5 min learnNew DelhiUp to date: Mar 17, 2026 08:57 AM IST
Nvidia GTC, certainly one of Nvidia’s largest conferences of the yr, begins Monday, March 16, and runs until Friday. It’s being held in San Jose, California, with CEO Jensen Huang set to ship his highly-anticipated keynote speech onstage on the SAP Middle.
Huang will doubtless reveal new chips, merchandise, and partnerships geared towards retaining Nvidia within the lead amid a quickly increasing area of opponents. The four-day GTC, brief for GPU Expertise Convention, has advanced into a serious occasion for the chipmaker through the years, the place it showcases its newest advances in chips, knowledge facilities, its chip programming software program CUDA, digital assistants often known as AI brokers, and bodily AI similar to robots, and so forth.
This yr’s convention, nonetheless, is carefully watched by traders on the lookout for indicators that Nvidia’s technique of placing again its earnings into the AI ecosystem is definitely paying off.
The occasion additionally comes at a time when the general AI chip market is rising, with some analysts reportedly estimating that Nvidia’s slice may shrink because the AI business’s focus strikes past coaching AI fashions to inference, which may run on other forms of chips, such because the TPUs (Tensor Processing Models) developed by Google and Amazon’s Trainium chips.
Right here’s what we’d doubtless see at Nvidia GTC 2026 in San Jose.
New AI chip
In December final yr, Nvidia agreed to pay $20 billion to AI chip startup Groq to licence its AI inference {hardware}, whereas recruiting a number of of its staff, together with its founder. The non-exclusive licensing deal is reportedly Nvidia’s largest buy of any expertise to date.
Now, studies counsel that Nvidia will doubtless introduce a brand new chip that has been developed by harnessing Groq’s inference expertise. Inference is without doubt one of the steps concerned in creating giant language fashions (LLMs), the place the mannequin is made to generate outputs on beforehand unseen knowledge. It’s step by step changing into a bigger and extra aggressive a part of the AI computing course of, at the same time as AI adoption goes mainstream and clients search out cost-effective methods to satisfy the booming demand.
Story continues beneath this advert
Groq integrations
In an earnings name final month, Huang mentioned that Nvidia will showcase at this yr’s GTC, the way it can plug Groq’s ultra-fast inference expertise into the firm’s current CUDA platform.
Analysts additionally count on Nvidia to roll out a brand new line of servers that can mix Groq’s specialised chips with Nvidia’s networking applied sciences to create a speedy and cost-efficient product, as per a report by Reuters.
CPU-only servers
Whereas CPUs (central processor models) have taken a backseat to Nvidia’s GPUs in recent times, analysts mentioned that Nvidia is prone to showcase CPU-only servers at this yr’s GTC. Within the latest earnings name, Huang reportedly talked up CPUs, signalling that the chip could possibly be again in focus.
Subsequent-gen GPUs
Nvidia is predicted to offer an up to date roadmap of its next-gen GPUs, together with the most recent Vera Rubin household of chips which is because of ship within the second half of 2026. In final yr’s GTC keynote, Huang had additionally shared a number of particulars about Feynman GPUs, named after American theoretical physicist Richard Feynman, which it plans to deliver to market someday within the subsequent two years.
Story continues beneath this advert
Chip tools
Earlier this month, Nvidia introduced it’s investing $2 billion every in Lumentum and Coherent, each of which make lasers for sending info between chips within the type of beams of sunshine. Analysts count on Nvidia to showcase how using these lasers inside co-packaged optics might assist velocity up the connections amongst its chips housed inside large knowledge centres.
Nemo Claw
Latest studies counsel that Nvidia is planning to introduce an open-source platform for AI brokers known as ‘NemoClaw’ to rival the favored OpenClaw, which has come beneath scrutiny for safety threats. NemoClaw will reportedly be designed with enterprise-grade safety, privateness safety, and scalable activity automation. It can doubtless be deeply built-in with Nvidia’s NeMo framework, the Nemotron mannequin sequence, and Nvidia Inference Microservices. The platform can also be mentioned to be hardware-agnostic, that means it might run on any system and be powered by Nvidia, Intel, AMD, or different processors.
Bodily AI
Finally yr’s GTC, one of many key moments onstage was Huang’s interplay with a Star Wars-like robotic powered by Newton, a physics engine that may simulate robotic actions in real-world settings. Newton has been developed in collaboration with Disney and Google DeepMind. The corporate additionally unveiled an AI basis mannequin for humanoid robots known as Groot N1 at GTC 2025.
This yr, too, robotics is predicted to characteristic prominently in Huang’s keynote with analysts predicting that bodily AI could possibly be a multi-trillion-dollar business within the coming many years.