After the funding cooldown, how can robotics companies get past the payment collection hurdle?

After robot companies secure orders, they must first spend money to deliver the equipment to customers.

A humanoid robot procurement announced by Tongji University in September closed at 198,000 yuan, requiring delivery within 15 days of the contract taking effect, with full payment after delivery and acceptance. Shanghai Kuangji Electronic Technology Co., Ltd., which took the order, must first complete the supply before receiving the payment corresponding to this revenue.

Robot companies that are expanding deliveries must simultaneously arrange more components, assembly, and on-site personnel. Customer money has not yet arrived, but procurement and wages are already due; R&D on next-generation products is also waiting for funding.

In September 2026, China's embodied intelligence primary market publicly recorded 46 financing deals, with a reference amount of approximately 11.624 billion yuan, down 40.1% from August; excluding the largest financing deal in each month, the amount for the remaining projects still fell 32.4%. When financing shows a monthly decline, companies that have already scheduled deliveries must recalculate how many projects their available funds can support before the next round of money arrives.

Order growth brings an outlay first

UBTech's revenue in the first half of 2026 grew 104.2% year on year to 1.269 billion yuan, of which revenue from the full-size embodied intelligence humanoid robot business was 590 million yuan. Over the same period, the group's operating cash inflow was 1.191 billion yuan and outflow was 1.940 billion yuan, a net outflow of 749 million yuan, roughly twice that of the same period last year.

Procurement is the largest cash expenditure item. Cash paid for goods and services was 1.092 billion yuan, compared with 525 million yuan in the same period last year; cash payments related to employees were 543 million yuan, compared with 395 million yuan in the same period last year. Both procurement and personnel expenses are increasing, with the former growing faster.

UBTech's operating cash flow and expenditure composition

Over the same period, an increase in the group's inventory corresponded to a cash flow deduction of 268 million yuan, and an increase in operating receivables corresponded to 408 million yuan. Of the funds already spent, some are tied up in stockpiling and production, and some are waiting for customer settlement. Revenue growth, profit, and cash collection follow different timetables.

This directly limits the number of projects that can be started simultaneously. The previous batch of equipment has not yet been paid for, while the next batch of components already requires payment; even if the factory has capacity and sales has orders, the company may still delay procurement because it lacks sufficient bridge funding. Delayed procurement in turn delays assembly, debugging, and acceptance, and the next payment arrives later. Financing could originally fill this gap, but once financing is delayed, several projects begin competing for the same pool of money.

The difference between companies lies in where this money comes from. UBTech had cash and cash equivalents of 2.326 billion yuan at the end of June, and its existing businesses continue to generate operating cash inflows. For a company with a certain revenue scale, collection speed first affects how many times funds can be turned over and when the next batch of projects can start. Cash reserves can provide a buffer, but the more projects there are, the more important the funding arrangements for each delivery batch become.

Agility, by contrast, is still in an early commercialization stage. In the first half of 2026, its net sales were approximately US$174,400, of which about US$171,700 came from deployment and professional services; over the same period, operating activities used approximately US$86.2 million in cash, and cash and cash equivalents at the end of June were approximately US$14.65 million. The funds were mainly used to advance commercialization, prepare for manufacturing, and develop next-generation robots. The primary problem it faces is that the scale of customer business is not yet sufficient to cover the entire company's investment.

UBTech and Agility operating data

Agility's proposed merger and related financing announced on October 30 are expected to raise approximately US$465 million, of which US$100 million in equity funding has already been received. This enables it to continue investing in new products and commercial deployment while existing revenue is still small. For such companies, collecting payments from a few projects can ease the delivery burden; funding to support R&D and expansion still mainly needs to be solved through financing.

Therefore, differences in scale ultimately come down to operating structure: companies with sustained revenue must prevent expansion from locking cash up in inventory and receivables; companies with still-small revenue must ensure that R&D investment continues until products mature. Both need collections, but collections play different roles in their respective funding arrangements.

Acceptance conditions determine how much money a single deal ties up

The Tongji University procurement placed payment after acceptance. The 15 days are the delivery deadline, and payment collection still requires acceptance; the settlement arrangements between the manufacturer and the supplier determine who bears the cost of goods and stockpiling investment first. The contract amount is written clearly, but how long bridge funding must last also depends on whether delivery can be completed as agreed.

For orders of the same amount, different prepayment ratios, phased acceptance, and final payment arrangements will create different funding pressure for the company. A large customized project with concentrated upfront investment, a broad acceptance scope, and back-loaded payment will tie up funds and engineers for a long time; a smaller order that can be delivered and settled in batches may instead free up resources faster and allow the next batch of production to follow.

For companies short of funds, an order's gross margin must also be viewed together with the funds tied up: to earn this money, how much must be advanced first, how long until it can be recovered, and whether another deal that could have been delivered on schedule is squeezed out during that period. An order with a higher book gross margin may also cause the company to face payment pressure on procurement and wages sooner.

Payment terms therefore affect product plans. If capabilities still under development are promised as part of overall acceptance, a feature that remains unfinished may hold up settlement for the entire batch of equipment. Only by accepting already deliverable features in batches first and separately defining the scope and cost of new requirements can completed work enter settlement in a timely manner.

