Is physical AI already on the eve of an explosion? Nvidia and Amazon's $1.4 billion joint investment reveals a new cycle in industrial capital

In June 2026, German humanoid robot company Neura Robotics announced the completion of Series C funding, with a total raise potentially reaching $1.4 billion and a valuation of approximately $7 billion. This round attracted major institutions such as NVIDIA, Amazon, Qualcomm, Bosch, the European Investment Bank, and stablecoin issuer Tether. This is not an isolated funding event. According to Dealroom data, since 2026, robotics companies have raised a total of $55.8 billion, setting a new record and nearly doubling the previous year's high. Capital is flowing into the robotics + AI convergence track at an unprecedented speed and density.

The industry’s strategic positioning for this track is becoming increasingly clear. NVIDIA CEO Jensen Huang summarized the evolution of AI technology into three paradigms—perception AI, generative AI, and agentic AI—while the next stage will be physical AI—"AI capable of running, reasoning, planning, and acting." Amazon Web Services and MassRobotics jointly launched the Physical AI Fellowship acceleration program with NVIDIA, and the second phase in 2026 is now open for global robotics startups to apply. Whether in terms of capital inflow or strategic deployment by leading tech companies, physical AI has moved from proof-of-concept to the eve of large-scale deployment.

Scope, Scale, and Market Structure of Physical AI

The core of physical AI is to enable AI to move beyond the digital realm into real-world physical environments. According to MarketsandMarkets, physical AI refers to the integration of artificial intelligence into physical systems such as robots, autonomous vehicles, drones, and industrial equipment, enabling these systems to perceive, analyze, and interact with the real world. Unlike traditional AI that generates text or images, physical AI outputs involve physical actions—moving objects, assembling, transporting—actual physical movements in the real world. Zhejiang Securities' industry deep-dive report notes that physical AI must answer two core questions: how the world will change next, and how the world will respond after physical actions occur. This technological capability underpins three key scenarios: autonomous driving, embodied intelligence, and industrial software.

Market size estimates vary widely depending on the scope, but there is a high consensus on growth direction. MarketsandMarkets projects the global physical AI market will grow from $1.5 billion in 2026 to $15.24 billion in 2032, with a CAGR of 47.2%. Using the broadest scope—covering all AI-enabled physical systems including industrial robots, autonomous vehicles, surgical robots, military automation, and smart infrastructure—the global market size in 2026 is approximately $383 billion, expected to reach $3.26 trillion by 2040. A macro perspective from hedge fund Coatue Management estimates the total physical AI market at at least $6 trillion, about 50% higher than the pure digital AI market. Jensen Huang further stated at CES 2026 that physical AI has the potential to reshape manufacturing and logistics industries worth about $50 trillion. Despite the significant differences in estimates, they all point to the same conclusion: the market scale of physical AI is transitioning from hundreds of billions to trillions of dollars.

Demand-side pressures are equally significant. About 2.5 billion people worldwide engage in physical labor, producing roughly $50 trillion in annual GDP. Accelerating aging trends mean labor shortages in manufacturing, logistics, and healthcare are continuously widening. Meanwhile, the costs of sensors, cameras, and robot-grade processors are rapidly decreasing, and the maturity of generative AI and agentic AI technologies is climbing. These three factors form a structural momentum driving physical AI deployment. During this window of simultaneous maturity on both demand and supply sides, large-scale capital inflows are a natural industry outcome.

Competitive Landscape and Product Differentiation of Global Physical AI Companies

The $1.4 billion funding for Neura Robotics is noteworthy not only for its size but also because it reveals a multi-tiered, multi-technology-path parallel competition pattern in the physical AI track. According to public data, leading humanoid robot companies by funding include: Figure AI with about $1.75 billion raised and a valuation of $39 billion; UBTECH with about $940 million; Apptronik with about $1 billion and a valuation of around $5 billion; Agility Robotics with about $330 million and a valuation between $1 billion and $1.75 billion; and Neura Robotics with a valuation of about $7 billion after this round. Additionally, Boston Dynamics continues to push Atlas humanoid commercialization within the modern automotive ecosystem.

