Nvidia’s marquee GTC keynote is almost here, and CEO Jensen Huang is set to use the stage to map out the next phase of the AI computing boom. Expect a flurry of hardware, software, and ecosystem news aimed squarely at scaling generative AI from research labs to everyday products—and at keeping Nvidia firmly in the driver’s seat.
How to Watch the Nvidia GTC Keynote Live Online
The keynote will stream free on Nvidia’s GTC event site and on the company’s official video channels, with a full replay typically posted shortly after the broadcast. If you’re attending in person, the main address is hosted at the SAP Center, and overflow viewing is usually available in nearby conference theaters. Pro tip: Nvidia often enables high-bitrate streams and live captions; choose the highest quality your connection supports for demo-heavy segments.
Registration for GTC sessions is free for virtual attendees and unlocks additional technical talks, hands-on labs, and developer Q&As that go deeper into what Huang announces. If you can’t tune in live, Nvidia generally publishes session recordings and slide decks in the event library for on-demand viewing.
Hardware Announcements to Watch at Nvidia GTC
Industry chatter points to a new Nvidia part targeting AI inference—the phase where models generate outputs for users. Faster, cheaper inference is the last major bottleneck for scaling generative AI. Analyses from a16z and SemiAnalysis estimate inference often accounts for 70–90% of AI compute spend once systems are deployed, so any uplift in performance-per-watt or performance-per-dollar will be consequential for cloud providers and enterprises alike.
Look for updates on the Blackwell platform’s rollout and networking stack, including NVLink, NVSwitch, InfiniBand, and Ethernet via Spectrum-X—areas where system-level gains can rival chip-level improvements. Expect emphasis on liquid cooling, rack density, and energy efficiency as hyperscalers chase lower total cost of ownership. Omdia and New Street Research have pegged Nvidia’s share of AI training accelerators near 80% in recent years; the keynote will likely outline how the company plans to defend that lead while pushing deeper into inference, where Google’s TPU, Amazon’s Inferentia, AMD’s Instinct accelerators, and Meta’s MTIA are intensifying competition.
Keep an ear out for roadmap breadcrumbs. Nvidia has previewed successors to current architectures on prior roadmaps, and Huang often teases memory configurations, interconnect upgrades, and software hooks that hint at what’s next without naming every SKU outright.
Software and AI Agents in Focus at Nvidia GTC
On the software side, Nvidia is expected to double down on its end-to-end stack: CUDA for acceleration, TensorRT and Triton for inference, and the growing suite of NIM microservices for deploying models behind APIs. Wired has reported that Nvidia is preparing an open-source platform for enterprise AI agents, referred to as NemoClaw, designed to orchestrate multi-step workflows and connect models to tools and data. If announced, it would squarely target the emerging market for agentic systems and mirror efforts from players like OpenAI and Anthropic while keeping workloads anchored to Nvidia’s runtime and hardware.
Watch for MLPerf references and real-world benchmarks—Huang frequently cites standardized scores and customer case studies to validate speedups. Expect demos that pair large language models with retrieval, vision, and speech, alongside guardrails, audit trails, and policy controls aimed at regulated industries.
Partnerships and Industry Demos to Expect at GTC
GTC is as much about ecosystem momentum as it is about chips. Anticipate updates across cloud partners—AWS, Microsoft Azure, and Google Cloud—as well as telecom operators building AI-ready networks. In automotive, Nvidia’s Drive platform has ongoing programs with brands like Mercedes-Benz and BYD; fresh milestones or software releases often land here. For robotics, Isaac-powered stacks typically feature factory automation from integrators such as Foxconn and a wave of AMR demonstrations.
Healthcare remains a showcase vertical. Look for BioNeMo updates, synthetic data generation, and imaging workflows with major providers and pharma companies; prior GTCs have highlighted collaborations with enterprises like Amgen and Recursion to accelerate discovery pipelines. In the public sector, expect mentions of “sovereign AI” initiatives as governments and national labs invest in domestic compute capacity.
Why This GTC Keynote Matters for the AI Landscape
The balance of power in AI is shifting from training headline-grabbing models to serving billions of daily inferences efficiently. If Nvidia delivers meaningful gains in inference throughput and latency—plus tighter integration across networking, compilers, and deployment tooling—it strengthens a moat built on hardware, software, and developer ecosystem lock-in.
At the same time, rivals are closing gaps. Google’s TPU generations have posted strong price-performance for specific workloads, AMD has notched wins with its Instinct line, and hyperscalers are rolling out custom silicon to rein in costs. The stakes are clear: better inference economics will dictate which platforms power the next wave of AI features in search, productivity apps, robotics, and vehicles.
Bottom line: Tune in on Nvidia’s official channels for the keynote, then dive into the developer sessions for the fine print. The biggest reveals often hide in the details—software flags, networking topologies, and reference designs that ultimately decide whether AI gets faster, cheaper, and greener at scale.