Do collaborative robots become a real growth line inside a test-equipment company?

Thesis

Chip test earns nearly all of Teradyne’s profit. In the quarter ended September 27, 2026, it produced $1,072.0 million of the company’s $1,306.0 million of revenue and $366.9 million of the $404.3 million of pre-tax income. The Robotics group, built around Universal Robots (collaborative arms, or cobots) and Mobile Industrial Robots (MiR, autonomous mobile robots), booked $95.8 million of revenue and $0.3 million of pre-tax income in the same quarter. That is the segment’s first profit in the periods we reviewed, and it follows pre-tax losses of $54.3 million in 2023, $77.6 million in 2024 and $99.4 million in 2025.

Robotics revenue is recovering: $287.0 million in the first nine months of 2026, up 31.1% from the prior-year period. But the segment’s weight in the company is shrinking even as it grows, because Semiconductor Test is expanding faster. Robotics was 10% of 2025 revenue and was 7.3% of nine-month 2026 revenue (our calculation from the segment table). At that proportion, a profitable Robotics line is a good outcome for capital allocation but does not drive the stock.

AI test demand drives the stock, and the market is debating how durable it is. Shares closed at $323.68 on October 27, $296.23 on October 29 and $274.37 on October 30 after the company reported a quarter that beat its own guidance but showed revenue down 1.7% from the second quarter (our calculation: $1,306 million against $1,329 million). Margins also slipped sequentially. Teradyne is a cyclical, concentrated, highly profitable test business with an option on a second business that has only just stopped losing money.

Company History

Teradyne was founded in 1960 and is headquartered in North Reading, Massachusetts, where it owns a 422,000 square foot campus. For most of its life it was a pure automated test equipment maker, selling testers that sort good chips from bad ones at wafer and package level.

The robotics push began in 2015. The company’s 10-K lists the sequence of deals: Universal Robots in 2015, Energid and MiR in 2018, Lemsys and AutoGuide in 2019, and most recently AET and Quantifi in 2025. Universal Robots introduced what the filing calls the first commercially viable cobot in 2008 and has sold over 110,000 cobots worldwide. Teradyne has since moved UR and MiR under one commercial organization; a January 2025 report notes that UR and MiR combined operations in 2024 and, in the first quarter of 2025, consolidated sales, marketing and services (The Robot Report).

The more recent corporate story is about test. In May 2024 Teradyne paid 483.1 million euros, or $524.1 million, for 10% of Technoprobe, an Italian probe card maker, and took a board seat. In March 2025 it created a Product Test segment by merging production board test, defense and aerospace, and wireless test, and it acquired Quantifi Photonics on May 31, 2025 to add photonic integrated circuit test. On April 8, 2026 it formed a joint venture with MultiLane in which it holds 75% for roughly $157.8 million, aimed at high-speed data connection test for AI data centers. On October 6, 2026 it agreed to make a $300.0 million minority investment in FormFactor, a probe and interface supplier, alongside a collaboration agreement.

Read together, the deal flow since 2024 points toward test, not robotics. Of the transactions above, only the 2015 to 2019 purchases went to Robotics.

Product

Semiconductor Test

The segment sells automated test systems and instruments used to test chips in production. The 10-K describes several families, including the ETS platform for analog and mixed-signal devices with SmartPin multi-site technology, power device testers for silicon carbide and gallium nitride, and an Integrated System Test group covering system level testers and storage testers. For AI work the relevant products are the testers used on compute processors and accelerators, where the customer is often a vertically integrated producer designing its own silicon, and memory testers used for high bandwidth memory and DRAM.

Management’s own description of 2025 is that the segment’s strategic shift toward AI-driven semiconductor testing meant AI-related demand drove the majority of revenue in the second half of 2025. In 2026 that broadened. In the third quarter, memory revenue hit a record $301 million, the fourth consecutive quarter above $200 million.

Product Test

This segment, created in 2025, holds board test and inspection, wireless test, photonic integrated circuit test, and defense and aerospace instruments. Third-quarter revenue was $138.1 million, up 56.4% from a year earlier on AI infrastructure demand for board, optical and high-speed interconnect test, plus the MultiLane joint venture. Pre-tax income was $41.4 million.

Robotics

Robotics sells two product lines. Universal Robots makes collaborative arms in a range of reaches and payloads, from the UR3e up to the UR30 and UR20, programmed through its PolyScope software. The filing cites an average payback of 12 to 18 months for customers, which is a company claim we could not independently test. MiR makes autonomous mobile robots for factory and warehouse material handling, including the MiR250, MiR600, MiR1350 and the MiR1200 Pallet Jack.

