Universal Robots and MiR: what Teradyne's filings and customers confirm

Teradyne, the US test-equipment maker that owns Universal Robots (UR) and Mobile Industrial Robots (MiR), reports cobots and mobile robots as a separate Robotics segment, which most mobile-robot makers do not. Its filings prove revenue. They do not prove how many robots are working. Its 2025 annual report says Teradyne Robotics "has sold over 110,000 cobots" and "over 11,000 AMRs", but it sells "principally through distributors and OEMs," so "sold" means shipped into a sales channel, not installed at a customer. The count is also updated unevenly: the previous annual report, filed February 20, 2025, still said "over 75,000" cobots, four days before UR announced it had passed 100,000.

The segment's growth now rests on one customer Teradyne does not name. It describes a "large e-commerce customer" whose revenue it tripled from 2025 to 2026, according to two published transcripts of the October 28, 2026 call. Amazon's own photos of its Vulcan picking robot show a UR-branded arm, and investigative reporters identified UR as the arm supplier in July 2025. Amazon has not named the supplier, and its own published Vulcan counts are a six-robot pilot and a planned 30-robot beta at one Spokane, Washington, building.

Named customers do confirm working fleets, mostly on vendor case pages: 43 MiR robots completing 1,000 missions a day at Stellantis's Caen transmission plant, 37 MiR robots at Sengkang General Hospital in Singapore, and 11 UR arms on a FIAT 500 line in Turin. Flex says in its own release and investor material that it runs UR and MiR robots in its factories, without a count. No customer publishes a payback period. The MiR Sengkang page says it is "too early for a full ROI report," which is the vendor's wording, not a quote from hospital staff.

Analysis based on Teradyne's 10-K and 10-Q filings for 2020 through the third quarter of 2026, its earnings releases and the prepared remarks it posts for each call, two independent transcripts of the October 2026 call, Amazon's own Vulcan pages and Amazon Science post, Flex's release and November 2026 investor presentation filed with the SEC, Rockwell Automation's 2025 10-K, MiR and UR case pages, and trade and investigative reports, all retrieved November 12, 2026. AI assistance was used in research and drafting, and the piece was checked by a separate AI fact-check and AI editorial review. Servospace did not visit any site, interview the companies named, or see operating data. Calculations are Servospace arithmetic on published figures and are labeled as such.

In this analysis

• The short answer
• What kind of evidence each number is
• What the segment line proves
• Why the sold counts cannot be subtracted
• The unnamed large customer
• Named customers, site by site
• Per robot and per site
• Marketing claims versus measured results
• Counterevidence and limits
• What a buyer should ask for
• What would change this

The short answer

• Units: Teradyne gives cumulative "sold" floors: over 110,000 cobots and over 11,000 AMRs in its February 2026 annual report. These are channel shipments since 2008 for UR, not robots running today. Teradyne gives no installed base, no active-robot count and, since the fourth quarter of 2025, no split of revenue between UR and MiR.
• Money: Robotics revenue was $95.8 million in the third quarter of 2026, up 27.6% from a year earlier, with pre-tax income of $0.3 million, according to the 10-Q. For the first nine months it lost $3.2 million before tax on $287.0 million of revenue. In 2025 it lost $99.4 million on $308.3 million.
• MiR is the smaller part. When Teradyne still split the numbers, MiR was $72 million of $365 million in 2024 revenue and $12 million to $20 million a quarter in 2025.
• Biggest customer: unnamed in every Teradyne document. The Amazon link is an inference from photos and reporting, not a Teradyne or Amazon statement.
• Named sites: Stellantis Caen (43 MiR robots), Sengkang General Hospital (37 MiR robots), Stellantis Mirafiori (11 UR cobots) and Flex (count not given). Only Caen gives a throughput figure, and none gives a measured payback.

What kind of evidence each number is

What the segment line proves

Teradyne's Robotics segment "includes operations related to the design, manufacturing and marketing of collaborative robotic arms and autonomous mobile robots," the 10-Q says. Revenue is the one number that comes with an audit trail.

