Robotics investment in 2026: where capital and acquirers are concentrating

The International Federation of Robotics (IFR) reported in September 2026 that the global operational stock of industrial robots surged 9% to a record 5 million units in 2025. Annual installations jumped 11%, and factories worldwide installed more than 600,000 new units over the year. The IFR forecasts installations to rise by 9% to 655,000 units in 2026 and to reach 806,000 units in 2029. That is a market growing steadily, not a speculative one.

North America is where growth stands out against a mixed picture. The IFR says the United States surpassed Japan in 2025 to become the second largest market for industrial robots, with installations growing 12% to almost 38,500 units. Its separate US release puts the figure at 38,000 units, up 11%, and notes that the US stands at 307 industrial robots in use for every 10,000 employees in the manufacturing industry. The two releases differ slightly in rounding; the direction is the same.

The regional contrast is sharp. Japan installed 36,219 units in 2025, a 19% decrease, and South Korea installed 30,000 units, down 1%. In Europe, Italy fell 11% to about 7,800 units and France fell 8% to almost 4,500. For an acquirer with a North American mandate, the US market is both the largest source of domestic demand growth and the place where customers are most willing to buy automation to offset labour constraints.

The implication is that the case for North American robotics assets does not rest on a distant future. There is a base of installed machines, a growing flow of new ones, and customers who need to integrate, maintain and upgrade them. That installed base is where recurring revenue sits, and it is where an acquirer can underwrite cash flows with evidence, rather than rely on projections.

A caution on the data: the IFR counts industrial robots installed in factories, which excludes many of the logistics, inspection and service robots that attract venture funding. The two bodies of evidence therefore describe different parts of the market. Industrial installations are the base on which acquirers can build, because customers, prices and volumes are visible. Venture rounds describe the frontier. A coherent map keeps the two apart and avoids using one set of numbers to justify valuations in the other.

What to do: build your market map from the installed base outward. Identify the segments in which US installations are growing, the integrators and software providers serving them, and the service revenue attached to the machines already deployed. Use the IFR’s country data to test any target’s growth claims against what the market is actually doing.

Venture capital has followed the demand. According to Crunchbase, robotics startups raised $15 billion in the full year of 2025, a record. The figure surpasses the $14.1 billion raised in the peak venture funding year of 2021. In 2026, Crunchbase says startups have so far raised $18.8 billion, so the 2025 total has already been eclipsed.

The distribution is as important as the total. In September 2025, Crunchbase reported that Figure, the developer of general-purpose humanoid robots, raised $1 billion in Series C financing at a $39 billion post-money valuation. In March 2026, Saronic raised $1.75 billion in Series D funding, setting a valuation of $9.25 billion, more than double its Series C level in 2025. Another robotics-AI company raised $1.4 billion, tripling its valuation to over $14 billion.

A few large rounds can account for most of a record year, and that tells a buyer something. Capital is being allocated to platforms with the largest claimed addressable markets, not evenly across the value chain. The companies attracting rounds of $1 billion or more are not the typical acquisition targets for a mid-sized corporate. They are competitors for talent and compute, and they set the valuation expectations that smaller companies then quote back to acquirers.

The implication is that the headline funding figures overstate the number of investable opportunities and understate the price expectations. A corporate development team looking at a robotics company with a modest revenue base may find the founders benchmarking themselves against the largest rounds. Distinguishing a company with customers and cash flow from one whose valuation depends on a large-platform comparison is the first piece of diligence.

Concentration also affects who sets the price of talent. Companies that have raised large rounds can pay for scarce machine-learning and robotics engineers at levels a corporate acquirer will find hard to match. For a buyer, this raises the importance of retention terms in any deal, and of understanding where the key technical knowledge resides. An acquisition that loses its core engineers within a year has paid for assets it can no longer develop.

What to do: build two lists. The first covers the platform companies whose valuations are set by large rounds, which you should track but not expect to buy. The second covers integrators, software providers and component specialists with real customers, which is where acquisition is realistic. Price the second list against revenue, margin and backlog rather than against the headline rounds.

Two transactions illustrate what strategic buyers value. In October 2025, ABB announced the sale of its Robotics division to SoftBank for an enterprise value of $5.375 billion. ABB Robotics had 2024 revenues of $2.3 billion, about 7 percent of ABB Group revenues, and an Operational EBITA margin of 12.1 percent. SoftBank said that its next frontier is Physical AI and expects the acquisition to close in mid-to-late 2026. On our arithmetic the enterprise value is about 2.3 times 2024 revenue.

