Physical AI diligence for dealmakers: PwC

Alex Baker

Technology, Media, and Telecommunications Deals Leader, PwC US

Email

Michael Fiore

Industrial Products Deals Leader, PwC US

Email

Aaron Tweadey

Deals Strategy and Value Creation, Principal, PwC US

Email

Harry Singh

Deals Strategy and Value Creation, Director, PwC US

Email

Key takeaways:

Physical AI blends hardware and software economics: it pairs capital-intensive hardware complexity with software-like data moats and recurring revenue, so it can’t be diligenced like pure SaaS or traditional industrials.
The market is accelerating and investable now: falling model costs, better sim-to-real transfer, policy-driven automation, and maturing liability frameworks are pushing physical AI toward a ~$450B core market by 2030 (>$1T including the ecosystem).
Deal activity is most actionable in industrial/logistics and connected infrastructure: these environments are more controlled, ROI is measurable, RaaS models are expanding adoption, and enabling layers (e.g., sensors/LiDAR, machine vision) are consolidating.
Value creation is engineering-led: scalable deployments, BOM and supply-chain resilience, design-for-manufacturability, recurring revenue attachment, and serviceability/field reliability (e.g., OTA updates, improving RMA rates) drive margins and durability.
Diligence must prove operating reality—not just “proprietary AI” claims: use teardown + replicability testing, manufacturing readiness and BOM analysis, integration/edge-cloud architecture review, and reliability/service economics to uncover scale, recall, concentration, and regulatory risks.

Physical AI refers to artificial intelligence embedded in machines, devices, and systems that sense, decide, and act in the physical world. Unlike software AI, which generates content or insights in digital environments, physical AI is deployed in systems such as robots, vehicles, medical devices, industrial equipment, and connected infrastructure. It combines AI models with sensors, motors, and control systems to perform tasks like navigation, inspection, object handling, and autonomous decision-making using real-world data.

Autonomous systems, connected devices, and AI-embedded industrial equipment are generating revenue, customer retention, and deployment track records that investors can underwrite. What hasn't kept pace is how most deal teams approach these investments. Traditional diligence and value creation frameworks were built for software or conventional industrials, and physical AI doesn't fit cleanly into either. It combines the hardware complexity of capital-intensive businesses with the data moats and recurring revenue of software. The investors who get this right will likely be able to update their analytical lens before they need it.

That means framing diligence questions around how well a target can turn technical promise into repeatable deployment and resilient economics, while also identifying where the real value creation levers sit once you own the asset. On the diligence side, it requires testing claims of proprietary AI against operating evidence, including device teardown, AI replicability testing, bill of materials resilience, manufacturing readiness, and field reliability, before underwriting scale, margin durability, and defensibility. On the value-creation side, the levers that move the needle—supply chain redesign, software monetization, and serviceability at scale—are often embedded in engineering and product decisions that traditional deal teams aren't set up to assess.

Let’s look at where deal activity is developing, which operating levers shape economics at scale, and how rigorous diligence separates product claims from operating reality.

The physical AI market is evolving

Four developments are making physical AI more relevant to technology, media, and telecommunications (TMT) deal teams in 2026.

The cost of running large-scale AI models is lowering barriers. The cost of running AI models dropped roughly 90% between 2023 and 2025, making it more practical to embed intelligence directly into factory equipment, vehicles, and industrial systems for tasks like quality inspection, navigation, and maintenance.
Simulation-to-real transfer crossed a quality threshold. Simulation technology now enables companies to train AI systems in digital twins of factories, warehouses, and roads, then deploy them in real facilities with minimal rework.
Policy is creating demand for automation in some markets. US and EU reshoring mandates and tariff escalations are compelling manufacturers to automate domestic production, creating demand driven by regulation.
Liability frameworks are maturing. Insurers have figured out how to underwrite the risks of warehouse robots, surgical systems, and connected safety equipment, removing a barrier that previously kept these technologies out of risk-sensitive industries.

The market reflects this momentum. We estimate the core physical AI market will reach approximately $450 billion by 2030, with high-growth segments expanding at 15% to 25%+ CAGR. When you include the broader ecosystem—mobility, connected infrastructure, and the technology layers that support them—the market approaches over $1 trillion. For investors evaluating the category, the focus may shift from market sizing to underwriting discipline, segment selection, and risk assessment.

Where market activity is concentrated

We’ve observed six key verticals where physical AI-related deal activity and commercial adoption are developing. For buyers, these verticals are not equally actionable. Among the differences that should be considered: stage of maturity, type(s) of capital needed, and risk-return profiles.