For scale, a January 2025 trade report put 2024 Robotics revenue at $365 million, of which Universal Robots was $293 million (down 3%) and MiR $72 million. Teradyne does not report the split in its filings. The company also owns a 200,000 square foot building in Odense, Denmark for Robotics operations and in 2025 began fitting out a 68,000 square foot U.S. site in the Detroit area.

Market

Customer

Test equipment is a concentrated-customer business. In 2025, the filing says, two customers specified more than 10% of consolidated revenue, at 12% and 10%, and one direct purchaser accounted for 19%, including revenue specified by others. The five largest direct customers were 44% of 2025 revenue, against 36% in 2024. Teradyne does not name them in these passages.

Geography has swung with the AI cycle. By customer site, Korea was 32% of third-quarter 2026 revenue against 12% a year earlier, Taiwan 14% against 37%, China 21%, Singapore 9% and the United States 8%. The shift from Taiwan toward Korea is consistent with memory becoming a larger contributor, though the filing does not attribute it that way. For the nine months, the United States was only 7% of revenue, which matters for tariff and export-control exposure.

Robotics customers are a different population: manufacturers, logistics operators and system integrators. Management says that in the third quarter electronics and semiconductor customers were the segment’s largest end-market, a notable overlap with the test business’s own customer base. The 2025 strategy was to pursue OEM partnerships, system integrators and large enterprise accounts in e-commerce, logistics, semiconductor and electronics.

Market Size

We did not find a primary-source figure for the total addressable market in semiconductor test that we could verify, so we size the opportunity from the company’s own results. Semiconductor Test revenue was $2,523.7 million in 2025 and $3,304.6 million in the first nine months of 2026 alone, a 101.5% increase on the comparable 2025 period.

For robots, the broader industry is growing. The International Federation of Robotics, as reported in September 2026, counted more than 600,000 new industrial robots installed in 2025, an 11% increase, with the global operating stock above 5 million units. We could not verify the cobot share of that total from a source we opened, so we do not state one. Teradyne’s cumulative cobot sales of 110,000 over roughly seventeen years, compared with annual industry installs above 600,000 robots of all types, shows how small the cobot niche is relative to the whole robot market.

Competition

Competitive Landscape

Teradyne names competitors segment by segment in its 10-K. In Semiconductor Test it lists Advantest, SPEA and Cohu. In Robotics it lists traditional arm makers KUKA, ABB, FANUC, Staubli and Yaskawa, emerging cobot makers Techman, Doosan, Jaka and AUBO, and mobile robot makers including Omron, Rockwell Automation, HikRobot and KION. In Product Test the list includes Keysight, Rohde & Schwarz and Anritsu. The filing adds that it also faces competition from emerging Asian companies and from internal supply at several customers.

The two main segments face opposite structures. Chip test is a small-number oligopoly with high switching costs, since a test program written for one platform is expensive to port. Cobots are a crowded field with Chinese and Korean entrants and a long tail of integrators. The first structure supports a gross margin near 59%. The second has produced a segment that needed a restructuring and a lower breakeven to reach a profit.

Competitors

Advantest

Advantest is the other scale supplier of system-on-chip and memory testers, and the competitor that matters most for the equity story. We did not source Advantest’s share or financials for this memo, so we make no share claim. Advantest was scheduled to announce fiscal second-quarter results on October 28, which one commentator called an important point of comparison for Teradyne. Any divergence between the two companies’ AI compute order trends would be the cleanest read on whether Teradyne is gaining or losing position.

Cobot and mobile robot rivals

Within Robotics, the named rivals fall into three groups. Incumbent industrial arm makers have the installed base and service networks. Collaborative specialists such as Techman, Doosan, Jaka and AUBO compete on price. Mobile robot specialists compete in warehouses where larger automation vendors already sell. Teradyne’s own January 2025 restructuring statement cited a rough patch for the industry and, in the group president’s words, an unusual pullback in low-mix, high-volume manufacturing. We read that as an admission that the core use case had softened, not that Universal Robots lost a specific fight.

Business Model

Teradyne sells capital equipment and attaches service. In the third quarter of 2026, products were 89% of revenue and services 11%, against 82% and 18% a year earlier. Services were $141.8 million in the quarter, little changed from $137.1 million a year earlier, so the mix shift reflects product growth rather than service loss.

