Nvidia's GTC: Agentic AI Bet
Nvidia used GTC to argue the next AI bottleneck is the surrounding stack—packaging, interconnects, software, and supply chains—not just GPU FLOPS, and it forecasted a $1 trillion market for agentic AI by 2027. The company’s networking arm has quietly become a multibillion-dollar business, and regulators are flagging that hyperscale AI power demand creates real grid‑stability risks that engineers will have to design around. (digitimes.com) (geneonline.com) (techcrunch.com) (eenews.net)
NVIDIA described the Vera Rubin platform as entering full production with seven new chips and rack-scale systems including NVL72 GPU racks, Vera CPU racks, Groq 3 LPX inference racks, BlueField‑4 STX storage racks and Spectrum‑6 SPX Ethernet racks. (nvidianews.nvidia.com) NVIDIA’s NemoClaw stack was announced to install Nemotron models and the new OpenShell runtime in a single command, enabling always‑on agent deployments from DGX and DGX Spark systems down to RTX PCs and workstations. (nvidianews.nvidia.com) NemoClaw’s security layer uses an OpenShell runtime that supports granular, YAML‑based permission policies for locally deployable agents, a design NVIDIA and coverage outlets flagged as aimed at enterprise safety and privacy. (zdnet.com) TechCrunch reported NVIDIA’s networking segment produced roughly $11 billion in a single quarter, a scale the outlet compared to Cisco’s networking business and which analysts called evidence the unit has become a multibillion‑dollar core business. (techcrunch.com) NVIDIA’s Groq deal last December involved roughly $20 billion for Groq’s assets, and GTC 2026 introduced the Groq 3 LPU and liquid‑cooled LPX racks integrated into the Vera Rubin lineup to boost low‑latency inference throughput. (cnbc.com) OEM and systems partners moved quickly: HPE unveiled Vera Rubin‑based systems at GTC and vendors including ASUS, Dell, GIGABYTE, MSI, Supermicro and HP were named as channels for new DGX Spark and DGX Station offerings. (crn.com) The North American regulator NERC has labeled emerging large AI data‑center loads a “high likelihood, high impact” reliability risk, publishing Large Loads guidance and white papers in 2025 and warning that rapid load swings—citing prior events such as a 1.5‑GW simultaneous disconnection in Virginia in 2024—could trigger cascading outages. (nerc.com) GTC 2026’s scale—reported attendance of more than 30,000 from over 190 countries—was used onstage to frame NVIDIA’s move from chips to a “full‑stack” AI infrastructure provider, positioning hardware, networking, software and partner systems as components of a single industrial buildout. (storagereview.com)