Industrial and smart infrastructure

Industrial and smart infrastructure are the most active segments. Factories, warehouses, ports, and logistics networks may be more conducive to adoption because tasks are repeatable, and conditions can be more controlled. As a result, the return on investment (ROI) may be more measurable in some use cases through labor, throughput and downtime metrics. Private investors are backing companies with a track record of rolling out warehouse automation at scale, while strategic capital continues to flow into robotics and automation capabilities, including ABB's ongoing expansion in industrial automation.

Autonomous mobile robots (AMRs), robotic picking, and AI-driven quality control continue to attract both financial and strategic buyers, with Robotics-as-a-Service (RaaS) models accelerating adoption by converting Capex to Opex.

Underpinning these systems is the sensor and hardware components, or actuator, layer. Consolidation in the Light Detection and Ranging (LiDAR) sector, as seen in recent mergers, highlights the importance of scale and economics in sensing technologies, while companies in machine vision and industrial sensing continue to hold key positions that are hard to compete away.

In connected safety and infrastructure, deal activity appears focused on companies that combine hardware, software, and recurring revenue.

Health and medical

Health and medical, while the smallest vertical by market size, is attracting investment dollars through business models where revenue is tied to each procedure, creating high switching costs, defensible intellectual property (IP), and recurring revenue. We also see investments flowing into rehabilitation robotics and hospital logistics automation, where labor shortages and aging populations are driving steady demand even as regulators struggle to keep up.

Entertainment

Entertainment is drawing interested investors into interactive animatronics, autonomous camera systems, and physical experiences powered by software with high margins and sticky customer relationships. Recent activity includes continued investment in experiential technology platforms specializing in interactive exhibits, animatronic environments, and sensor-driven installations, as well as established leaders in broadcast infrastructure, mobile production units, and remote camera systems. Capital is also flowing into immersive production and real-time rendering platforms that integrate LED volume stages, projection mapping, and game engine software with physical environments to power live events and location-based experiences.

While still an emerging segment, these investments reflect growing conviction that entertainment is shifting toward systems that blend physical hardware with software and where it can generate recurring revenue.

Humanoids and service robots

Humanoids and service robots are absorbing some of the largest venture capital (VC) rounds in the category today. Several companies in this sector have crossed the $1 billion valuation mark, and new funding continues to flow into companies building robots that operate in human-designed environments. The market is expected to reach $80 billion by 2030—reshaping work, operations and growth. Early deployments are already underway in logistics and hospitality, though training data and the ability to handle objects with humanlike precision remain key constraints holding back broader adoption.

Autonomous driving

Autonomous driving continues to attract the largest share of VC capital in physical AI, with recent funding rounds exceeding billions of dollars backing advances in neural perception systems, sensor fusion, and end-to-end learning architectures that enable vehicles to see and understand their surroundings in real time. Despite this, regulatory complexity, liability frameworks, and deployment risk across diverse geographies continue to keep most platforms in the venture phase, with commercialization timelines remaining a key area of uncertainty for investors.

Aerospace and defense

Aerospace and defense are seeing active VC investment into autonomous drones, unmanned vehicles, and multi-agent systems, with recent funding rounds reaching into the billions across leading platforms. Investment is targeting advances in onboard perception and navigation, real-time edge inference, secure communications, and coordinated multivehicle autonomy. Technologies with both military and civilian applications are attracting particular interest, with significant rounds backed by both venture and strategic investors reflecting the growing overlap between defense and commercial use cases.

Foundational enablers

VC dollars are also flowing into foundational technology layers that cut across verticals. World models are emerging as a critical frontier, with leading research initiatives building AI systems that can reason about physics, space, and causality––capabilities that current language models lack. Additional capital is flowing into next-generation AI architectures designed to process fast-changing, real-world data directly on the device without relying on cloud infrastructure, enabling faster and more reliable decision-making at the edge.

Segments to monitor for maturity signals

Some segments may remain earlier-stage, with adoption and diligence proof points still developing. Medical and surgical robotics, autonomous last-mile delivery, and AI-enabled building infrastructure may reach the maturity investors are looking for in the next couple of years.

In the meantime, much of the technology being built for humanoid robots—sensors, AI models, motor control systems—will eventually show up in simpler, more focused industrial machines.

The investors who start building relationships in these spaces now, whether through direct engagement, co-investment alongside VC firms, or tracking key proof points, will likely be first in line when these categories are ready.

Value creation levers

Traditional value creation levers still apply, but in physical AI they’re more tightly coupled with engineering, supply chains, and product design.

