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

Nvidia's physical AI revenue exceeded $9 billion on a trailing 12-month basis, up from $6 billion in fiscal 2026, implying a 50% increase in quarterly run rate. The article highlights a potential $960 billion physical AI market by 2033, a $200 billion addressable market for stand-alone Vera CPUs, and consensus EPS rising to $15.64 by fiscal 2029 from $4.77 in fiscal 2026. Overall, it frames Nvidia as well positioned to benefit from expanding AI infrastructure and edge AI demand.

Analysis

The market is still pricing NVDA as a data-center capex proxy, but the more important option value is now in its ability to become the operating system for physical autonomy. That matters because the spend profile changes: instead of a single hyperscaler buying tens of thousands of GPUs in one tranche, you get a longer-duration, multi-vertical rollout across robotics, mobility, and industrial automation. The result should be a smoother revenue curve, higher software attach, and more durable gross margins than a pure hardware cycle would imply.

The second-order winners are not just NVDA’s obvious OEM partners, but the ecosystem that can turn prototypes into manufacturable fleets. TSM benefits if physical AI accelerates edge compute demand into custom silicon, advanced packaging, and more diversified wafer demand; GM gains if factory automation improves throughput faster than the market is discounting; UBER is effectively a distribution layer for autonomy, but only if the unit economics beat human-driver economics in enough geographies. The less obvious loser is any legacy industrial automation vendor that depends on closed architecture and slow deployment cycles — physical AI compresses adoption timelines and shifts value toward whoever controls the model stack and developer tooling.

The key risk is that the market is extrapolating a decade of adoption from a handful of pilot programs. Robotaxis, humanoids, and factory twins are all technically compelling but operationally bottlenecked by regulation, safety certification, and integration costs; those delays can easily push monetization 12-24 months to the right without breaking the long-term thesis. A near-term reversal would likely come from guidance that physical AI remains de minimis versus core data-center demand, which would force investors to unwind the “new TAM” premium quickly.

Consensus may be underestimating how much of NVDA’s upside is already embedded in the stock, while still underestimating how much optionality exists in adjacent names. If physical AI becomes real, the supply chain intensity favors picks-and-shovels around fabs, packaging, and fleet deployment more than the application layer in the first wave. That argues for owning the platform enablers and expressing skepticism through names where the narrative is ahead of deployment economics.

Key Decisions for Investors

• Maintain a core long NVDA, but finance it with short-dated upside overwrites into earnings cycles; the stock still has multi-year secular upside, but near-term implied volatility likely overprices the physical-AI narrative versus actual revenue contribution.
• Add a relative-value long TSM / short INTC pair for the next 6-12 months; if physical AI drives more heterogeneous edge silicon and packaging demand, TSM captures the incremental mix shift while INTC remains more exposed to execution risk and slower design-win conversion.
• Initiate a small long UBER position on a 6-12 month horizon only if paired with evidence of city-by-city autonomy rollout; otherwise use call spreads rather than common, as the market may be paying for a 2028 narrative well before economics are proven.
• Buy a GM call spread or capped-risk equity position into any weakness over the next 3-6 months; factory automation is a margin lever if executed well, but the trade should be sized for binary integration risk and slower-than-expected adoption.
• If NVDA rallies sharply on physical-AI headlines without corresponding order commentary, consider a pair trade long NVDA / short a basket of overextended industrial automation names; the goal is to own the platform while fading the first-order narrative premium elsewhere.