At CES 2026, under the spotlight, Jensen Huang, founder and CEO of NVIDIA（NVDA）, pointed to a slide featuring a compact robot and said, “I love Serve Robotics（SERV）.” That single remark not only sent the shares of the relatively obscure company soaring over 14% in a single day but also triggered profound reflections in the capital markets regarding the value proposition of “Physical AI.”
What makes Serve Robotics so special that it became the exemplar of what Huang calls the “next generation of AI” — Physical AI — and attracted intense scrutiny from Wall Street analysts?
Huang clearly stated in his keynote: “The next generation of AI is physical AI.” This is not an empty slogan but represents NVIDIA’s strategic evolution from a pure-play computing power provider to an enabler of intelligent solutions. The core value of Serve Robotics lies in its translation of AI algorithms from data centers to the real-world “last hundred meters” of city sidewalks.
Serve’s delivery robots are, in essence, mobile, miniaturized AI edge-computing units, continuously handling complex tasks such as visual navigation, pedestrian avoidance, and path planning. This perfectly aligns with NVIDIA’s grand narrative of pushing AI from the virtual world (like ChatGPT) into the physical realm (robotics, autonomous driving). Investing in Serve is, therefore, a bet on the materialization of the NVIDIA ecosystem within tangible, real-world scenarios.
Northland Securities dubbed Serve “one of the best investments in physical AI,” with the core logic being that it appears to have cracked the key bottleneck for commercial deployment. Compared to the technologically dazzling but costly autonomous vehicles, Serve’s sidewalk robots sidestep complex public road regulations and high hardware costs by focusing on the relatively enclosed, low-speed, and predictable sidewalk environment.
This “niche-focused” strategy has enabled rapid scaling of its partnerships with giants like Uber Eats and Shake Shack. Analysts project Serve’s revenue will skyrocket from $2.5 million in 2025 to $25 million in 2026, validating the viability and explosive growth potential of its business model within this specific domain.
Although its primary application is currently food delivery, the long-term investment thesis hinges on the scalability of its technology. Its autonomous mobile robot platform could, in the future, be equipped with sensors for urban inspection, short-haul logistics, retail restocking, and even security patrols. This week’s acquisitions in robotics by Grab and Mobileye have already demonstrated the intense demand among giants for “last-mile” automation solutions.
Serve’s scarcity lies in the vast repository of real-world, urban sidewalk operational data it has accumulated through commercial partnerships — a contextual data barrier difficult for latecomers to replicate quickly. This data is crucial for the continuous training and refinement of AI models, forming its core competitive moat.
The market enthusiasm has significantly inflated Serve’s valuation. Even based on the projected 2026 revenue of $25 million, the stock trades at a hefty forward Price-to-Sales (P/S) ratio of 40x — undoubtedly a premium valuation for a development-stage company. Investors must remain clear-eyed about the following:
Jensen Huang’s endorsement is far from a simple shout-out; it represents a top-tier technological and strategic validation for Serve Robotics. The deep-seated logic is that Serve serves as a “perfect case study” connecting NVIDIA’s AI empire to the vast landscape of real-world economic applications. Its story is not merely about robots delivering meals but represents an early chapter in how AI steps out of server racks to take on real-world physical tasks.
For investors, it presents a tempting entry point into the Physical AI revolution, yet one that requires caution regarding the significant volatility and risks inherent in any early-stage venture.
