NVIDIA’s $1T AI Forecast
At GTC Jensen Huang pegged AI‑driven orders at $1 trillion by 2027 and positioned NVIDIA as central to autonomous AI across businesses, space, and robotaxis — the keynote emphasized new software stacks to accelerate that push ( ). The message: NVIDIA sees hardware + software as the platform for the next wave of autonomous, agentic systems. (cnet.com)
NVIDIA said its Vera Rubin platform is already in volume production and described a full rack ecosystem that includes NVL72 GPU racks, BlueField‑4 storage racks and Spectrum‑6 Ethernet racks to scale agentic inference. (nvidianews.nvidia.com)) HPE and other infrastructure partners announced Rubin‑based systems and services for private‑cloud AI deployments at GTC, positioning OEMs to ship Rubin hardware to enterprise customers. (crn.com)) NVIDIA unveiled the Groq 3 LPU—built from technology acquired in a roughly $20 billion Groq asset deal—and said the Groq 3 wafers are being produced by Samsung Foundry on a 4nm node with shipments expected in the third quarter. (cnbc.com)) The company released Dynamo 1.0 as a production, open‑source “inference operating system” for AI factories and named initial adopters among cloud providers and production users, citing customers such as Perplexity, PayPal and Pinterest. (nvidianews.nvidia.com)) NVIDIA unveiled the Agent Toolkit and NemoClaw (which bundles the OpenShell runtime and Nemotron models) and listed enterprise partners including Adobe, SAP and Salesforce as early collaborators on agent deployments. (nvidianews.nvidia.com)) NVIDIA’s Nemotron‑3 Super model was published as an open‑weight model with roughly 120 billion parameters, a 1‑million‑token context window and a hybrid mixture‑of‑experts architecture tuned for multi‑agent, high‑throughput inference. (arxiv.org)) NVIDIA and Uber announced a robotaxi rollout plan that begins in Los Angeles and the San Francisco Bay Area in 2027 and scales to 28 cities globally by 2028 using NVIDIA DRIVE hardware and software. (money.usnews.com))
