as well as RTX PRO 6000, 5000, 4500 and 4000 Blackwell GPUs — now with support for 4:2:2. For example, with the RTX 5080 and 5090, creators can import 5x 8K30 or 20x 4K30 streams at once, or 9x 4K60 to do multi-camera editing and preview every angle without stutters. And with the RTX PRO 6000, this is boosted up to 10x 8K30 or 40x 4K30 streams.

When it’s time to export, video editors that use the GeForce RTX 50 Series ninth-generation NVIDIA video encoder can get a 5% improvement in video quality on HEVC and AV1 encoding (BD-BR), resulting in higher-quality exports at the same bitrates. Plus, a new Ultra High Quality (UHQ) mode available in the latest Blackwell encoder boosts quality by an additional 5%.

Finally, DaVinci Resolve Studio 20 beta adds support for three-way split-frame encoding — a technique where an input frame is divided into three parts, each processed by a different NVENC encoder. GeForce RTX 5090 Desktop and Laptop GPUs include three NVENC modules each, leading to significantly faster encoding speeds — more than 37% faster than the last generation.

Plug in to NVIDIA AI PC on Facebook, Instagram, TikTok and X — and stay informed by subscribing to the RTX AI PC newsletter.

Follow NVIDIA Workstation on LinkedIn and X. See notice regarding software product information.

Read More

NVIDIA this week recognized the contributions of partners in Europe, the Middle East and Africa at the annual EMEA Partner Day hosted by the NVIDIA Partner Network (NPN). 

Fourteen awards spanned eight categories. The recipients were honored for their outstanding efforts, dedication and innovative spirit in using NVIDIA technologies.

The 2025 NPN award categories and winners are:

Go-to-Market Excellence

Swisscom was honored with the Go-to-Market Excellence award for its Swiss AI Platform, which helps address the challenges of industries spanning financial services, government, high-performance computing and enterprise, as well as the growing need for sovereign AI. 

Swisscom’s successful marketing efforts included a digital “30 days of AI” campaign and a robust business-to-business sales campaign. Launched in just eight months, the platform onboarded several customers within two months of its launch, showcasing remarkable speed and execution. 

Industry Innovation

EDAG Group was recognized with the Industry Innovation award for its innovative platform that enables customers to create advanced industrial metaverse applications and digital twins. Key features of the platform, which is built with NVIDIA Omniverse, include data-driven decision-making, virtual training, quality assurance at the component level and AI-powered project management. These resulted in improved collaboration, increased efficiency and reduced costs for customers.

Consulting Partner of the Year

Deloitte was awarded the Consulting Partner of the Year award for creating customer success stories across EMEA. By combining its industry expertise and AI implementation capabilities with NVIDIA technologies, Deloitte has driven demand through digital marketing and events, engaging early adopters in pioneering use cases like sovereign AI infrastructure and long-term severe weather forecasting. 

Deloitte serves a diverse portfolio of clients — spanning finance, government, energy and retail — and has made substantial progress in training and certifying its team, building showcases that serve as crucial accelerators in client projects.

Software Distributor of the Year

TD SYNNEX was named the Software Distributor of the Year. Recognizing the critical role of software in NVIDIA’s full-stack technologies, TD SYNNEX developed the NATALA Tool, which offers resellers and customers a unique, personalized model to consume NVIDIA AI Enterprise software. This significantly contributed to the organization’s exceptional revenue performance in software.

Distributor of the Year

PNY received the Distributor of the Year award for achieving stellar revenue across EMEA in the past year, making significant contributions to the growth of NVIDIA’s partner ecosystem across the region.

Rising Star

Rising Star awards honor partners with the greatest revenue growth over the past year. The winners for each region are:

• Northern Europe: Computacenter UK
• Central Europe: Swisscom
• Southern Europe and the Middle East: 2CRSi

Star Performer

Star Performer awards recognize partners demonstrating excellence in sales across the entire NVIDIA portfolio. The winners for each region are:

• Northern Europe: Vesper Technologies
• Central Europe: Amber AI & Data Science Solutions
• Southern Europe and the Middle East: Solutions by STC

Star Performer, Software

The Star Performer, Software awards celebrate partners dedicated to selling NVIDIA’s full-stack technologies, with a focus on software. The winners for each region are:

• Northern Europe: WWT
• Central Europe: Amber AI & Data Science Solutions
• Southern Europe and the Middle East: Computacenter France​​

Learn how to join the NPN, or find a local NPN partner.

Read More

Necessity is the mother of invention. And sometimes, what a person really needs is hot chocolate served to them by a robot — one named after a pop star, ideally.

Angle Bush, founder and CEO of Black Women in Artificial Intelligence (BWIAI), began her AI journey in 2019 with the idea to build a robot named Usher that could bring her cocoa. As she scoured robotics tutorial videos for ways to bring her vision to life, Bush found herself captivated by something even bigger: artificial intelligence.

“As I’m doing this research, I’m finding more about artificial intelligence, and I’m hearing it’s the fourth industrial revolution,” she said.

But when Angle started attending AI events, a lack of diverse representation became glaringly obvious to her.

“I wasn’t quite seeing a full reflection of myself,” she said. “Surely you can’t have a revolution without Black women.”

From this realization, BWIAI was born.

Bush joined the NVIDIA AI Podcast to share more about the organization’s mission to reshape the AI community by educating, engaging, embracing and empowering Black women in the field.

Not five years after its founding, BWIAI brings together members from five continents and collaborates with key industry leaders and partners — serving as a supportive community and catalyst of change.