Overall acceptance and phased acceptance

Agility's Digit 5 order provides a more concrete example: as of May 2026, a customer's roughly $300 million multi-year order corresponds to 1,000 robots under a three-year RaaS service arrangement, with fulfillment contingent on meeting contract milestones, product functions, and specifications. The multi-year demand the customer signs must first become a deliverable product, and then services and revenue are provided and earned period by period. When robotics companies arrange manufacturing capacity for orders, they must also allocate R&D resources for these preconditions.

For business operations, committing to a feature that is not yet mature costs more than development expenses. Once it becomes a condition for acceptance of the entire order, the related equipment and on-site teams may end up waiting together. The scope set when sales pursues an order is passed along the contract into R&D schedules and funding plans.

The time engineers spend on site is also a cost

A robot places a bin onto a conveyor belt, and the camera captures only a few seconds; integrating this action into a factory process involves incoming material posture, grasp positions, conveyor line takt, and safety systems. If a certain step requires repeated manual remediation, the entire workstation must continue debugging, and costs continue to accrue even after the equipment is delivered to the site.

The Schaeffler Taicang Embodied Intelligent Robot Digital-Intelligent Factory, which opened on October 20, has five workstations and six humanoid robots, covering bin handling, parts sorting, press-fitting and assembly, inspection, and packaging, along with training, data collection, and digital twin capabilities. What delivery faces is the entire process, and the robot itself is only one part of it.

Robot work cell deployment diagram

If every customer requires redeveloping interfaces, rebuilding task flows, and long-term on-site staffing, then as revenue increases, project teams must expand in tandem. Hardware can be produced in advance, but field engineers can hardly move fully to the next order before the previous one is wrapped up. The ceiling on delivery is limited by both funding and manpower.

This is also where R&D is easily led by delivery. Fixing a fault that blocks acceptance is often more urgent than exploring next-generation capabilities, because the former directly determines whether a sum of money already invested can be recovered. Core R&D personnel continuously staying on site to wrap up work crowds out general product iteration; the next order still lacks mature tools and must continue to rely on the same group of people.

Whether on-site investment can be converted into tools usable for subsequent projects determines whether the company becomes heavier and heavier as it grows. Adapting grasping solutions to similar bins, deploying software to reduce changeover configuration, and syncing fault fixes to already delivered equipment all reduce repetitive work. In contrast, interfaces and processes that serve only a single customer often still require continuous maintenance following customer needs. Both types of work can generate contract revenue, but their help for subsequent delivery differs.

GXO and Agility's collaboration has already reached cross-site expansion. The two companies piloted in Flowery Branch, Georgia, at the end of 2023 and signed a multi-year RaaS agreement in June 2024; as of October 2026, Digit has cumulatively handled more than 100,000 bins at that site. On October 26, the two parties announced plans to begin deployment in Statesville, North Carolina, and Lebanon, Tennessee, starting in early 2027, first using Digit 4 and migrating after Digit 5 is commercialized.

GXO and Agility deployment progress

The value of this collaboration lies in the same customer continuing to purchase similar handling tasks. New sites still require adaptation, but the robotics company now has task experience, software, and service processes it can carry over, as well as a clearer deployment plan. For the business side, what matters is whether new sites can use fewer debugging personnel, allowing the original team to handle more deliveries.

Service fees stretch out the funding arrangement

Cross-site deployment also brings another business choice: who holds the equipment, and how costs are recovered.

Agility offers two models: RaaS and equipment purchase. Under RaaS, Agility retains ownership of the robots, the customer pays deployment fees and ongoing service fees, and the company first invests in hardware and then recovers funds through subsequent charges, while also bearing maintenance and equipment wear. Under the equipment purchase model, the customer acquires the robots, pays the hardware price upfront, and then there is additional software and maintenance service revenue.

Funding arrangements for equipment purchase and RaaS

Service fees reduce the customer's one-time burden of buying equipment, but leave the supplier with a longer capital recovery process. The more newly deployed robots there are, the more equipment must be invested upfront, while service revenue comes in gradually. Even if the customer pays each installment on time, the supplier still has to address funding needs during the expansion phase. Here, there is no overdue collection, yet money is still short.

The charging method is therefore tied to the expansion speed a company can withstand. Companies with thicker cash reserves and the ability to arrange long-term funding are better positioned to wait for service revenue to be recovered installment by installment; companies with smaller funding buffers are more in need of customer prepayments, equipment sales proceeds, or phased deployment to ease upfront investment. Service contracts can build long-term customer relationships, but the company must also bear the equipment and O&M expenses within that relationship.

Customer prepayments also have usage constraints. The company must first set aside enough for manufacturing, deployment, and after-sales costs for this order before it knows how much funding remains for other projects. If it continuously uses the next order's prepayment to fill the previous order's gap, then once new orders slow down, the delivery expenses of old projects will pile up together. Problems that are not easily exposed when contracts continue to grow may emerge concentratedly when orders slow.

After the funding cooldown, robot companies' operating choices will become more concrete: whether they can turn down a custom order that ties up too much capital, whether they can let R&D staff return to product iteration as planned, and whether they can have equipment and service funds ready before adding new sites. These choices need to be supported by both delivery and collection.

Once a batch of orders passes acceptance, engineers can move on to the next deployment, the recovered funds can support the next round of procurement, and on-site experience can be used for the next delivery—only then does growth have the conditions to continue. The initiative that comes with collecting payments is reflected precisely here.

This article is from the WeChat public account "Feng Wei Feng Dabai," author: Feng Dabai, published by 36Kr with authorization.