From Robot Brains to Nervous Systems: 3 Stocks Leading Physical AI

Robots are no longer a distant-future fantasy; they are an investment theme already priced into today's stock market. The hosts open with a striking claim: accelerators small enough to sit inside a robot can now perform two quadrillion operations per second, and they make it vivid by noting that counting one operation per second would force you to start 66 million years ago, when the dinosaurs vanished. The claim sounds theatrical, yet Nvidia's own engineering record backs the substance: according to the Nvidia developer blog, the Jetson Thor platform reaches 2070 teraflops of FP4 compute, roughly two quadrillion operations each second. To me, that number moves robotics stocks out of the fantasy basket and into a measurable hardware story.

The narrator then builds the foundation: physical AI means intelligence stepping out of the chat window and into machines that move in the real world. Grasping a coffee cup feels trivial to us; eyes locate it, the brain judges distance, the hand reaches out. For a robot, the same act demands sensing the environment, deciding instantly, and executing motion with almost no delay, which is why compute must live inside the machine rather than in the cloud. The show frames this need as a three-layer stack: the brain that decides, the nervous system that carries signals, and the testing layer that verifies everything. Nvidia's Jetson Thor documentation tells the same story, describing a platform built for real-time multimodal perception at the edge.

A three-layer stack: brain, nervous system, testing

Nvidia plays the brain in this story, and the hosts place its Jetson Thor robot computer at the center of the stock thesis. Built on Blackwell graphics architecture, the platform packs 128 GB of memory into a 130-watt power envelope, promising 7.5 times the compute of the previous Orin generation at 3.5 times the energy efficiency, all confirmed in the Nvidia developer blog. On software, there is GR00T, the generalist robot model introduced through a 2024 announcement carried by GlobeNewswire, which learns coordination and dexterity by observing human movement inside the Isaac robotics platform. According to FierceElectronics' GTC notes, Apptronik's Apollo humanoid and Agility's Digit walked that ecosystem's showcase floor. Nvidia, in other words, sells not just accelerators but the learning infrastructure of robots.

Now to the earnings side, because this show is a stock analysis, not a product launch. Nvidia reported second-quarter fiscal 2027 revenue of $96.2 billion , up 106 percent year over year, with data-center revenue of $89 billion , up 117 percent. Gross margin stood at 75 percent and diluted earnings per share came in at $2.46. Management guided third-quarter revenue to $108 billion . Reuters reports the company expects fiscal 2028 revenue to grow about 70 percent, with 2 million additional accelerators to be deployed alongside AWS across 2027 and 2028. The show's cited projections of $400 billion in 2027, $700 billion in 2028, and $924 billion in 2029 revenue are analyst estimates relayed by the hosts, not reported figures, and should be read as such. A 0.37 PEG ratio makes the valuation look cheap, but that ratio leans on aggressive growth assumptions. Institutional flow data looks choppy across the year, suggesting smart money has not bought the story unconditionally; the quarter is strong, yet much of the triumph may already be priced in.

The second layer belongs to Analog Devices, likened by the hosts to the robot's nervous system. The logic is simple: however powerful the brain, a robot drops the cup if touch and position data arrive late. The company's sensors, motor controls, and tactile sensing solutions exist to move that data to the brain with minimal delay. The show notes that management flagged the data-center and industrial segments as growth engines on its latest earnings call, with the industrial arm nearing $2 billion in quarterly sales. In my view this is the least discussed yet most indispensable link in the robotics chain: a rival can fabricate an accelerator, but decades of analog know-how are far harder to copy.

Two stocks across the nervous system and testing layers

Analog Devices' reported numbers support the show's thesis. According to the company bulletin published via Nasdaq, Analog Devices posted record fiscal third-quarter 2026 revenue of $4.02 billion , led by the data-center and industrial segments. Trailing-twelve-month operating cash flow reached $5.5 billion with $4.9 billion of free cash flow, and the company returned $1.7 billion to shareholders during the quarter. The earnings trajectory shared on the show points from $12.87 toward nearly $20 by 2028, against a market valuation near $200 billion . A 20.6 percent net margin with 29.4 percent one-year earnings growth explains why this quiet giant sits on institutional flow radars. From a portfolio angle, ADI shines less brightly than Nvidia but offers a steadier cash-backed profile.

The third stock is Keysight Technologies, occupying the most critical role of all: the testing and integrity layer. The hosts describe its solutions for validating AI model accuracy and battery systems, including robotic battery rigs that reproduce real operating conditions in the lab. Why does it matter? One faulty sensor or dying cell means downtime in a factory robot and potentially lives at stake on the road, a point consistent with Keysight's own bulletin on engineering challenges. The electronic-industrial arm reportedly booked about $0.5 billion in quarterly revenue, while orders topped $2 billion for a second straight record quarter. The testing layer may look small, but its demand compounds as robotics scales.

Keysight's earnings confirm the momentum. Per its company bulletin, third-quarter revenue rose from $1.35 billion to $1.85 billion ; GAAP net income hit $397 million with earnings of $2.30 per share. On an adjusted basis, earnings reached $3.07 per share and operating cash flow printed $437 million . The trajectory shared on the show runs from $11.46 in 2026 past $15 in 2028, and management raised the full-year outlook. Record orders, as reported by Keysight, suggest investor confidence stretches beyond the current quarter. On valuation , growth versus price looks calmer than Nvidia's, which makes the testing layer a stabilizing element in a portfolio .

Market context and portfolio logic

These three stocks deserve to be read together with the wave behind them. Research carried by MorganStanley sees the humanoid market reaching $5 trillion by 2050, with nearly 1 billion robots, 90 percent of them working in industrial and commercial settings. A Barclays study cited by CNBC expects the market to hit $200 billion within a decade, while Wedbush analyst Dan Ives talks in trillions over ten years. SoftBank chief Masayoshi Son argues the next trillion-dollar company will emerge from physical intelligence. Yet Chinese makers reportedly hold 97 percent of humanoid shipments, so Western technological leads do not yet translate into manufacturing scale. That backdrop explains the show's infrastructure-layer picks: whoever builds the robot, the brain, the nervous system, and the test gear remain necessary.

My summary of the hosts' logic is crisp: join the robotics revolution not through a single robot maker but through three stocks entrenched in the stack's indispensable layers. Nvidia brings growth, Analog Devices brings cash balance, Keysight brings low-visibility but critical test revenue. Still, this is no rally invitation: valuations may already discount high expectations, memory-supply constraints feature in Nvidia's own warning, and Chinese competition could squeeze margins. My investor takeaway is to put the trio on a watchlist and track cash and order trends quarter by quarter around earnings season rather than loading up at once. Portfolios compound on discipline, not excitement.