NVIDIA GTC 2025 Preamble: Not Written By AI

Here we go: The NVIDIA GTC 2025 Preamble. With t-minus 24 hours until GTC kicks off...a few musings.

There is a ton of speculation on this week’s conference and with the recent market volatility (mostly unrelated to tech), the AI trade has come under fire. But make no mistake—the AI Trade is still red hot, and I expect that we will see and hear further confirmation this week when NVIDIA CEO Jensen Huang takes the stage at what is today the world’s most important AI conference.

So, what is going to be in focus? Here are some of my thoughts ahead of the event. I’ll do my best to steer from the obvious.

The Big Announcements—

Datacenter will continue to be the epicenter of the AI demand narrative. How much Capex will be spent and how fast can hyperscalers continue to cycle from generation to generation of NVIDIA platforms.

We will see Blackwell Ultra deliver a mid-cycle update to provide needed additional memory.

But it will be the Rubin announcements that will get the biggest headlines. The annual cadence is the driver of the companies DC business and with more than $300 billion in projected CAPEX in 2025 coming from just the largest companies, it will be seeing the next generation that will encourage investors and customers that the NVIDIA investment path remains the right one for future-proofing data centers and portfolios.

We can also expect an explosion of interest in co-packaged optics moving the optical engine closer to the switch and increasing the efficiency of GPUs. Powerful.

Also, the enterprise has remained underwhelming in adoption and deployment of AI. Lots of tinkering and PoC, but we have even NVIDIA get closer to the likes of HPE, Dell, and more recently Cisco to improve its positioning for more enterprise workloads and diversification beyond the big hyperscalers. An important area to keep an eye on.

It is Bigger than Datacenter Though—

At CES Jensen discussed the “Chat GPT Moment” for robotics. The opportunities for AI beyond the DC, powered by the real world will be exponential even compared to the DC. It is a cloud to edge and back stack that will enable this, but it is critical for NVIDIA to land that it is more than just a DC GPU company returning its focus on Automotive and the future of physical robots and humanoids. Part of why the stock has been rangebound has been a lack of confidence that there is a catalyst beyond DC and the DC CAPEX risk has become high. Continuing to convince the market in these emerging fields + a drill down on Quantum, which feels appropriate after the yo-yo of Quantum since Jensen’s CES comments.

The CPU+GPU+QPU era is nearing.

We will also hear more on graphics (gaming) and client innovation — but I put that lower on the list.

The Subtle Undertones to Listen For—

Thematically, it is all about convincing the world that NVIDIA’s datacenter lead is palpable and that its innovation engine goes far and wide beyond just the data center.

Recent comments about XPUs and custom chips from NVIDIA have indicated strong confidence in its market position. Our channel checks do indicate though a growing interest, demand, and expectation of more performant and capable XPUs that will offer significantly better economics. I’ve written extensively about this, and I do believe these will be more pointed toward the largest hyperscalers looking to improve AI cost efficiencies and improve margins and decrease dependencies. But this is a notable consideration.

I think it is important here to also note, we continue to expect XPU growth to meaningfully ramp in 2026, but we don’t feel it is all at the detriment of GPUs and NVIDIA. We see the $500+ Billion TAM expansion from growing utilization of AI to expand the market opportunity leaving strong forward revenue and EPS growth for NVIDIA while also seeing the likes of Broadcom, Arm, and others in the custom chip space expanding. And there you have it. Huge week ahead.

Strap in everyone. It is GTC time.