Three things stand out. First, the third quarter of 2026 was the "highest third quarter revenue in the segment's history," as Teradyne puts it, but it was below the second quarter's $99.9 million, which still lost money. Second, the profit was thin: $0.3 million on $95.8 million is about 0.3% of revenue (Servospace arithmetic). Third, the growth is not even by region. The 10-Q's revenue note shows third-quarter Robotics sales in the Americas rose to $39.0 million from $28.5 million and Asia Pacific to $25.3 million from $13.9 million, while Europe, the Middle East and Africa fell to $31.6 million from $32.8 million.

What drove it, in Teradyne's words: "Electronics and Semiconductor is now our largest vertical for Robotics," its third-quarter prepared remarks say, as contract manufacturers turn to robots for data-center tray assembly and test. "AI-related" revenue was about 20% of Robotics sales in the first nine months of 2026. In the second quarter, Teradyne said US sales were 32% of Robotics sales, per its Q2 2026 remarks.

For comparison, Rockwell Automation, which bought Clearpath Robotics and its OTTO Motors division in October 2023, lists autonomous mobile robots inside the Motion Control business of its Intelligent Devices segment and does not report their revenue separately in its fiscal 2025 10-K. Teradyne's segment line, thin as it is, is more than a buyer gets from most mobile-robot makers.

Why the sold counts cannot be subtracted

The tempting move is to subtract one year's "sold" count from the next and call the difference annual shipments. The filings do not support it.

On February 24, 2025, four days after the FY2024 report repeated "over 75,000," Universal Robots announced that it had sold 100,000 cobots. So the FY2024 figure was a stale floor, and differencing the two latest reports would show 35,000 cobots in a year in which Robotics revenue fell 15.5%. The real gap between UR's 100,000 announcement and the 110,000 in the next annual report is about a year, but both are "over" figures, so even that 10,000 is not a measured shipment number.

Three more cautions. The subject of the sentence changed from "Universal Robots" to "Teradyne Robotics" in the latest report, and MiR absorbed AutoGuide in September 2022, so the 11,000 AMRs may include AutoGuide units; the filing does not say. "Sold" through distributors can include robots sitting in distributor stock or integrator labs. And none of the reports gives a retirement or replacement rate, so the cumulative count says nothing about how many 2010s cobots still run.

The unnamed large customer

Teradyne has referred to one large customer in most quarters since mid-2025, without naming it. The sequence, from its own posted remarks and the October 2026 call:

The fourth-quarter 2025 figure gives the only dollar floor: more than 5% of about $89.4 million is more than about $4.5 million in one quarter (Servospace arithmetic). Teradyne gives no 2026 dollar figure, so "tripled" cannot be turned into dollars or robots. On the same call, Smith said visibility in Robotics, "aside from our largest customer, is pretty short term," with a large portion of revenue coming "from turns," industry shorthand for orders booked and shipped within the same quarter. Both transcripts agree on those phrases and on the customer's goal: to "quickly flip over a distribution center from manual to automated."

Is it Amazon? Teradyne has not said so, and neither has Amazon. The case for it:

• Robotics president Jean-Pierre Hathout told The Robot Report in December 2025, "we've landed a major deal with a major e-commerce player," in explaining a new US hub in Wixom, Michigan, meant to be closer to that customer. The publication said it believed the customer was Amazon.
• Hunterbrook Media, with Citrini Research, reported in July 2025 that Amazon's Vulcan arms are made by Universal Robots, pointing to UR labels and logos in Amazon's own images and video. It said neither company replied. Hunterbrook discloses that an affiliated investment firm may take positions based on its reporting.
• Servospace looked at the hero image on Amazon's Vulcan page: the arm's joint cap carries the UR logo.

That makes UR the likely arm supplier for Vulcan. It does not prove that Vulcan is the "large e-commerce customer" business in Teradyne's remarks, and it says nothing about volumes. Amazon's own numbers are small and dated. Its Amazon Science post of May 9, 2025 said the Vulcan pilot "involved six Vulcan Stow robots" in Spokane and that a beta trial "will involve another 30 robots in the same facility," followed by a larger deployment in Germany. The same post says both Vulcan robot types have a second arm with a hook for the pod's elastic bands, so robot counts and arm counts are not the same. Amazon's June 2026 Europe post says Vulcan is "expanding to more European sites" and handles more complex picking in Hamburg, with no count. What Amazon says Vulcan can do, about 75% of item types at speeds comparable to staff, is covered in our piece-picking evidence guide.

