What Wall Street’s looking for from Jensen Huang’s keynote GTC address

GTC, Nvidia’s big event this week, stands for GPU Technology Conference.

Yet, the media and Wall Street expect the main thrust of Jensen Huang’s keynote address will be how the chip designer is going beyond GPUs, the processor that enabled the AI boom.

Analysts have been hyping the event as a major potential catalyst for the chip designer. That’s despite recent high-profile affairs for the stock tending to see the shares do poorly.

“We expect a very bullish update around enterprise AI demand and expect Jensen to come out with a ‘no holds barred’ positive outlook on the AI industry,” wrote Wedbush Securities analyst Dan Ives. “This will be a much needed confidence boost for tech investors navigating a very tricky tape.”

Proving out the power of the technology, and the arms race to acquire powerful chips among hyperscalers, is no longer a top-of-mind consideration for investors.

This conference is likely to reflect the shift in sentiment: it’s no longer about what we can dream on, but the dollars and cents this technology can deliver. As such, efficiency and the breadth of Nvidia’s full-stack solutions will be in focus.

Groq

Late in 2025, Nvidia struck a “non-exclusive licensing agreement” with AI inference specialist Groq in an effective acquihire that saw some of its most important employees join the company.

The seeds of that relationship are expected to start to bloom today. Shortly after Nvidia’s Q4 earnings, The Wall Street Journal reported that Nvidia would be revealing “a new system for ‘inference’ computing” at this conference that incorporates a chip designed by Groq.

“We expect NVDA to unveil a next-gen inference rack with (SRAM-based) LPU chips inside at the upcoming 2026 GTC,” wrote Bank of America analyst Vivek Arya, suggesting that this could launch in 2027 or 2028.

Per Arya, Groq’s LPUs offer low latency, energy efficiency, and relative simplicity — and thanks to that, fewer supply chain or execution concerns.

“All of these lead to a chip that is ideal for single-stream or low-batch inference workloads,” he concluded.

During the quasi-veiled spat between OpenAI and Nvidia early this year, Reuters reported that the ChatGPT maker was “unsatisfied” with the inference performance of Nvidia’s AI chips.

CPUs

Nvidia’s recent “multi-year, multi-generational strategic partnership” with Meta includes an enhanced role for its CPUs in data center environments.

Dion Harris, Nvidia’s head of AI infrastructure, told CNBC that “CPUs are becoming the bottleneck in terms of growing out this AI and agentic workflow.”

The outlet reported that the chip designer is likely to unveil a “CPU-only rack” at the conference, with these chips aiming to deliver efficiency improvements for the entire data center package.

After Nvidia’s Q4 results, JPMorgan analyst Harlan Sur wrote:

“We suspect, though, that management has left a lot of ‘meat on the bone’ for GTC next month, including additional details on NVDA’s engagement with META on large-scale Grace/Vera CPU deployments (our view is that NVDA CPUs are being deployed alongside META MTIA ASIC XPUs), the incorporation of Groq’s low-latency inference architecture into upcoming platforms, and perhaps an updated quantification of backlog for the next several quarters.”

Optics

Nvidia recently invested $2 billion apiece in Coherent and Lumentum, two advanced optics companies.

These firms offer solutions that use light rather than electricity to move data around. As part of these partnerships, Nvidia is making purchase commitments for their offerings, so expect to hear more about the efficiency gains available from co-packaged optics and their integration with Nvidia’s product road map.

“The industry’s transition from copper to fiber, and from pluggable optics to Co-Packaged Optics, is addressing a genuine system-level bottleneck,” wrote Ben Bajarin, CEO and principal analyst at Creative Strategies. “The demand tailwind facing network infrastructure companies is real and now visible in public filings.”