Deloitte, Nvidia Target Industry with 'Physical AI'

This collaboration is an extension of a multi-year partnership; Deloitte was named Nvidia's Global Consulting Partner of the Year in 2021 for developing services around Nvidia's Omniverse, Metropolis, and DGX platforms. The new solutions focus on "Physical AI," enabling machines to perceive and interact with their environment, a market where 58% of enterprises already have some deployment. The offering is built on Nvidia Omniverse, a platform of libraries and microservices for creating industrial digital twins and robotics simulations. Deloitte will use Omniverse to build immersive simulations and leverage Nvidia's Isaac Sim for robotics development and Jetson Thor for synchronizing workloads between the edge and the cloud. To accelerate adoption, Deloitte is also launching a new Physical AI Center of Excellence in Shanghai. A core component is "agentic AI," which operates with a high degree of autonomy to manage complex tasks and achieve goals with minimal human input. In manufacturing, this means AI agents can independently analyze supply chains, predict machine failures, and re-route logistics in real-time to avoid disruptions, with some factories recording up to 25% lower energy costs. This move pits Nvidia's versatile GPU architecture against the rise of custom ASICs for AI workloads. While GPUs offer flexibility for a wide range of models, custom silicon can provide superior performance-per-watt and cost efficiency for specific, mature AI tasks. This trade-off is central to the "build vs. buy" decision for hyperscalers, who are increasingly designing their own chips like AWS's Trainium and Inferentia to optimize for cost and efficiency at scale. The Go-to-Market (GTM) strategy for such complex enterprise AI solutions is itself being transformed by AI. Modern GTM platforms are moving beyond single-point solutions to create unified intelligence engines that connect marketing, sales, and customer data. These systems use AI to identify high-intent accounts from behavioral data, automate personalized outreach, and provide real-time competitive intelligence. For enterprise ML teams, the total cost of ownership (TCO) for AI extends far beyond hardware. Inference costs can be 10 to 20 times higher than initial training costs, and expenses for data engineering, model maintenance, and specialized talent can constitute a majority of the budget. Subscription fees for AI tools often represent less than 40% of the actual total cost. The venture capital landscape reflects intense interest in the AI hardware space, with global funding for AI startups reaching $270.2 billion in 2025, accounting for over half of all VC investments. Notable recent funding rounds include AI chip startups like Ricursive raising $300M, Etched.ai securing $500M, and MatX raising $500M to develop custom hardware for large language models, signaling a well-capitalized challenge to incumbent chip designers.