The model is cyclical by construction. Customers buy testers when they ramp new chip generations or add capacity, and order timing is lumpy. The filing states that backlog is not necessarily indicative of future sales because customers may delay deliveries or cancel orders without advance notice. The cost base is partly variable. Third-quarter selling and administrative expense was 15.2% of revenue against 22.0% a year earlier, and engineering and development was 12.5% against 16.2%. That operating leverage is the reason profits rose far faster than sales, and it works in reverse if revenue falls.

The Robotics model sells hardware through distributors, integrators and OEM partners, with software and an accessory ecosystem (UR+) around it. Its economics depend on volume against a fixed engineering and overhead base. Teradyne said at the January 2025 restructuring that the cuts reduced the segment’s breakeven revenue from $440 million in 2024 to $365 million in 2025. That framing is useful. Actual 2025 revenue was $308.3 million, below the target, and the segment lost money, which is what the arithmetic implies.

Capital allocation is balanced across acquisitions, buybacks and a dividend. In the first nine months of 2026 the company spent $165.6 million on acquisitions, repurchased $173.5 million of stock and paid $61.0 million in dividends. Buybacks were far lower than the $518.7 million in the first nine months of 2025, as cash went toward growth and deals.

Financials

The latest reported quarter is the third quarter of 2026, ended September 27, 2026, released on October 27. Revenue was $1,306 million, up from $769 million a year earlier and down from $1,329 million in the second quarter. GAAP diluted earnings per share were $2.17 (non-GAAP $2.29), against $2.38 in the second quarter and $0.75 a year earlier. GAAP gross margin was 59.1%, against 59.8% in the second quarter, and non-GAAP operating margin was 31.6%, against 33.7%. For the fourth quarter the company guided revenue to $1,250 million to $1,350 million and non-GAAP EPS to $1.95 to $2.25.

For the nine months, revenue was $3,917.4 million against $2,106.7 million, operating income $1,312.8 million against $356.9 million, and diluted EPS $7.08 against $1.85. For fiscal 2025, revenue was $3,190.0 million, net income $554.0 million and diluted EPS $3.47, so 2026 is already running at more than twice the 2025 profit pool.

Semiconductor Test and AI compute

Semiconductor Test revenue crossed $1 billion for the third consecutive quarter. It grew 76.9% year on year in the third quarter on compute and memory for AI, but fell sequentially from $1,121.8 million in the second quarter (a 4.4% decline by our calculation). The segment’s pre-tax margin in the quarter was about 34% ($366.9 million on $1,072.0 million, our calculation). The 2025 pattern was similar in direction: segment revenue rose 18.8% and pre-tax income rose to $700.7 million, driven by AI compute and networking while memory held steady on share gains in high bandwidth memory.

Management is positioning for continued demand. CEO Greg Smith said demand extended across the AI value chain, from memory and hard disk drive test to high-speed interconnect and robotics. The cash and balance sheet signals are mixed. Inventory rose to $575.0 million at quarter end from $379.6 million at December 31, 2025, and receivables rose to $1,110.8 million from $786.9 million. That is consistent with preparing for shipments and also with a risk of excess stock if demand stalls.

Robotics: revenue, restructuring and margins

Robotics revenue fell from $375.2 million in 2023 to $364.8 million in 2024, citing continued weakness in industrial automation and softer Universal Robots sales, then to $308.3 million in 2025, a 15.5% decline on lower sales of both cobots and mobile robots. The 10-K notes the fourth quarter of 2025 was the third consecutive quarter of sequential growth. That recovery carried into 2026: six-month revenue was $191.2 million against $143.9 million, and the second quarter reached $99.9 million, up 33.4% year on year.

The third quarter did not extend the climb. Revenue of $95.8 million was up 27.6% year on year but about 4% below the second quarter (our calculation), so the segment has plateaued around $90 million to $100 million per quarter for the last three quarters. Annualized, that is near $380 million, only modestly above the $365 million breakeven level management cited in January 2025.

The restructuring accounts for most of the margin change. The 2025 10-K records $29.4 million of severance charges in total, of which $24.3 million related to a Robotics restructuring affecting about 400 employees. The January 2025 first round was reported as 10% of the group’s workforce. Robotics cash severance payments of $15.3 million were made in 2025, with the rest expected by the end of the third quarter of 2026. Company-wide restructuring and other charges fell to $2.0 million in the third quarter of 2026 from $6.6 million a year earlier.

The result: Robotics pre-tax income improved from a loss of $17.6 million in the third quarter of 2025 to a profit of $0.3 million. Over nine months the loss narrowed to $3.2 million from $73.3 million. The filing credits higher sales volume, improved product mix and lower operating expenses from restructuring. A profit of $0.3 million on $95.8 million of revenue is a margin of 0.3% (our calculation): breakeven, not yet a return on the capital put in. Across 2023 to 2025 the segment lost a combined $231.3 million pre-tax (sum of the three annual figures cited above).