An EDGAR full-text search for "Universal Robots" in filings from June 2025 to November 12, 2026 returned Teradyne and four other filers, none of them Amazon or a named customer in this analysis.

Named customers, site by site

Flex is the strongest customer voice because the statement sits in its own filing. A slide in Flex's November 10, 2026 investor presentation, furnished to the SEC on Flex's 8-K, says Flex makes key components and deploys UR and MiR robots in its own factories (the slide text is laid out in columns, so it is paraphrased here). Flex is also a Teradyne supplier, and the slide calls it "Customer, supplier and innovator." Figures on that slide, a "4x volume ramp in 6 months" and "100% on-time delivery," are not tied to any robot fleet and are not robot performance figures. None of the named sites says what the robots cost, and none was confirmed in a customer-issued release with a count.

Per robot and per site

Only one site gives both a count and a daily figure.

About 23 missions a robot a day is an average across three different flows and two robot sizes; a mission between the paint shop and the lines is not the same job as a box run to a workstation. The figure is useful mainly as a scale check: it describes steady internal transport, not high-frequency picking.

Marketing claims versus measured results

The payback ranges appear in the business description of a filing, but they read as marketing. No customer in this analysis publishes a payback figure to set against them.

Counterevidence and limits

• Losses and restructuring. Robotics lost $99.4 million before tax in 2025 and restructured around 400 employees that year, recording $24.3 million of severance, per the 2025 10-K. It also took a $6.0 million inventory provision on Robotics products.
• A missed break-even target. In February 2025, Teradyne said restructuring would cut Robotics "break-even revenue from $440M in 2024 to $365M in 2025." Robotics revenue in 2025 was $308.3 million. Of the quarters shown in this analysis, the third quarter of 2026 is the only one with a pre-tax profit.
• Goodwill risk. Robotics carried $416.4 million of goodwill at the end of 2025, which the auditor flagged as a critical audit matter, and Teradyne found a triggering event for an interim test in mid-2025 without recording an impairment. The 10-K warns that changed assumptions "could result in" impairment in a future period.
• Concentration and visibility. Growth now depends on one large customer that Teradyne will not name, while the rest of the business is short-cycle orders.
• Competition. The 10-K lists traditional robot makers (KUKA, ABB, FANUC, Staubli, Yaskawa), other cobot makers (Techman, Doosan, Jaka, AUBO) and AMR rivals including Omron, Rockwell, HikRobot and KION.
• Old or undated case pages. The Mirafiori page is undated. Case pages are written by the vendor and do not say whether the robots are still running.

What this analysis does not cover: the MiR1200 pallet jack and pallet-moving fleets, which are in our autonomous forklift evidence guide, and order-picking robots such as Locus, in the collaborative picking guide.

What a buyer should ask for

What would change this

The next checkpoints are Teradyne's fourth-quarter 2026 results, which it has not yet dated (the fourth-quarter 2025 call was February 3, 2026), and its 2026 annual report, which should update the sold counts; any Amazon statement naming its Vulcan arm supplier or giving a fleet count; and Flex's next annual report. This guide will be reviewed when one of those is published, or within 90 days.

How the sources were used

Teradyne's SEC filings are treated as the company's audited or reviewed figures, and its posted prepared remarks as management statements. Quotes from the October 2026 question-and-answer session are used only where the Motley Fool and GoodMoat transcripts agree; one transcript's partner names ("Intrinsic AI," "Se-Ro") differ from Teradyne's own posted remarks ("Scale AI," "Sereact"), so the remarks were used. Amazon and Flex pages and filings are treated as the customers' own statements. MiR and UR case pages are treated as vendor claims, with named customer staff noted. Investigative and trade reports are labeled as such.