The second is Symbotic’s purchase of Walmart’s robotics business. Symbotic’s own filing states that $200M is to be paid at closing, with Walmart paying Symbotic $520M for a development program, including $230M at closing. Walmart commits to purchase 400 systems if performance criteria are achieved, which Symbotic said represents over $5B of future backlog. The buyer paid for a customer commitment and a roadmap, and the seller funded the development.

The two deals have different structures but a common logic. SoftBank bought an established franchise with a global installed base, a profitable business and a brand. Symbotic bought capability and received a committed anchor customer. In both cases the price reflected cash flow or contracted volume, not a demonstration. Neither deal was a bet that a robot works in the lab; both were bets that it is already being deployed.

The implication for corporate development is that the strongest precedents for robotics M&A involve either carve-outs of established units or acquisitions tied to a customer commitment. Transactions where the seller is a large industrial group simplifying its portfolio are likely to continue, because ABB’s decision shows that robots can be worth more to a technology-focused owner than inside a diversified electrification group.

The ABB transaction also carries a lesson on process. The seller had earlier considered a spin-off of the robotics division and chose a sale, which indicates that industrial groups are comparing the value of separation routes carefully. A buyer approaching a division that a parent has reviewed in this way should expect a well-prepared separation package, and should be ready to answer questions about the standalone cost, shared services and transitional support that make such a transfer workable.

What to do: map the large industrial groups with robotics units that no longer fit their core, and approach them with a view on separation as well as price. For customer-anchored acquisitions, insist that the commitment is contractual and conditional on performance milestones you can monitor. A multi-year purchase commitment is only as valuable as the terms under which the customer can walk away.

Unit sales of robots depend on the manufacturing cycle, and the IFR’s 2025 data shows how uneven it is. Installations rose in the United States and India, where the IFR reports almost 10,500 units, up 15%, and fell in Japan, South Korea, Germany, Italy and France. Germany, by far the largest market in Europe, saw sales fall 8% to fewer than 25,000 units. Customers in several mature markets are not buying more machines.

This matters because hardware growth is easy to forecast badly. A company whose revenue is mostly robot sales will move with capital spending by automotive, electronics and general industry, and with currency and trade policy. The 2025 data shows that even a year of record global installations contained sharp declines in major markets. A target whose revenue depends on one customer industry or one region carries risk that a headline growth rate hides.

Services and software behave differently. Once a robot is installed, it needs programming, maintenance, upgrades and integration with the rest of the plant. The installed stock of 5 million units is a larger and steadier source of work than annual installations of more than 600,000. The companies that sell, install and maintain systems across several vendors are less exposed to any one manufacturer’s cycle and are more likely to have recurring revenue.

The implication is that an acquirer should ask, for each target, which part of revenue depends on selling new machines and which on running those already installed. Revenue from the installed base is more predictable and usually earns a higher multiple. Revenue from new installations is cyclical and should be valued with that in mind. A mix that looks balanced on paper may be cyclical in practice if the service contracts are short.

There is a competitive dimension too. Chinese suppliers have gained domestic share in the world’s largest robot market, and that tends to push down the price of standard hardware globally. A North American buyer that depends on selling standard arms into price-sensitive segments will meet that competition directly, whereas one that sells applications, software and service faces it only indirectly. This is a further reason to favour targets whose proposition is not a machine specification.

What to do: split target revenue into new installations, software subscriptions, services and spares, and test each against the customer contracts. Look at retention and renewal data for service and software revenue, and at the time to deploy and the margin on each installation. A business that shortens deployment time for its customers has a more defensible position than one that simply sells more units.

The IFR’s US data shows both strength and concentration. US installations rose 11% to 38,000 units in 2025, but the automotive industry remains the largest adopter, reaching 13,500 units, just 1% below the previous year. That means roughly a third of US installations went to one industry, which has long experience with automation and the in-house engineering to deploy it. Other industries are adopting more slowly.

The density figure points the same way. At 307 industrial robots in use for every 10,000 employees in manufacturing, the US is far from the saturation seen in the most automated economies. That gap is an opportunity for suppliers who can make automation workable for companies without large engineering teams. It is also a reason for caution: the barrier in those firms is not the price of a robot but the difficulty of applying it.

Longer-term estimates point to a larger prize. McKinsey research, as reported by AI Business in 2025, suggests the general-purpose robotics market could reach $370 billion by 2040, with 50% of that value coming from China, and that investment has topped $1 billion annually since 2022. This is a secondary report of a long-range estimate and should be treated as a scenario, not a forecast. It does, however, show that serious analysts expect a market beyond the factory floor.

