Robots are Coming: Sizing Up the Opportunity

🤖 Robots are coming. Here is the opportunity:
👤 Agentic AI
👨‍💼Humanoid Robots
🚘Autonomous vehicles

In this week’s Tech Edge we dive into the key takeaways from last week’s Consumer Electronics Show (CES) and size up the robotics and autonomous driving opportunity. Our weekly digest is intended to keep you on the cutting edge of investments in data infrastructure, software, and cybersecurity.

🤖 Robots are coming… There were several interesting announcements from Nvidia at CES this week, including an AI PC (Project Digits). But the most significant theme that came out was Nvidia’s focus on enabling Robotics. Nvidia’s CEO, Jensen Huang, believes that the ChatGPT moment for the Robotics industry is fast approaching and highlighted three types of robots: Agentic AI, Humanoid Robots, and Autonomous Vehicles. Let’s dive into each one: 👤 Agentic AI Agentic AI is similar to copilots used today, but unlike AI chatbots that use generative AI to provide responses based on a single interaction, Agentic AI can “think.” These robots can use reasoning and iterative planning to solve complex, multi-step problems autonomously which is referred to as “test time scaling”. Per Nvidia CEO, “Agentic AI is the perfect example of test time scaling.”…this means that it requires significantly more sophisticated types of inference and models that “think” rather than pull out answers from pre-trained models. To learn more about this topic, refer to the Generative AI ➡️Thinking AI 🤔 section of our AI Primer. “People used to tell me, Jensen, inference is easier. Training is hard… There is nothing easy about thinking. Anybody who thinks inference is easy, they’re just not thinking.”– Jensen Huang on test time scaling at CES. Agentic AI uses a four-step process for problem-solving:Perceive: AI agents gather and process data from various sources, such as sensors, databases and digital interfaces.

Reason: A large language model acts as the orchestrator, or reasoning engine, that understands tasks, generates solutions. This step uses techniques like retrieval-augmented generation (RAG).

Act: By integrating with external tools, agentic AI can quickly execute tasks based on the plans it has formulated.

Learn: Agentic AI continuously improves through a feedback loop, or “data flywheel,” where the data generated from its interactions is fed into the system to enhance models. This ability to adapt and become more effective over time is the differentiated feature vs. what we are using today.Applications: The potential applications of agentic AI are vast, covering nearly every service business function, ranging from customer service, content creation, cybersecurity, healthcare, etc.. Companies that can benefit are the ones that provide co-pilot-like tools today and can charge more for extra features such as Microsoft, Gitlab, ServiceNow, Salesforce, Hubspot, Shopify, etc. Regulatory score: Easy ✅

👨‍💼 Humanoid Robots Unlike Agentic AI where test time inference is key, for Humanoid Robots and Autonomous driving model training is key. Specifically for Humanoid Robots, the main challenge is processing imitation information. While this tends to be a laborious process, it will likely require synthetically generated motions, allowing imitation training to occur without physical human demonstration and speeding up training times. The long-term annual revenue opportunity from Humanoid robots is massive. We estimate that 1M robots deployed would be a $150B opportunity. To put this in context, in 2023, roughly 15M people were employed in the private sector manufacturing industry in the US. The opportunity can significantly broaden if the use cases broaden out from manufacturing to other areas and can reach over $1T.

While humanoid robots may appear like SciFi to most people, the interesting point that Jensen brought up is that these robots can go into existing infrastructure making it very easy to scale and regulate (e.g. there are already industrial robots and co-bots used in factories today). Regulatory score: Medium ✅ 🚗 Autonomous Vehicles Over the next 2-3 years we expect that each car company will have to map out an autonomous strategy. The two dominant platforms that companies can choose to pursue are Nvidia and Tesla. “Every single car company will have to be autonomous or you’re not going to be a car company.” – Jensen Huang at CES Nvidia provides the companies with an opportunity to use their platform but requires significant investment and knowhow. Companies like Mercedes have been partners for many years and are sharing 50% of the software revenue generated with Nvidia in exchange for access to the platform. The list is expanding, with Toyota being the most recent new partnership announced at CES. Nvidia expects $5B in automotive revenue in FY26, almost 5 times its FY24 automotive revenue of US$1.1B. For context, our conversations with the company in 2021, prior to when Chat GPT took off, sized up the automotive opportunity as comparable to Data Center. Data Center surprised to the upside, now approaching $200B – will Automotive follow? Tesla, on the other hand, provides a full solution where companies don’t have to build anything in-house. This could be particularly attractive for companies that don’t have the capital or capabilities to invest in autonomy, but can leverage their brand and manufacturing experience. Over time, we believe that Nvidia and Tesla will be able to capture significant market share. We believe that Tesla’s revenue will ramp up quicker due to its internal use case. Nvidia will be relying on partners which will take longer to scale but offers a similar opportunity set. In our base case, we assume 100K autonomous vehicles being a $10-20B opportunity in the next 3 years (~in-line with Nvidia’s $5B assumption by ’26 at ~50% market share); ramping to 1 million vehicles by 2035, which would be a $120-180B opportunity. Longer term, at 100 million vehicles, the opportunity would be in the trillions. For comparison, the number of registered vehicles in the US in 2022 was 283M, making the 100 million assumption not outlandish.

While from regulatory perspective autonomous vehicles are the hardest, they also have the most to benefit from the current administration potentially setting up federal regulation. Regulatory score: Hard 🔴 Even if only 10% of the robotics opportunity materializes across the three different types of robots, it would have a profound impact on the economy. Productivity improvement in the US has stalled over the past 20 years and even longer in the rest of the world. AI could have a transformational impact in re-accelerating GDP growth. Tune in to our 2025 Outlook Webinar to learn more 👇

Watch our CIO, Ivana Delevska, discuss the robotics opportunity with Charles Payne on Fox Business.