My take? The issue isn’t a lack of algorithms. It’s that today’s systems remain largely blind to the physical world where businesses operate. Models can spin up brilliance with words and images, but struggle when asked to understand inventory on a shelf, a part on a factory line, or a space a robot must navigate.

That gap is where Physical AI emerges.

Beyond words and pixels

Physical AI refers to AI trained on high-quality 3D datasets. Instead of reasoning only in text or two-dimensional images, these systems can understand scale, geometry, material, and spatial context. A robot stocking a shelf, a virtual assistant designing a living room, or a drone inspecting energy infrastructure all require this kind of embodied intelligence.

In many ways, Physical AI is not a new algorithm but a new substrate. Just as cloud provided the compute and storage layer for software, Physical AI requires an infrastructure of 3D digital twins: accurate, machine-readable representations of products, spaces, and interactions. Without that foundation, applications collapse under their own ambition.

The investor’s blind spot

For investors, the relevance is straightforward. Every major tech cycle has been anchored in infrastructure plays. Data centres and semiconductors powered the Cloud. App stores and broadband powered mobile. Foundation models and GPUs powered generative AI.

Now the race is on to build the infrastructure that allows AI to bridge into the physical. Companies like NVIDIA, Meta, and Alphabet are quietly placing bets on digital twin platforms, simulation engines, and robotics interfaces. They understand that the next wave of AI value, whether in commerce, logistics, healthcare, or design, will come from systems that can perceive and interact with the real world.

As Jensen Huang of NVIDIA recently put it: “Physical AI is the next multi-trillion-dollar market opportunity.” That may sound bold, but the logic echoes past transitions. When the market underestimates a new data layer, outsized returns follow for those who spot it early.

Industries where Physical AI lands first

Where will adoption happen first? Three areas stand out.

• Retail and e-commerce: from automated planograms to immersive product visualization, digital twins already boost conversion rates, cut return costs, and speed up time-to-market
• Industrial operations: factories and warehouses are fertile ground for Physical AI. Training robots to navigate aisles, manipulate inventory, or manage energy usage requires precise, simulation-ready models of real environments
• Healthcare and design: from surgical training to architectural planning, industries where precision and safety matter are rapidly turning to simulation-driven AI to reduce risk and accelerate learning

Each vertical carries risks. Data standardization is still fragmented, and the capital intensity of building 3D pipelines can deter mid-market players. Yet modular approaches are emerging: tools that allow enterprises to start small, prove ROI, and scale without rebuilding their entire tech stack.

Where simulation becomes infrastructure

For years, digital twins were treated as experiments, useful for pilots or marketing campaigns, but rarely central to enterprise strategy. That perception is changing.

Consider a Fortune 50 retailer that decided to replace traditional studio photography with digital twins. Previously, every product launch meant shipping items to studios, coordinating shoots, and waiting weeks for content. The process was slow, costly, and fragmented across categories.

By adopting a digital twin pipeline, the retailer began generating photorealistic, machine-readable 3D assets at scale. The results were transformative:

• 30% reduction in imagery costs by eliminating expensive studio production
• 63% faster production times, allowing new SKUs to go live in days instead of weeks
• Double-digit conversion lifts, with add-to-cart rates rising 17–32% depending on category
• Lower return rates, as customers engaged with more accurate and interactive visuals (360° views, AR-ready assets)

What started as a cost-saving measure quickly became an infrastructure shift. Those same assets now power everything from merchandising automation and catalogue updates to AI-driven personalisation and supply chain simulations.

This illustrates the inflection point: when digital twins move from marketing collateral to operational backbone, simulation stops being an experiment, it becomes infrastructure.

Closing the reality gap

Generative AI has dazzled with words and images, but the real prize lies in teaching machines to deal with the stubborn details of the real (physical) world. History shows us that progress comes when we stop sketching on the whiteboard and start laying bricks.

Physical AI is that next brick-and-mortar moment for technology. It’s less about dazzling outputs and more about building a scaffolding sturdy enough for new industries to climb. Without it, AI remains a bright idea with shaky footing, like the humanoid robots we’ve all seen trying to play football and tumbling with every third step. With it, you get a platform solid enough to carry the weight of factories, supply chains, hospitals, and cities.

Every great structure is remembered by its skyline, but its value depends on the stones hidden underground. Physical AI is that foundation stone. Investors who put capital and conviction there will shape the outlines of the next economy.

For more startup news, check out the other articles on the website, and subscribe to the magazine for free. Listen to The Cereal Entrepreneur podcast for more interviews with entrepreneurs and big-hitters in the startup ecosystem.