The balance sheet carries a reminder. The goodwill assigned to the Robotics reporting unit was $416.4 million at December 31, 2025, and the auditor flagged the unit’s impairment test as a critical audit matter, since the fair value estimate depends on projected revenue growth rates and market multiples. The unit passed in 2025, but the next test, in the fourth quarter of 2026, will rely on exactly the revenue trajectory discussed above.

Cash and capital

Cash and equivalents were $460.7 million at September 27, 2026, up from $293.8 million at year-end, and short-term debt was zero after being $200.0 million at year-end. Operating cash flow was $313.6 million in the quarter and $1,047.9 million for nine months, against nine-month capital expenditure of $241.8 million. The $300 million FormFactor investment is expected to close in the fourth quarter of 2026, which will draw on this cash. The company paid a quarterly dividend of $0.13 per share.

Valuation

All market data is as of the October 30, 2026 close. TER closed at $274.37. With 156,124,751 shares outstanding as of October 26, 2026, the implied market capitalization is about $42.8 billion (our calculation from those two cited figures).

On trailing earnings, GAAP diluted EPS for the twelve months to September 27, 2026 was about $8.70, built from fiscal 2025 EPS of $3.47, plus nine-month 2026 EPS of $7.08, minus nine-month 2025 EPS of $1.85. That gives a trailing price-to-earnings ratio of roughly 31.5 times ($274.37 divided by $8.70). Trailing revenue, built the same way from $3,190.0 million, $3,917.4 million and $2,106.7 million, was about $5.0 billion, which puts the market capitalization at roughly 8.6 times trailing sales. These are trailing multiples on a period of unusually fast growth; they will compress mechanically if the fourth-quarter guide of $1,250 million to $1,350 million holds and flatten the year-on-year comparison.

For context on how far the valuation has moved: the 10-K cover reported a market value of shares held by non-affiliates of about $12.7 billion at June 29, 2025. That measure excludes affiliates and differs from total market capitalization. Even so, it indicates that the equity value has roughly tripled in sixteen months. Recent price action has been volatile: the stock closed at $345.49 on October 21, fell to $308.75 the next day, rose to $323.68 on the day of the earnings release, fell to $296.23 two sessions later and closed at $274.37 on October 30, down 7.4% on the day. Our reading of the post-earnings decline is that the market is pricing the sequential revenue dip, lower margins and a flat fourth-quarter guide as signs of a plateau, though we found no company statement that confirms the cause.

The segment mix matters for how to value this. Semiconductor Test generated $1,262.3 million of nine-month pre-tax income. Robotics generated a loss of $3.2 million. At today’s market value, the Robotics group is a rounding error in earnings terms, and its stand-alone worth rests on revenue multiples applied to roughly $380 million of annualized sales, the same method the company’s own impairment test uses.

Key Opportunities

Broadening AI test demand beyond compute

Management says demand is spreading beyond accelerators into memory, hard disk drive test, high-speed interconnect and robotics. The company has put capital behind this with the MultiLane joint venture and the FormFactor stake, whose collaboration agreement targets silicon photonics, high-performance compute, advanced packaging and high bandwidth memory. If these fill in as separate revenue streams, the test business is less dependent on one customer type. Product Test already grew 56.4% in the third quarter.

Robotics operating leverage from a lower cost base

With the restructuring complete and revenue near $96 million a quarter, each additional dollar of Robotics revenue should flow through at a higher margin than before. The segment moved from a $17.6 million loss to profit on $20.7 million of additional year-on-year revenue, which implies an incremental pre-tax margin near 87% (our calculation: $17.9 million improvement over $20.7 million revenue gain). That figure is flattered by cost cuts and will not persist. It does show the cost base was the main problem. Management has said it is committed to reinvesting in research and development in 2027, which would reduce that leverage and signals intent to grow rather than harvest.

Semiconductor and electronics as a Robotics end-market

The segment’s largest end-market in the third quarter was electronics and semiconductor customers. That is where Teradyne already has deep relationships, and where factory labor shortages and onshoring (the Detroit site is part of this) could support demand. The 2025 strategy specifically targeted semiconductor and electronics verticals. This is the one place where the two halves of the company can sell to the same buyers.