The implication is that the opportunity lies in widening the set of customers who can adopt, not in selling more machines to those who already do. Suppliers who offer packaged applications, pre-trained skills, remote monitoring or robots-as-a-service lower the threshold for a small manufacturer or a warehouse operator. These models also convert one-off sales into recurring revenue, which is the property that acquirers and lenders value.

A practical test of any adoption claim is the time and cost to deploy a second and third installation of the same application. The first installation absorbs engineering effort and learning. If later installations are not markedly faster and cheaper, the company is selling projects rather than a product, and its margins will reflect that. Acquirers should ask for deployment histories by customer, with hours, costs and commissioning delays, and treat missing data as an answer.

What to do: identify the application categories where adoption outside automotive is growing, such as food, logistics and general manufacturing, and the companies that have made adoption easy for customers there. Assess how far the target’s proposition depends on the customer having engineers. The more it depends on them, the smaller the addressable market, whatever the technology.

Physical AI is the term for systems that combine machine learning models with machines that act in the real world. SoftBank has described it as its next frontier, and venture investors are pricing it accordingly. Figure’s $39 billion post-money valuation and another company’s tripling to over $14 billion in seven months show how quickly expectations have moved. The valuations assume that general-purpose capability will arrive and that customers will pay for it at scale.

Evidence of deployment at that scale is thinner than the capital. The deals and installation data we have cited concern established industrial robots and warehouse automation, where customers, backlog and margins can be verified. The humanoid and foundation-model rounds are priced on potential. That is not an argument against investing, but it implies a different risk profile and a different kind of diligence for each part of the market.

For an acquirer the central risk is paying for capability that cannot yet be deployed reliably. A model that works in a demonstration may fail in a plant with variable lighting, parts and people. The cost of closing that gap, through data collection, safety validation and field service, may be larger than the initial development and falls on the buyer after completion. Valuations that ignore this cost are likely to disappoint.

The implication is that valuation discipline matters more in this segment than in mature industrial automation. A corporate development team can reasonably take positions in physical AI through minority investments, partnerships and options, which limit exposure while the technology is proved. Full acquisitions are better reserved for companies that already generate revenue from real deployments and whose software can be shown to transfer across customers.

Safety, liability and regulation add to the diligence agenda. Machines that move in shared spaces carry product-liability exposure, and standards for general-purpose systems are still developing. A buyer should understand who bears responsibility when a system fails, what insurance is available and how the target’s certification position compares with comparable products. These are not reasons to avoid the sector, but they are costs that belong in the valuation rather than in a footnote.

What to do: structure physical AI exposure in stages. Begin with commercial pilots or minority stakes that include rights to deploy, then convert to control as milestones for reliability, safety and customer retention are met. Include technical diligence by people who have run robots in production, and value the target on the evidence of deployments, not on the size of the last funding round.

A robotics acquisition thesis should start with the customer’s job. Which task is being automated, how long does it take to deploy, who maintains the system, and what return does the customer achieve? Targets that can answer these questions with data are rare and valuable. Those that answer with a vision are not yet businesses. This difference explains much of the dispersion in outcomes across the sector, whatever the underlying technology.

We would screen on five attributes. The first is repeatability: how many installations of the same application, and how fast each is deployed. The second is software content and margin. The third is customer concentration and contract terms. The fourth is the share of revenue from the installed base. The fifth is the integration burden on the acquirer, including product, supply chain and people. A target that scores well on all five is likely to be worth the premium.

Deal structure should match risk. For targets with proven deployments, a conventional acquisition with a modest earn-out is reasonable. For those still proving technology, staged investment and milestone payments protect the buyer. For carve-outs from large industrial groups, such as ABB Robotics, the quality of the separation plan, including shared services, supply contracts and intellectual property, will determine whether the value moves intact to the new owner.

Geography matters as well. The US market is growing and has the density gap that suggests room for adoption, while parts of Europe and Asia are contracting in unit terms. An acquirer should ask whether a target’s growth depends on regions that are shrinking, and whether its products are adapted to the US customer base, including standards, service networks and financing models.

What to do: agree a short list of the five criteria with your board, score the long list against them and publish the scores internally. Combine the scores with a view on the strategic fit with your existing business. Start conversations early with the three or four companies that meet the criteria, because the best targets are rarely for sale and the right moment to approach them is before they raise their next round.

Karter McKinley supports corporate development teams with M&A and corporate development: building the market map, screening and approaching targets, and structuring staged investments and acquisitions alongside specialist technical, legal and tax advisers. This is information, not investment advice, and each engagement is scoped separately.