NVIDIA Will Corner the Industrial Software Market With New GPU Architecture and API Offerings

Upon release later this year, Blackwell GPUs will be rolled out in data centers hosted by cloud providers such as AWS, Google, IBM, and Microsoft and will be offered on the respective cloud platforms. The B100 and B200 GPUs will be available for cloud operations and will complement the current H100 and H200. Innovation comes from the combination of two B200 GPUs to create the GB200 Grace Blackwell Superchip which will see a performance increase of 30x for inference workloads while also consuming 25x less energy. The level of improvement in performance and efficiency with the GB200 will allow users to run demanding AI workloads such as Mixture-of-Experts which exceed 1 trillion parameters. At maximum usage, the new superchip will be able to run LLMs up to 10 trillion parameters, making GB200 the fastest and most efficient way to operate the largest and most comprehensive Generative AI models.

Available first on Microsoft Azure, followed by AWS, GCP, and IBM, the 5 new upcoming APIs will allow further portability between 3D applications and the OpenUSD platform Omniverse. With enhanced data extraction and integration to Omniverse, users will be able to build photorealistic digital twins with accompanying data inputs from complex datasets exported from an initial modeling software such as 3DEXCITE, and portions of the Siemens Xcelerator platform.  

With the upcoming release of the new Omniverse Cloud APIs, industrial software providers have been showcasing new and improved methods to deploy LLM, digital twins, and photorealism. This is just the start as further case studies, beta releases, and test runs will be implemented over the next 6 months as the APIs are further explored and the connection between 3D modeling applications and NVIDIAs OpenUSD platform Omniverse is refined. Additionally, when the new GPU200 is released at the end of the year, all functionality utilizing cloud and Generative AI will be vastly increased. The speed and efficiency in deploying and training AI models will exceed the current Hopper generation GPUs, however due to supply limitations and the current compute power of the Hopper GPUs, the GB200 will be reserved for high end clients that require leading edge training. Below are the most promising impacts from the upcoming use cases of Omniverse Cloud APIs from GTC:

Ansys- Ansys will adopt Omniverse APIs to aid in data interoperability, with RTX enabled simulation and enhanced cloud rendering. Ansys will first deploy these connections to its AVxcelerate software, used to simulate environments for autonomous vehicles through radar, lidar, and optical sensors.

Dassault Systèmes- Dassault will connect Omniverse APIs to its 3DXCITE software along with the overarching 3DEXPERIENCE platform to effectively assign data to digital twins and simulation using Omniverse cloud capabilities.

Hexagon- Hexagon will connect Omniverse APIs to its reality capture and digital reality platforms. This connection will come through the OpenUSD framework and will improve latency times on hyper-realistic renderings of 3D objects and the simulation required to produce an effective digital twin.

Siemens- Teamcenter X will connect to Omniverse APIs allowing for RTX rendering and design data integration with NVIDIA Generative AI. This will allow Siemens to design digital twins of large manufacturing projects with enhanced collaboration, AI insights, and top-notch photorealism.

NVIDIA went all in to showcase the breath to which new APIs are allowing stronger ties between design data and virtual 3D representation in Omniverse along with the potential power of the upcoming GB200 to further enable LLMs, digital twin creation, and robotics simulation. The tentative shortlist for industrial software vendors planning on utilizing this technology is Ansys, Cadence, Dassault Systèmes, Hexagon, Microsoft, Rockwell Automation, Siemens, and Trimble. This list will only grow as competitors realize the danger of not partnering with NVIDIA for Omniverse.

Demand for functional and aesthetically pleasing digital assets are increasing, with the only viable engine capable of running such operations being Omniverse. The closest competition to Omniverse currently comes from Unity and Unreal engines which are more prevalent in the online gaming scene and cannot compete with Omniverse on the hardware or software side of simulation, rendering, and AI. Furthermore, with NVIDIAs market position of focusing on the intersection of AI, robotics, and digital twins within the Omniverse platform, manufacturers and manufacturing software suppliers have a dedicated partner that has a tailored solution fit for industry use. While only a handful of industrial software providers have partnered with NVIDIA and fewer with demonstrated use cases, the partnerships will undoubtable grow. NVIDIA has a proven winning strategy with Omniverse by virtue of RTX GPUs for complex computations which outperform competitors, making it the logical partner to accelerate the development of photorealistic and immersive experiences. With this in mind, it is imperative that partnerships between NVIDIA and the remaining industrial software providers are set in stone before the initial launch of the new cloud APIs and GB200. With Generative AI and digital twins being the talk of the town in the manufacturing world, industrial software providers need to capitalize on this hype by partnering with NVIDIA and providing potential use cases for the connection of Omniverse to in-house software. If industrial software providers choose not to partner with NVIDIA the ramifications will be subpar renders of high-quality 3D simulations, lack of capability to deploy high end large parameter Generative AI models, and future use of GB200 for compute power. Choosing an alternative provider to NVIDIA runs the unnecessary risk of significant user loss in CAD, PLM, MES, and simulation software as large manufacturers desire the best-in-class features Omniverse Cloud brings to the table.