AI Portfolio

3 Can't Miss Announcements from GTC: NVIDIA's (NVDA) Biggest Event of the Year

Eric Bleeker

Published: March 19, 2025 11:21 am

NVIDIA (Nasdaq: NVDA) is hosting its biggest event of the year: GTC. Yesterday NVIDIA CEO Jensen Huang gave a keynote which not only was the company’s opportunity to push back on concerns weighing down its stock, but also to announce future product updates. 

The keynote was long, but we watched it and distilled Huang’s comments down to three key points that will shape the direction of NVIDIA’s stock in the years ahead. Will the company’s share price head to $80 per share or $200 per share next? It depends on how well NVIDIA executes in the three key areas we’ve identified below. 

Trending Video

▶️

1. NVIDIA Wants to Change the AI Story 

As we noted at the top of this article, GTC is an event that was created to share recent innovations from NVIDIA and build support for its ecosystem of chips and software. Yet, while there are dozens of panels, workshops, training labs, and sessions at GTC; the highlight of the show is Huang’s keynote which is targeted at Wall Street. 

Huang wasted no time directly addressing ‘bear’ arguments that have been pointed at NVIDIA in recent months. After the company’s shares sold off more than 15% on January 27th following DeepSeek’s R1 announcement, NVIDIA’s CEO directly addressed the panic at GTC. He said:

“This last year, this is where almost the entire world got it wrong. The computation requirement, the scaling law of AI, is more reslient and in fact, hyper-accelerated. The amount of computation we need at this point as a result of agentic AI, as a result of reasoning, is easily 100 times more than we thought we needed at this time last year.” 

Another ‘bear case’ that has targeted NVIDIA is that the company is overly reliant on just a few customers. To be sure, NVIDIA’s customer concentration is immense. 

A small group of four companies – Amazon, Microsoft, Google, Meta – are expected to spend more than $320 billion on capital expenditures this year. The bulk of that money is being spent on building AI data centers. 

While the estimates for how much these companies are planning to spend this year have continued growing – in November, the estimate for 2025 spend was closer to $280 billion – Wall Street has shifted from enthusiasm about spending gains to worry that these companies have reached their spending limit. 

So, how does NVIDIA get Wall Street to shift from a ‘glass half empty’ viewpoint to one that’s a glass half full? 

The biggest driver for NVIDIA’s stock would be a change of story. The more NVIDIA can demonstrate there are coming demand channels the market isn’t accounting for today, the more upside its stock possesses. 

Jensen Huang began making this pitch in earnest. First, the company provided more details on why they believe spending on AI data centers can hit $1 trillion by 2028 (a significant jump from today’s levels). The crux of this argument is the quote we featured above: that compute intensity for AI models has increased 100-fold from expectations last year. 

In addition, NVIDIA laid out a future where robotics could become the next multi-trillion industry. There were some tangible business announcements from the company – NVIDIA announced a partnership with GM to power their self-driving technology – as well as technology meant to ‘seed’ this market. 

For example, NVIDIA announced GROOT N1, a foundation model for building humanoid robots. 

NVIDIA’s overall pitch was best summarized by the below chart which details AI use cases for every industry:

• Robots, AV
• Robotic Manufacturing 
• Edge, 6G 
• GPU Clouds 
• Cloud Service Providers 
• Enterprise IT 

Right now, the bulk of NVIDIA’s revenue is from the Cloud Service Providers and GPU Clouds (companies like CoreWeave). 

Areas like Enterprise IT, Edge, Roboticic Manufacturing, and Robots are all ‘greenfield’ opportunities. The challenge is that NVIDIA is simply so large that new opportunities need to be massive to move the stock. For example, few investors realize that NVIDIA’s automotive unit is growing at a red-hot 103% rate (higher than their data center growth). 

The ‘challenge’ is that revenues for that unit are at a $2 billion run rate. Even if that unit 5X’s in the next two years, it’s still just a few percent of NVIDIA’s total revenue. 

Similarly, robotics feels like it will be massive and the next major trend in technology. NVIDIA is setting itself up very well in this market, but absent a path to robotics producing tens of billions annually, it will have little impact on the company’s stock. 

