Artificial Intelligence (AI) Is Moving Beyond Data Centers. Nvidia Has Already Turned This Opportunity Into a Multibillion-Dollar Business

Nvidia said its physical AI revenue exceeded $9 billion over the trailing 12 months, up from $6 billion in fiscal 2026, implying a 50% increase in the run rate. The article highlights a long runway from robotaxis, robotics, and autonomous systems, plus a new Vera CPU opportunity that management says could open a $200 billion addressable market and add $20 billion in revenue this year. Analysts now expect fiscal 2029 EPS of $15.64 versus $4.77 in fiscal 2026, reflecting a 48% CAGR.

Analysis

The market is likely underappreciating that physical AI is not a single end-market but a second-wave platform pull-through for the entire compute stack. Once robots, vehicles, and factory digital twins move from pilots to fleet deployments, the winners skew toward vendors that own both training and inference plus the software tooling layer; that favors NVDA structurally, but also creates incremental demand for TSM’s leading-edge capacity and select integration partners. The important second-order effect is that physical AI adoption is capex-light for end customers only after the initial systems are proven, so revenue can inflect sharply once unit economics cross a threshold rather than ramp linearly.

The bigger read-through is to Intel and General Motors’ ecosystem. INTC is not a direct winner from the narrative unless it can credibly capture edge inference or factory automation sockets; otherwise, the existence of a fast-growing Nvidia-led platform raises the risk that Intel remains boxed out of the highest-growth workloads. GM is a cleaner beneficiary from factory automation than from autonomy; the near-term gain is margin expansion via throughput and scrap reduction, while the longer-term optionality on autonomy remains binary and capital intensive.

The main risk is timing mismatch: physical AI is a multi-year adoption story, while the stock is pricing in some of that optionality immediately. If enterprise capex tightens, or if robotaxi/humanoid deployments slip 12-18 months, multiple expansion could stall even as long-term TAM remains intact. The contrarian view is that consensus may be overconfident on monetization cadence but still underestimating the breadth of ancillary demand: tooling, simulation, networking, and power delivery could see earlier revenue acceleration than end-device shipments.

Key Decisions for Investors

• Maintain long NVDA into the next 6-12 months, but prefer call spreads over outright equity to capture upside while limiting multiple-compression risk if physical AI revenue ramps slower than expected.
• Overweight TSM versus the broader semicap complex for 6-18 months; physical AI increases demand for advanced-node capacity and packaging, with better earnings visibility than pure-play software beneficiaries.
• Use UBER as a tactical long only on pullbacks; treat robotaxi optionality as a 12-24 month catalyst, not a near-term earnings driver. Consider a small starter position with tight downside, since execution slippage can re-rate the multiple quickly.
• Fade INTC rallies tied to AI edge rhetoric over the next 3-6 months unless there is evidence of share gains in automotive or industrial inference; relative short INTC vs long NVDA remains the cleaner expression.
• For GM, favor a small long only as a productivity/automation beneficiary, not as an autonomy compounder. Pair GM long against a weaker auto OEM if seeking industrial automation exposure without paying for full robotaxi upside.