These companies differ significantly in technical approaches and business models. Figure AI, founded in 2022 by serial entrepreneur Brett Adcock, expanded rapidly through VC-heavy funding, receiving investments from NVIDIA, Microsoft, OpenAI’s startup fund, and Amazon founder Bezos during Series B. Its Figure 03 household robot is priced around $20k, targeting the consumer market. Apptronik adopts an industrial alliance model, with about $1 billion raised, strategic partnerships with Google DeepMind, GXO Logistics, and Mercedes-Benz, and its Apollo robot designed as a universal platform with both bipedal and wheeled configurations, advancing toward mass production in Texas and California. Agility Robotics focuses on logistics, with its Digit humanoid deployed in Amazon warehouses, with investments from Amazon, NVIDIA, and SoftBank. Boston Dynamics exemplifies another path—after acquiring 80% of Atlas from Hyundai for $880 million, leveraging automotive manufacturing resources to accelerate commercialization.

The Chinese market also features a clear multi-layered competitive structure. Over 200 humanoid robot concept stocks are listed on A-shares, with a combined market cap exceeding 6.1 trillion yuan. YuShu Technology has passed the STAR Market IPO review and is expected to become the first humanoid robot stock in A-shares in Q3. Tesla’s Optimus V3 is expected to start mass production in summer 2026, and BYD has announced entry into the humanoid robot track with project code “Yao Shun Yu,” planning to deploy 20k units in its own factories in 2026. The Xi’an Robot Industry Park has an initial phase with a capacity of 50k units annually, priced below 200k yuan each. From the supply chain perspective, companies like Midea Group, Shenghong Technology, Lens Technology, Inovance, and Ganfeng Lithium are deeply involved in the humanoid robot track.

Particularly noteworthy is NVIDIA’s key role in the entire physical AI ecosystem. As a global leader in GPU and edge computing chips, NVIDIA’s Isaac GR00T development platform has become a universal base in the industry. NVIDIA partnered with YuShu Technology to launch the first humanoid robot reference design H2 Plus based on the open-source Isaac GR00T platform, and announced the next-generation chip Feynman, designed specifically for physical AI, expected in 2028. This three-layer structure of chips + algorithms + platforms positions NVIDIA as an infrastructure provider in the physical AI ecosystem—complementing AWS’s strategy of engaging physical AI startups through cloud computing resources. In March 2026, Neura Robotics announced a strategic partnership with AWS to expand the Neuraverse platform globally via AWS.

Case Study: Neura Robotics — An Example of Investment Logic in Physical AI

The $1.4 billion funding, $7 billion valuation, and participation from over ten top-tier institutions make Neura Robotics’ Series C one of the most emblematic deals in the physical AI field in 2026. Analyzing this case provides insight into the core logic behind current industry capital choices in physical AI investments.

Technologically, Neura Robotics employs a multi-form product strategy, with product lines including the humanoid robot 4NE1, consumer-grade wheeled robot MiPA, and warehouse transport series MAV, all driven by the AURA AI navigation system. This multi-product approach allows simultaneous data collection across industrial, logistics, and consumer markets, creating a data feedback loop to improve algorithms. The company states that the funds will be used for three main areas: global deployment of humanoid robots, expansion of production and delivery capacity, and R&D of next-generation physical AI systems. These three directions correspond to the typical stages of physical AI companies—from “technology validation” to “scaling” and “paradigm upgrading.”

On the capital side, Neura Robotics’ investors are highly diversified, including chip giant Qualcomm, tech giants Amazon and NVIDIA, industrial infrastructure players Bosch and Schaeffler, and even the relatively regulated Tether. Notably, Tether’s investment is purely equity without any blockchain protocol or token issuance, indicating that institutional investors’ interest in physical AI has moved beyond conceptual hype into substantive industry capital allocation. This cross-sector capital synergy suggests that physical AI is evolving from a single hardware technology track into a multi-industry integrated platform.