Many of these new opportunities – Edge, Robotics, Robotic Manufacturing – remain something that could be interesting in the next 3-5 years.

Yet, how NVIDIA performs in the next year will be more dependent on whether Wall Street buys that AI data center spending can hit $1 trillion by 2028 and how well NVIDIA is fending off key rivals like Broadcom (Nasdaq: AVGO). 

2. NVIDIA’s Roadmap Remains Extremely Aggressive 

Another important aspect of Huang’s keynote is putting more details behind the company’s roadmap. NVIDIA has laid out an extremely aggressive plan for releasing new chips and systems that make it difficult for any competitors to keep up. 

For example, while NVIDIA began scaling its Blackwell platform last quarter, in the second half of the year it plans to begin production of its Blackwell Ultra systems. 

Then next year, NVIDIA will begin releasing its next platform named ‘Rubin.’ NVIDIA also revealed more details behind what it’s releasing in 2027 and 2028. 

In 2027, the company is releasing ‘Rubin Ultra NVL576’ a system NVIDIA claims is 14X more powerful than the top-of-the-line Blackwell Ultra systems entering production this year. Rubin Ultra is incredibly ambitious, with 4X the number of GPU dies in a single server rack. 

There are three important points to make about NVIDIA’s product roadmap:

1. The upgrades NVIDIA is planning make it very difficult for any leading cloud service provider to not stay on the treadmill of significant upgrades each year. Simply put, not being on the cutting edge makes their offerings vastly inferior to competitors. NVIDIA’s rate of progress is simply too significant for any major customer to ‘stand pat.’ 
2. NVIDIA’s rate of progress isn’t just difficult for a rival like AMD (Nasdaq: AMD) to keep up with, it continues to keep down competitive pressures from application-specific integrated circuits (ASICs) being developed by all major cloud providers (more on this in the next section). 
3. The power demands on systems planned for the coming years continue to soar. For example, Rubin Ultra has five times the power demand per rack compared to Blackwell Ultra. That’s going to require much more spending in areas like liquid cooling and is certainly music to the ears of a company like Vertiv (NYSE: VRT). Yet, the company’s stock was down nearly 5% yesterday. Vertiv’s shares are down 45% from their 52-week high even though industry news continues to move in their favor. 

3. Pushback on the Broadcom Story 

Finally, a key message from yesterday’s GTC Keynote was pushback on the threat posed by ASIC companies like Marvell (Nasdaq: MRVL) and Broadcom. 

Every single cloud provider – in addition to companies like OpenAI and Apple – is engaged with companies like Marvell, Broadcom, Mediatek, and Alchip on designing custom chips for running AI workloads. 

Designing custom chips makes sense from the perspective of cloud companies. First, it creates an opportunity for differentiation. For example, Amazon has been trying to undercut the price of using NVIDIA’s GPUs with its Trainium chips that are designed in cooperation with Marvell. 

Google was the first company to go big on desinging custom chips – its TPUs were designed in collaboration with Broadcom – and it has become a massive advantage for the company. 

Second, designing custom chips puts more pricing pressure on NVIDIA. If cloud companies have a viable alternative, they can be more aggressive when negotiating prices. From this perspective, the threat to NVIDIA isn’t that ASICs provide an ‘existential’ threat, but they could continue creating margin pressure on the company’s 70%+ gross margins that are essential to growing profits in the years ahead. 

NVIDIA spent significant time detailing why their offerings are superior to custom chips. Their argument boils down:

1. The company’s roadmap is aggressive enough that buying its chips still provides a lower TCO (total cost of ownership) than custom chips can provide 
2. Changes to AI – like recent breakthroughs in reasoning – are still fluid enough that general-purpose chips (that NVIDIA makes) are superior to custom chips that need to serve very specific functions and are far more inflexible. 

For now, we believe that the market for AI is big enough that both NVIDIA and a company like Broadcom can succeed. However, this is shaping up to be one of the biggest battles in technology in the years ahead. 

Get the Full Story on AI 

Last year we published our comprehensive report on AI named ‘The Next NVIDIA.‘ We’ll soon be removing it making this your last chance to get this complimentary research report before its gone for good. One of the stocks we featured last year has more than tripled since we launched this report. Make sure to grab a copy before it’s gone for good!