BWIAI and its partners offer hands-on learning experiences and online resources to its member community. They also launched a career assessment agent to help members explore how their interests align with emerging career paths in AI, as well as technologies and coursework for getting started.

“We have people in television, we have university professors, we have lawyers, we have doctors,” Bush said. “It runs the gamut because they are an example of what’s happening globally. Every industry is impacted by AI.”

Time Stamps

2:45 – Bush discusses BWIAI’s partnerships and initiatives, including its autonomous hair-braiding machine.

6:30 – The importance of educating, engaging, embracing and empowering Black women in AI.

10:40 – Behind BWIAI’s AI career assessment agent.

12:10 – Bush explains how removing barriers increases innovation.

You Might Also Like… 

NVIDIA’s Louis Stewart on How AI Is Shaping Workforce Development

Louis Stewart, head of strategic initiatives for NVIDIA’s global developer ecosystem, discusses why workforce development is crucial for maximizing AI benefits. He emphasizes the importance of AI education, inclusivity and public-private partnerships in preparing the global workforce for the future. Engaging with AI tools and understanding their impact on the workforce landscape is vital to ensuring these changes benefit everyone.

Tara Chklovksi, Anshita Saini on Technovation Pioneering AI Education for Innovation

Tara Chklovski, founder and CEO of Technovation, returns to discuss the importance of inclusive AI. With Anshita Saini, a Technovation alumna and OpenAI staff member, Chklovski explores how Technovation empowers girls through AI education and enhances real-world problem-solving skills. Saini shares her journey from creating an app that helped combat a vaping crisis at her high school to taking on her current role at OpenAI. She also introduces Wiser AI, an initiative she founded to support women and underrepresented voices in AI.

Imbue CEO Kanjun Qiu on Transforming AI Agents Into Personal Collaborators

Kanjun Qiu, CEO of Imbue, explores the emerging era where individuals can create and use their own AI agents. Drawing a parallel to the personal computer revolution of the late 1970s and ‘80s, Qiu discusses how modern AI systems are evolving to work collaboratively with users, enhancing their capabilities rather than just automating tasks.

Read More

NVIDIA is collaborating with Google Cloud to bring agentic AI to enterprises seeking to locally harness the Google Gemini family of AI models using the NVIDIA Blackwell HGX and DGX platforms and NVIDIA Confidential Computing for data safety.

With the NVIDIA Blackwell platform on Google Distributed Cloud, on-premises data centers can stay aligned with regulatory requirements and data sovereignty laws by locking down access to sensitive information, such as patient records, financial transactions and classified government information. NVIDIA Confidential Computing also secures sensitive code in the Gemini models from unauthorized access and data leaks.

“By bringing our Gemini models on premises with NVIDIA Blackwell’s breakthrough performance and confidential computing capabilities, we’re enabling enterprises to unlock the full potential of agentic AI,” said Sachin Gupta, vice president and general manager of infrastructure and solutions at Google Cloud. “This collaboration helps ensure customers can innovate securely without compromising on performance or operational ease.”

Confidential computing with NVIDIA Blackwell provides enterprises with the technical assurance that their user prompts to the Gemini models’ application programming interface — as well as the data they used for fine-tuning — remain secure and cannot be viewed or modified.

At the same time, model owners can protect against unauthorized access or tampering, providing dual-layer protection that enables enterprises to innovate with Gemini models while maintaining data privacy.

AI Agents Driving New Enterprise Applications

This new offering arrives as agentic AI is transforming enterprise technology, offering more advanced problem-solving capabilities.

Unlike AI models that perceive or generate based on learned knowledge, agentic AI systems can reason, adapt and make decisions in dynamic environments. For example, in enterprise IT support, while a knowledge-based AI model can retrieve and present troubleshooting guides, an agentic AI system can diagnose issues, execute fixes and escalate complex problems autonomously.

Similarly, in finance, a traditional AI model could flag potentially fraudulent transactions based on patterns, but an agentic AI system could go even further by investigating anomalies and taking proactive measures such as blocking transactions before they occur or adjusting fraud detection rules in real time.

The On-Premises Dilemma

While many can already use the models with multimodal reasoning — integrating text, images, code and other data types to solve complex problems and build cloud-based agentic AI applications — those with stringent security or data sovereignty requirements have yet been unable to do so.

With this announcement, Google Cloud will be one of the first cloud service providers to offer confidential computing capabilities to secure agentic AI workloads across every environment — whether cloud or hybrid.

Powered by the NVIDIA HGX B200 platform with Blackwell GPUs and NVIDIA Confidential Computing, this solution will enable customers to safeguard AI models and data. This lets users achieve breakthrough performance and energy efficiency without compromising data security or model integrity.

AI Observability and Security for Agentic AI

Scaling agentic AI in production requires robust observability and security to ensure reliable performance and compliance.

Google Cloud today announced a new GKE Inference Gateway built to optimize the deployment of AI inference workloads with advanced routing and scalability. Integrating with NVIDIA Triton Inference Server and NVIDIA NeMo Guardrails, it offers intelligent load balancing that improves performance and reduces serving costs while enabling centralized model security and governance.

Looking ahead, Google Cloud is working to enhance observability for agentic AI workloads by integrating NVIDIA Dynamo, an open-source library built to serve and scale reasoning AI models across AI factories.

At Google Cloud Next, attend NVIDIA’s special address, explore sessions, view demos and talk to NVIDIA experts.

Read More