The setup for GTC next week (Keynote: Monday, March 16) is that the near-term bar is already high because of the company’s guidance for the April quarter, provided on February 24th. For the April quarter, the Street was looking for roughly 64% growth, and the company guided to something closer to 79% at the high end, a number they will likely exceed. That raised guidance was met by a collective shoulder shrug, with shares of NVDA trading down 9% over the following two days, compared to the Nasdaq being down 2%.

My sense is that Jensen will emphasize the concept that the utility of AI reached an inflection point starting back in November. It is a theme he mentioned on the last earnings call and a good place to start GTC, because the recent improvement in utility is remarkable and likely having a significant impact on Nvidia’s demand outlook.

As for the July quarter, the Street is expecting a step up to 84% y/y growth, given it is the first quarter where the China H20 comp headwind is removed. Any read on July will also likely be positive, yet largely ignored by investors because the focal point has shifted to CY27.

The reason investors are ignoring favorable near-term updates is a belief that growth rates will be halved in each of the next two years. For CY26, the Street is looking for 70% growth, declining to 30% in CY27 and 16% in CY28. The central challenge for NVDA shares is a concern that growth next year will step down meaningfully; the Street is looking for 31% growth in CY27, compared to the 68% growth expected in CY26.

The GTC Keynote (typically held twice a year) gives us a slightly different read on the business compared to the quarterly earnings. It’s Jensen giving a 90-minute masterclass on how to think about the layering of AI infrastructure demand curves, along with the state of utility regarding the underlying models.

At GTC in DC at the end of October, the headline message was: “Physical AI is small today and will be meaningful in the future.”

Next week, I expect Jensen to continue to highlight physical AI as a major driver and expand the conversation to include comments on how Rubin changes inference economics, and whether networking is becoming a bigger part of the growth story.

The question is: Will his layered AI growth outlook persuade investors to get back on board?

Similar to the last earnings call, I expect Jensen to emphasize that Rubin improves the economics of inference for customers; therefore, demand for Rubin will outpace the 30% expectations for CY27.

Specifically, I’ll be listening for language that outlines Rubin’s impact on the cost per token, throughput, and performance per watt. Those are the details that should lead investors to gain confidence that Nvidia’s position is defensible against AMD and custom silicon, and have the potential to impact how investors think about CY27.

In CY26, Networking is expected to be 16% of total revenue compared to 14% in CY25. In CY27, the segment is expected to dip back to 14% of revenue. This translates to reported growth in CY25 of 141%, CY26 of 80%, and 23% growth in CY27. In other words, investors are expecting segment growth to hit a wall in CY27.

My sense is that performance increasingly depends on how well the full system handles data movement and power, which lays the groundwork for the segment to exceed Street expectations and grow at more than 23% next year.

Back at GTC in DC, Jensen’s keynote kicked off with a video titled “Physical AI.” Looking at the presentation through the eyes of an investor, the company appeared to be addressing the central question from Wall Street: How can the Nvidia growth party continue beyond 2026?

Physical AI systems are different from the chatbots and agentic AI that have been the focus over the past three years and the catalyst for massive investment from hyperscalers, which I believe accounts for about 60% of revenue. Physical AI can understand, reason, and interact with the real world. That means these models will excel at computer vision and applied physics, skills chatbots do not prioritize.

Some of the biggest use cases for this AI include a range of “light” physical AI on your phone or glasses processing the world around us, as well as autonomous driving systems, specialized robots, and humanoid robots.

I believe Physical AI is going to be a meaningful growth driver for Nvidia starting in CY28. Today, only about 1% of Nvidia’s sales are used to deliver physical AI, which are reported as the Automotive and Robotics segment. I believe by CY30, the Automotive and Robotics segment will grow into a $50B+ annual business, up from about $3B a year today. That means by CY30, physical AI could account for 8% of a $600B total revenue pie, up from our estimate of $350B in CY26 (where the Street sits at $320B).

Disclaimer