Scaling depends on whether deployments can be standardized and performance stays consistent across environments. Companies that reduce implementation time and complexity scale faster.
Supply chain resilience is a central value driver. Semiconductor and memory shortages have disrupted production. Having backup suppliers, redesigning products to reduce reliance on hard-to-source components, and regionalizing manufacturing are all important levers.
Design for manufacturability has a direct impact on margins. The choices a company makes early in product design—how many parts, which materials, how it's assembled—lock in the cost structure and ability to scale.
R&D allocation should be balanced across hardware, software, and AI. Overinvesting in one area creates bottlenecks in another.
Ecosystem strategy matters. Some build everything in-house for tighter control; others open their platform to attract more users. The right choice depends on use case and market dynamics.
Recurring revenue transition improves predictability and valuation. Software, analytics, and data services tightly integrated with the physical system are key enablers.
Serviceability and field reliability directly impact unit economics at scale. Companies that can update and fix their devices remotely without sending a technician on-site have a significant cost advantage. RMA (Return Merchandise Authorization) rates serve as a leading indicator of product quality, with improving rates of signal maturity, while flat or rising rates point to design or supply chain problems.

Risk considerations

Physical AI carries specific risks that differ from traditional industrial assets and pure software.

Deployment timelines stretch hold periods. Physical AI rollouts involve installation, integration, calibration, and training. Enterprise deployments can take six to 18 months per site.
Hardware recall risk has no software equivalent. A defective sensor or actuator requires physical retrieval, replacement, and potential liability exposure. A single component failure in a safety-critical application can halt deployments across customers.
Customer concentration is common. Many growth-phase physical AI companies derive 30% to 50% of revenue from fewer than five customers. Diligence should test how diversified the sales pipeline really is and what contractual protections are in place.
The talent bottleneck is real. Engineers who span mechanical design, embedded systems, and machine learning are exceptionally rare. A company's ability to recruit and retain this talent is often a good predictor of execution speed.
Regulatory uncertainty persists. Autonomous vehicles, surgical robots, and drone systems operate under evolving frameworks. A single adverse ruling can freeze a category for months.

What to look for in diligence

Traditional diligence approaches are often insufficient for physical AI. Investors should evaluate the technical and operational foundations that determine scalability, defensibility, and whether margins can hold up over time.

Here are some concrete actions to help you move from product claims to operating evidence.

Test product defensibility. Use hardware teardowns to understand architectural differentiation, AI replicability testing to assess how easily core functionality can be reproduced, and customer validation to pinpoint what actually drives differentiation.
Assess manufacturing readiness. Look for evidence that the product is designed for manufacturability, cost discipline, and repeatable production, with processes in place to manage quality, yield, and continuous improvement.
Analyze bill of materials (BOM) resilience. Examine component-level cost structure, supplier dependencies, redesign opportunities, and exposure to tariffs or supply chain volatility.
Evaluate system integration. Assess whether the architecture supports modularity, upgradability, data capture, feedback loops, and the right balance between edge and cloud processing.
Pressure-test serviceability and field reliability. Review over-the-air update capability for firmware, AI models and security patches, along with RMA rates by product line, trend data, and the economics of modular repair versus full-unit replacement.

Traditional commercial and financial diligence miss the risks that determine whether a platform can scale. Defensibility claims based on "proprietary AI" often don’t survive structured replicability testing. Bill of materials costs presented at prototype volumes frequently mask margin compression at scale. Consider an integrated approach that combined device teardown, full-stack assessment, and manufacturing analysis that can help you surface these issues before they become post-close surprises.

Timing considerations

Now (2026): Industrial automation, logistics, and connected safety platforms appear to show more developed commercial and diligence signals than some other segments. The business models work, the deployments are repeatable, and the field is consolidating. Some investors may evaluate platform or consolidation strategies, subject to thesis, diligence findings, and portfolio objectives. Though investors should be careful about trade policies, export controls, CHIPS Act incentives, and the broader US-China decoupling mean that where a company sources its chips, sensors, and components is now a deal-level issue. Companies that have sorted out their supply chains are worth more, and those that haven’t are seeing impacts to valuation.

Near-term (2027-2028): Medical and surgical robotics, aerospace defense dual-use platforms, autonomous last-mile delivery, and entertainment automation are approaching PE (private equity)-grade maturity. Investors may choose to monitor proof points and develop market relationships.

Longer horizon (2029 and beyond): Humanoid robotics, autonomous driving at full scale, and general-purpose service robots remain in the venture phase. These segments may remain earlier-stage and could require different risk tolerance, time horizons, and diligence criteria.

Returns may depend on the timing of entry, quality of diligence, execution capability, and market conditions.