Kindred Ventures Founder on Investing Trends From CES

Key Concepts

• Level 4 Autonomy: Full self-driving capability, requiring no human intervention in specific conditions. Differentiated from Level 2+ assistance.
• Autonomy Intelligence Platform: An operating system enabling automakers to integrate self-driving capabilities. (e.g., Nuro’s platform)
• Nvidia’s Full-Stack Solution: A comprehensive offering including chips, open-source models, and datasets for developing self-driving car platforms.
• Inference vs. Training: Inference refers to using a trained AI model to make predictions, while training involves creating the model. Demand for inference compute is predicted to surpass training compute.
• HBM (High Bandwidth Memory): A type of computer memory crucial for AI processing, currently facing a shortage.
• “The Flip”: The point at which demand for inference compute exceeds supply, signifying a major shift in the AI landscape.

The Rise of Robo-Taxis and Nvidia’s Dominance in Autonomy

The discussion centers on the evolving landscape of autonomous vehicles, particularly robo-taxis, and the pivotal role Nvidia is playing in shaping the industry. The partnership between Uber, Nuro, and Lucid was highlighted as a significant development, demonstrating a “capital light” model for deploying self-driving vehicles. This involves leveraging Nuro’s autonomous technology and Lucid’s vehicles within the Uber network, aiming for a launch in the Bay Area later this year. This approach allows Uber to focus on fleet management rather than the substantial investment in hardware development.

Nuro and the Autonomy Intelligence Platform

Nuro stands out as a key player, having been actively testing autonomous vehicles on Bay Area streets with cameras actively recording. Their recent launch of an “autonomy intelligence platform” is crucial, functioning as an operating system that allows third-party automakers to integrate self-driving capabilities into their vehicles. This platform is a core component of the Uber-Lucid partnership, enabling Level 4 autonomy – true self-driving – rather than the more common Level 2+ driver-assistance systems.

Nvidia’s Strategic Play: Becoming the “Fed of AI”

However, the most significant development discussed is Nvidia’s aggressive push into the autonomous vehicle space. Nvidia isn’t simply selling chips; they are offering a complete “full-stack solution” – a starter kit including chips, open-source models, and open datasets. Their strategy is to foster competition among developers, encouraging them to build self-driving platforms tailored to various applications (ride-sharing, delivery, logistics) while simultaneously increasing demand for their GPUs.

As stated, Nvidia aims to be “like the Fed for A.I.,” setting the direction and providing the necessary tools (models, data, and compute) for others to follow. This is a deliberate effort to maintain and expand their dominance, countering arguments about a potential decline in their market position. Nvidia is an investor in Nuro, making the Uber-Nuro partnership a key showcase for their technology. The shift from focusing solely on robotics to emphasizing “autonomy and mobility” signifies a recognition that the field is reaching a critical point.

Market Consolidation and the “Rule of Thumb”

The conversation emphasizes a predictable pattern in Silicon Valley: initial proliferation of competitors followed by consolidation down to two or three dominant players. Currently, the autonomous vehicle landscape includes Nuro, Waymo, Tesla, and several Chinese competitors, with a notable absence of significant players in Europe. The high resource requirements – including extensive training data and years of on-road testing – are driving this consolidation. This trend is expected to extend to robotics, with Boston Dynamics (owned by Honda) emerging as a potential leader, and generative media/AI agents.

Robotics and the Practicality of Humanoids

Boston Dynamics’ humanoid robot was also discussed, with a noted shift in perspective. The speaker, previously skeptical of humanoid development, now believes Boston Dynamics has created a practical and functional humanoid robot. This represents a significant step forward in the field.

The “Flip” and the Compute Battle Royale

A crucial point raised is the impending “flip” – the moment when demand for inference compute surpasses demand for training compute. This signifies a shift from building AI models to deploying and using them. This shift is expected to exacerbate the existing shortage of High Bandwidth Memory (HBM), leading to a “battle royale” for compute resources. The speaker likened this to a significant event in the cryptocurrency world, highlighting its importance.

Key Takeaways

The discussion paints a picture of a rapidly evolving autonomous vehicle industry, with Nvidia positioning itself as a central and controlling force. The Uber-Nuro-Lucid partnership represents a practical approach to deploying robo-taxis, while Nvidia’s full-stack solution aims to democratize (and ultimately profit from) the development of autonomous systems. The anticipated market consolidation, coupled with the impending “flip” in compute demand, suggests a period of intense competition and strategic maneuvering in the coming year. The emergence of practical humanoid robots, like those from Boston Dynamics, further underscores the accelerating pace of innovation in robotics and AI.