What the Future of AI Means

Banstanks/Adobe Stock

AI is moving from screens into the physical world: first as a collaborative partner in our everyday work, and increasingly as embodied humanoid robots that will share homes, factories, and public spaces over the next decade. This is no longer a sci-fi thought experiment; it is an investment theme and a practical reality that will touch how we write, work, care for family, and build wealth.

AI as an everyday collaborator

For most of us, the first wave of this change shows up not as a robot walking through the front door, but as an invisible partner sitting inside our laptops and phones. Modern AI can already help structure a note to a client, draft a tricky email, translate technical jargon into plain English, or tailor a presentation to different stakeholders in seconds. Very soon, asking AI for help on a memo, a marketing plan, or a legal summary will feel as natural and unremarkable as sending a quick text to a colleague.

And the line between this “soft” AI and robots in the physical world is thinner than it appears. The same models that help with writing and analysis are being wired into machines that can see, move, and manipulate the world around us. The idea that you might one day ask your home robot to “clean up after dinner and set out the kids’ backpacks for tomorrow” is closer than most people think.

Humanoid robots today

Humanoid robots are already out of the lab and into the field. Systems like Tesla’s Optimus, Agility Robotics’ Digit, Boston Dynamics’ electric Atlas, and Unitree’s G-series robots can walk, climb stairs, navigate cluttered environments, and use multi-fingered hands or grippers to move objects reliably in structured settings. They are starting where the economics are clearest: warehouses, factories, and logistics operations that run 24/7 and struggle to hire and retain enough people for physically demanding, repetitive work.

These early humanoids can combine cameras, depth sensors, force feedback, and AI models to follow spoken instructions, unload containers, pick and place items, and even act as greeters or guides in commercial spaces. But this is still the first inning. Costs are high, safety and regulatory standards are tightening, and robots are much more comfortable in a warehouse aisle than in your living room, where the “data set” includes kids’ toys, pets, and spilled juice.

The next 3–5 years

Over the next three to five years, the real scaling story will likely play out in industrial and logistics environments. Manufacturing floors, distribution centers, and large commercial properties are natural habitats for this first generation of capable humanoids: the tasks are repetitive, the environments are structured, and the return on investment is easier to quantify. At the same time, we will start to see the first serious attempts at household robots — expensive, early-adopter products that take on a narrow band of chores under human supervision.

Behind the scenes, the technology is improving at an impressive pace. Research teams are already demonstrating humanoid manipulators that succeed on diverse object-handling tasks roughly nine times out of ten, while learning to generalize to new tasks without being hand-coded for each one. That kind of reliability is essential if you want a robot to safely handle your groceries, not just move boxes on a test bench.

Musk and Huang: from possible to probable

Two leaders worth listening to on what is “possible” versus “probable” in humanoid robotics are Elon Musk and Jensen Huang. Musk has been very clear: in his view, demand for humanoid labor will be effectively insatiable, with hundreds of millions — potentially around a billion — robots working alongside humans globally by the 2030s and 2040s. If he is directionally right, the impact on labor markets, productivity, and the meaning of “having a job” will be profound.

Jensen Huang, as CEO of Nvidia, comes at the problem from the compute side, but his message rhymes with Musk’s. He often says that “everything that moves” will eventually be robotic, and that we are heading toward a “ChatGPT moment for general robotics” as AI, sensors, and simulation finally converge. Huang talks openly about a world with on the order of a billion humanoid robots in the mid-2030s, trained first in photorealistic virtual environments and then deployed into the messy real world. In his framing, life gets easier for many people — but the upside belongs disproportionately to those willing to pair their curiosity and ideas with the leverage of these new tools.

10+ years: robots in the family

Stretch the timeline beyond a decade, and the picture becomes even more personal. Demographically, many economies are aging, birth rates are falling, and care work is becoming both scarcer and more expensive. In that context, it is easy to imagine a typical family relying on one or more robots to assist with elder care, support children’s routines, manage everyday logistics, and provide a layer of safety and monitoring for health events.

In practical terms, this could mean a robot that does heavy lifting, repetitive cleaning, basic food preparation, and nighttime checks, while also acting as a patient tutor or language partner for kids. It will not replace parents, nurses, or teachers — but it will change how their time is spent, potentially freeing more of it for high-value emotional and intellectual work. If these systems scale the way many forecasts suggest, households could come to see intelligent machines less as gadgets and more as part of the extended “support network” that keeps life running.

The nervous system of technology: semiconductors

Underneath all of this — cloud AI, on-device copilots, humanoid robots — is one essential ingredient: semiconductors. If data is the fuel of the AI era, then chips are the nervous system that makes the whole organism function: they sense, process, communicate, and control. A single humanoid robot will rely on a hierarchy of silicon — high-end AI accelerators for perception and planning, specialized processors for vision and audio, microcontrollers for every joint and actuator, dedicated chips for safety, connectivity, power management, and often a connection to remote GPUs in the cloud.

Investors used to be told, during the Gold Rush and its many echoes, to “sell picks and shovels to the miners.” In today’s language, that advice points directly at semiconductors and the broader manufacturing, tooling, and design ecosystem around them. At our firm, we have put that belief into practice with a dedicated option-driven strategy for semiconductor exposure*. The strategy is designed with a goal of moving roughly one-for-one with the market on the downside but participating one-to-many when the sector rises, with convex upside rather than linear returns. We believe that this strategy is an example of how thoughtful use of derivatives can amplify a high-conviction thesis without abandoning risk discipline.

For anyone who cares about crypto, high-performance computing, industrial automation, or AI more broadly, the case for focusing on semiconductors — the nervous system of this emerging humanoid era — is straightforward. If this future resonates with you, it may be the right time to talk about how to own the “picks and shovels” of the next great technology cycle.