GTC ’25: Impressions from “The Superbowl of AI”

By David Sapoznikow, VP at Material 

Vince Lombardi’s “Packers Sweep.” Joe Gibbs’ “Counter Trey.” Nick Sirianni’s “Brotherly Shove.” Every so often an NFL visionary comes up with an unstoppable play, deploys it with complete transparency and executes it flawlessly against hapless competitors. This brings us to GTC 2025, self-styled by NVIDIA CEO Jensen Huang as the “Superbowl of AI.” It marks a key milepost along the company’s epic journey. It also provides Jensen with a platform to share his vision for the future to eager audiences and beleaguered competitors, reinforcing NVIDIA’s palindromic mantra: “I AM AI.” 

Looking across Jensen’s recent keynotes, you can see an audacious vision steadily progressing toward business reality. His evolving narrative referenced 2023’s “ChatGPT moment” placing AI in the public consciousness, followed by 2024 as the year AI investments began delivering ROI and enabling the development of AI agents, all building up to this year. His 2025 keynote pushes the AI frontier out even further, incorporating game-changing, AI-powered agentics and tangible looks at Physical AI as the combination of the virtual (Omniverse) and physical (robotics).  He dares all comers to try and keep up.  

Jensen to Market: Full Speed Ahead 

Jensen used his 2025 keynote to focus attention on the health of the market and the sizeable opportunities emerging in the near future. Clearly taking aim at recent industry angst, he noted that demand for AI compute continues to rapidly increase regardless of cost concerns, energy requirements or competitive thunderbolts out of the PRC. Speaking directly to potential questions raised by DeepSeek (including whether the need for compute would now decrease), he demoed DeepSeek R1 for the audience to show how it enabled more accurate responses to queries than traditional LLMs while generating 20x more tokens in the process. More tokens require more power and, Jensen suggested, more throughput to handle those tokens for customers demanding speed. NVIDIA’S roadmap is designed to handle these increasing volumes, quickly. Just like that, Jensen reversed the logic of an argument against his company and then positioned NVIDIA as the only suitable supplier for the next few years. 

The keynote delivered a dense set of announcements seemingly intended to overwhelm skeptics with evidence of NVIDIA’s stable position atop the industry. He reassured partners that Blackwell GPUs remain in full production and high demand, expanded last year’s sneaky-good NIM Microservices with a Microsoft collaboration and debuted new CUDA libraries and an LLM model (Llama Nemotron), reinforcing the strength of the company’s software capabilities. To the ecosystem and competitors, he claimed NVIDIA’s position as a leader in inference (beyond their recognized leadership in training AI) while delivering a boundary-pushing product roadmap evolving Blackwell to Rubin (’26) to Feynman (’28) and introducing Dynamo as the “OS for the AI Factory.” For the enterprise, he outlined how NVIDIA’s Omniverse and Cosmo platforms are powering the potentially massive Robotics market – possibly the next $1T industry(!).  

Gaps and Opportunities 

As AI moves further from general capabilities to specific AI-assisted products, services and experiences, the rate of adoption will depend on the ability to articulate specific use cases and the corresponding value of AI. NVIDIA may be leading the ecosystem’s vanguard rapidly towards Physical AI, but most businesses and consumers are still early or pre-Generative AI. They struggle to understand how to use today’s tools, much less have clarity on how they might improve their businesses or lives with these emerging products and applications.  

Failing to effectively engage with customers may eventually carry a cost.  NVIDIA excels at producing emotionally-resonant customer stories that are simultaneously accessible to technologists and their business-oriented counterparts, but they are leaving important case studies involving end customers and Generative AI to others (notice only nine Generative AI examples on their portal). Enabling CSPs, like Google, who are today’s customers and tomorrow’s competitors, to define which audiences should be doing what and why, is risky if NVIDIA’s technological leadership shrinks and visionary status erodes. End user demand ultimately determines what sells, and companies that create without sufficient regard for customer preferences risk delivering glorified science projects. 

While Jensen presented NVIDIA as the dominant market leader, there were hints at potential sources of vulnerability and future challenges. Specifically, 

• No response to the potential competitive threat posed by major NVIDIA customers, such as the big hyperscalers, who are developing their own AI hardware.   

• No POV on the implications of recent advancements in quantum computing on the trajectory of AI and NVIDIA’s market position. 

• The absence of discussion on security was noticeable given AI’s profound implications, for good and bad, on the topic. 

• The TSMC logo was visible directly behind Jensen, providing a visual reminder of NVIDIA’s heavy supply chain dependency and the risks of geopolitical tensions. 

• The Dynamo announcement delivered an uncomfortable and public broadside at longtime partner VMWare, perhaps prompting others in the ecosystem to ponder potential competitive threats from NVIDIA. 

• The almost exclusive focus on Jensen coupled with the absence of other NVIDIA leadership raises questions about succession plans and bench depth. Jensen was the only human on stage for over two hours and even performed most voice overs. 

 Looking Ahead 

A moment during the keynote captured NVIDIA’s current state: in the middle of a demo, Jensen picked up two cables and found them unexpectedly tangled. He cooly untangled them live on stage with a couple of clever quips to the audience, never missing a beat. Just like its CEO, NVIDIA presents as a market leader on top of its game, supremely confident in its ability to retain that position regardless of challenges and surprises to come its way.     

The 2025 keynote reaffirmed NVIDIA as the perceived leader in the AI space, setting the course and pushing the industry forward. They continue to deliver remarkable growth while nimbly adjusting to emerging challenges in a phenomenally dynamic space. Their technology stack, driven by value, network effects and a loyal following render its continued market leadership all but assured for the foreseeable future. NVIDIA has no clear rival – today – who can match its vision, technology or ecosystem lock-in.  

But there’s an old military saying that “the enemy has a vote.” Given how rapidly AI technology is evolving, NVIDIA’s seemingly irresistible force may yet encounter a competitive immovable object sooner than expected. The Packers Sweep met its match with the Dallas “Flex” defense. Washington’s Counter Trey was overwhelmed by Chicago’s “46.” The Brotherly Shove may disappear courtesy of the NFL’s rules committee. That day may come for NVIDIA, but it is not yet time to bet against Jensen and company.