Key Risks

Concentration and cyclicality in AI test

In 2025 a single direct customer accounted for 19% of revenue, and the top five direct customers for 44%. The filing states that the loss of, or a significant decrease in demand from, a major customer could have a material adverse effect. Third-quarter revenue and Semiconductor Test revenue both slipped sequentially, and guidance implies a flat fourth quarter. A pause in AI capital spending would hit a cost base that has been scaled for growth.

Margin and working capital pressure

Gross margin fell from 59.8% to 59.1% sequentially, and non-GAAP operating margin from 33.7% to 31.6%, in part because selling and administrative expense rose to $199.0 million and engineering and development to $163.1 million. Inventory is up about 51% since December (our calculation), and the company recorded a $25.8 million inventory provision in 2025, including $6.0 million in Robotics. Further write-downs are possible if demand stalls.

Robotics may not clear its cost of capital

Revenue declined in two consecutive years and the segment’s market is structurally difficult. The IFR reported an 11% rise in all industrial robot installs, but we cannot show that cobots took share, and Teradyne’s own segment growth of 31.1% in nine months is a rebound from a depressed 2025 base, not proof of a new growth rate. A break-even segment carrying $416.4 million of goodwill is exposed to an impairment charge if forecasts are cut. An impairment would be non-cash, but it would also put the strategic rationale back on the table.

Acquisition and investment execution

The company is adding deals while the stock is volatile. The FormFactor stake of $300.0 million is minority and non-controlling, so it creates mark-to-market and strategic dependence without control. The MultiLane joint venture produced $131.6 million of goodwill and a redeemable noncontrolling interest held by the seller. The 10-K’s own risk factor warns that benefits from acquisitions may not be fully realized.

Geographic and trade exposure

Korea and Taiwan together were 46% of third-quarter revenue (32% plus 14%, our calculation), and China was 21%. The company states that trade-law compliance has limited its ability to compete in certain regions. It paid tariffs under the IEEPA program since 2025 and began receiving refunds after U.S. Customs started accepting claims on April 20, 2026, with no material effect so far. Export-control changes could affect China-linked demand in either segment.

What Would Prove Us Wrong

Our view is that Robotics is turning from a drag into a small, roughly breakeven business, and that the test franchise carries the investment case. These signposts would contradict that.

1. Robotics fails to hold profit. If the Robotics segment reports a pre-tax loss in the fourth quarter of 2026 on revenue of at least $90 million, the claim that restructuring fixed the cost base is wrong. The September quarter result of $0.3 million leaves no cushion. We would check this in the 10-K for fiscal 2026, due in early 2027.
2. Robotics scales faster than we expect. If Robotics revenue exceeds $120 million in any quarter of 2027, while segment pre-tax margin exceeds 5%, then cobots are a growth line and our view of the segment as a stable, low-weight business is too cautious. The comparison point is $99.9 million in the second quarter of 2026, the segment’s best recent quarter.
3. AI test demand rolls over. If Semiconductor Test revenue falls below $1,000 million in any quarter through the end of 2027, the three-quarter run above $1 billion would be revealed as a peak and the premise that test cash flows persist weakens. The most recent print is $1,072.0 million, and the first checkpoint is first-quarter 2027 guidance relative to the fourth-quarter range of $1,250 million to $1,350 million.
4. The Robotics goodwill is impaired. If Teradyne records a goodwill impairment on the $416.4 million Robotics balance in the annual test as of December 31, 2026, then the market-based valuation of the unit has fallen below its carrying value, which would contradict a recovery thesis. The disclosure would appear in the fiscal 2026 10-K.

Summary

Teradyne earns its profit from chip test. In the third quarter of 2026 Semiconductor Test brought in $1,072.0 million and $366.9 million of pre-tax income, on AI compute and memory demand that more than doubled the segment’s nine-month revenue. The quarter also showed a plateau: total revenue was down sequentially, margins slipped, inventory is up about half since December, and the stock fell from $323.68 to $274.37 between October 27 and 30.

Robotics has been repaired but is not yet growing. After three years of declining sales and widening losses, a roughly 400-person restructuring and a lower cost base moved the segment to a $0.3 million profit on $95.8 million of revenue. Revenue has recovered about 30% in 2026 but levelled at $90 million to $100 million per quarter, the segment is 7% of company sales, and it carries $416.4 million of goodwill. Calling cobots a growth line would take revenue above $120 million a quarter at a healthy margin, and we have not seen it.

For the equity, the debate is the durability of AI test volumes against a market capitalization near $43 billion, about 31.5 times trailing GAAP earnings (all as of October 30, 2026). The next data points are the fourth-quarter 2026 results and the goodwill test at year-end.