However, there are important risks to identify. First, whether the $1.4 billion can be fully received depends on the company’s ability to meet certain development milestones, including mass production capacity, order fulfillment, and commercialization progress. Second, Neura’s previous fundraising of about $55 million in 2023, which grew to over $1 billion in three years, reflects rapid sector enthusiasm but also increases market expectations and pressure on delivery and valuation. Third, the overall humanoid robot field faces high homogenization risk—many leading companies share similar technical routes, application scenarios, and target customers, requiring more operational data to establish and verify differentiation.

Investment and Risk Framework for Physical AI

Based on the above, physical AI as an investment track can be summarized into three interconnected logical layers.

First is the infrastructure layer. Chips and computing power are the core underlying support. NVIDIA, with its GPU advantage and ecosystem of robot software platforms, dominates this layer. Qualcomm and other chipmakers enter via edge computing SoCs. Hardware in the physical AI market accounted for the largest share in 2025-2026. This layer’s investment logic is relatively mature, but the competitive landscape has stabilized, with incremental growth mainly driven by downstream application expansion increasing computational demand.

Second is the robot body and platform layer. Currently the most heavily funded area globally, including companies like Figure AI, Apptronik, Agility Robotics, Boston Dynamics, UBTECH, and YuShu Technology. This layer combines hardware manufacturing and software algorithms, with the highest investment barriers and the greatest divergence in technical routes. Differentiation lies in three aspects: mechanical design (bipedal vs. wheeled vs. hybrid), AI decision architecture (centralized vs. distributed), and scene focus (industrial logistics vs. household service vs. public safety). No clear leader has emerged, so investing here requires evaluating both engineering capability and algorithm strength; technical lead in one dimension alone is insufficient for long-term advantage.

Third is the industry solutions and data services layer. This involves providing end-to-end automation solutions for specific scenarios based on underlying chips and platforms, and accumulating real-world physical data during operation. Amazon AWS and NVIDIA’s Physical AI Fellowship exemplify early-stage efforts—offering technical and computational resources to help startups overcome R&D hurdles. This layer’s business model resembles SaaS, but commercial maturity requires more time.

Correspondingly, risks include: the high uncertainty of humanoid robot mass production—supply chain stability, quality control, and cost curves are yet to be validated at scale; the convergence risk of technical routes—current models and visual-language-action approaches are still in early, non-converged stages, and breakthroughs could render previous investments obsolete; safety and boundary issues—Deepu Talla from NVIDIA emphasized that physical AI development involves full lifecycle management from data generation to safe deployment, with any failure potentially causing irreversible physical consequences, slowing deployment compared to pure digital AI; macroeconomic fluctuations, geopolitical impacts on supply chains, and regulatory changes in major economies also influence valuation and development pace.

Conclusion

From NVIDIA and Amazon’s combined $1.4 billion investment to the global $55.8 billion annual robotics funding, the physical AI track is at a critical inflection point driven by both capital and industry. This track uniquely embodies the paradigm shift of AI from digital to physical, involving deep integration with semiconductor chips, sensors, motion control, and industrial automation.

For investors and researchers, understanding the essence of physical AI—enabling AI to perceive, reason, and execute physical actions in a closed loop—is fundamental to building effective analysis frameworks. Tracking the technological differentiation, commercialization progress, and capital structure changes of leading companies is key to identifying investment timing and industry turning points. Whether physical AI can reach the tens of trillions of dollars in scale envisioned by Coatue and Jensen Huang depends on breakthroughs in technology maturity, mass production feasibility, and safety control. But at this moment, it is relatively clear that physical AI is no longer a distant sci-fi concept but an industry on the cusp of large-scale deployment.