That's how much Cognex increased its quarterly operating profit from a year earlier.
Revenue reached a record $313 million. Profit margins expanded, cash generation improved, and management raised its full-year outlook.
Its stock still finished the week down 12.6%.
Rockwell Automation told a similar story. Its factory-control software business grew 15%, margins improved, and cash generation increased. Its shares fell 11%.
These weren't struggling businesses announcing deteriorating results. They were established industrial technology companies showing meaningful operational improvements while investors sold their stocks.
And that raises an interesting question for Physical AI.
Are investors discovering problems in factory automation, or are they simply becoming less willing to pay for the future of AI?
This week's earnings, economic data and industry developments suggest the answer is more complicated than either explanation alone.
The AI selloff reached far beyond Silicon Valley
November 5 was one of the most difficult trading days for technology stocks in recent years.
The Nasdaq Composite plunged 5.52%. The Nasdaq 100 fell 6.4%, wiping out approximately $2.5 trillion in market value. The Philadelphia Semiconductor Index fell about 12%, while Nvidia lost more than 9%.
According to Reuters, the selloff followed the U.S. midterm elections, as investors worried that greater political scrutiny could slow data-center construction.
The concerns extended beyond Washington. Electricity costs, local opposition, permitting delays and the enormous financing requirements of AI infrastructure were already making investors question how quickly the industry's spending plans could be realized.
UBS analysts estimated that roughly 30% of planned U.S. data-center capacity additions for 2028 could face risks from construction pauses.
That doesn't mean 30% of projects have been canceled. It illustrates how much uncertainty surrounds an expansion that investors had increasingly treated as inevitable.
For Physical AI, the distinction matters.
A company selling equipment to semiconductor factories is exposed to a different customer budget than one supplying inspection cameras to packaging plants or software to warehouse operators.
But when investors suddenly reduce their expectations for AI growth, many of those companies can fall together.
Our Black Scarab Physical AI 15 reflects that tension.
The index gained just 0.36% between the November 3 and November 10 closes, outperforming the S&P 500's 0.63% decline and the Nasdaq's 1.21% loss.
Only five of its fifteen constituents rose.
And two sensor companies accounted for much of the positive momentum.
The headline index return concealed an increasingly divided market.
November 4–10
Weekly market performance
November 3 to November 10 closes. Price returns exclude dividends and FX. Japan starts from November 2 because November 3 was a market holiday. Source: Black Scarab.
Cognex and Rockwell delivered better results than their stocks suggested
Last week, we identified Cognex and Rockwell as important tests of whether improving U.S. factory activity was translating into stronger demand for automation.
Both provided encouraging evidence.
Cognex, which supplies machine vision systems that help factories inspect, identify and measure products, reported quarterly revenue growth of 13%. Excluding an unusual payment received in the previous year, growth reached 19%.
More importantly, operating profit grew 74%, and operating margin climbed from 20.9% to 32.2%.
Think about that difference.
For every $100 in sales, Cognex generated roughly $32 in operating profit, compared with $21 a year earlier.
Cognex
Operating margin
GAAP operating margin, comparable third quarters. Source: Cognex, November 4.
Free cash flow also increased to approximately $104 million.
These aren't projections of what AI might accomplish in five years. They're financial results from equipment being sold into real industrial markets.
Management raised its full-year profitability outlook. Its previously announced agreement to acquire RealSense for approximately $500 million, disclosed on September 22, would expand its reach from traditional machine vision into robotic depth sensing and perception.
Yet the stock declined.
Part of the selling preceded the November 4 earnings announcement, and the following session coincided with the broader technology rout. The earnings themselves were also broadly in line with Wall Street's expectations, meaning strong growth did not necessarily translate into a positive surprise.
Rockwell's results require similar context.
At first glance, quarterly revenue growth of just 1% looked disappointing.
But that figure included an $85 million accrual for refunds to customers affected by earlier tariffs, reducing reported and organic growth by approximately four percentage points.
Beneath that accounting effect, Rockwell's Software & Control division delivered 15% sales growth, with operating margin increasing to 34%.
Quarterly free cash flow climbed to $602 million from $405 million, although tariff refunds also contributed to the increase.
Management expects organic sales growth of 6–9% in fiscal 2027, including roughly one percentage point from the easier comparison created by this year's refund program.
There are reasons for investors to remain cautious. Neither company is insulated from semiconductor investment, both face changing customer budgets, and their future growth still has to justify their valuations.
But the operating improvements are difficult to dismiss.
The lesson isn't that factory automation is immune to an AI correction.
It's that a falling stock price and a deteriorating industrial business are not necessarily the same thing.
The lidar rally revealed another divide
While Cognex and Rockwell struggled in the market, two companies supplying the sensing systems behind autonomous machines delivered remarkable gains.
Ouster shares rose 31.9% during our measurement period. Hesai gained 19.3%.
The companies benefited from encouraging business developments, although the timing matters.
Ouster reported earnings after the November 3 close, just before our performance window began. Quarterly revenue increased 50% to approximately $59 million, supported by industrial and robotics customers.
The company shipped more than 27,000 lidar and camera sensors, with applications spanning warehouses, construction, mapping and autonomous systems.
That is tangible evidence of growing demand for the technologies that help machines understand their surroundings.
But Ouster still reported a $14 million net loss.
Revenue growth is accelerating, yet the company has more work to do before that growth translates into consistent profitability.
Hesai offers an especially revealing comparison.
Its November 10 results showed quarterly revenue growth of 34%, with lidar shipments reaching nearly 730,000 units.
Most of those units still served automotive driver-assistance systems rather than robotics.
The company's established lidar business generated RMB145.5 million in operating profit, an 88% increase from a year earlier.
Its newer Strategic Growth Initiatives division, which includes robotic actuation and spatial-intelligence technologies, generated RMB90.5 million in revenue but lost RMB82.9 million from operations.
Hesai · Q3 2026
From lidar profit to group profit
Strategic Growth Initiatives includes robotic actuation and Kosmo. Operating profit and loss, rounded. Source: Hesai, November 10.
In other words, Hesai's profitable sensing business is helping finance its expansion into new Physical AI markets.
That is not necessarily a problem. Companies often invest heavily before emerging businesses reach scale.
But it is an important distinction for shareholders.
Hesai's established business is already generating operating profits. Its newer ventures still need to demonstrate that revenue growth can eventually cover their development and operating costs.
We've previously examined one of those ventures, Kosmo, which aims to turn captured physical environments into reusable training data for robots.
The opportunity is intriguing. The financial returns remain to be demonstrated.
And although Hesai reported strong results on November 10, its shares had already gained substantially before the announcement. The entire weekly rally cannot be attributed to those earnings.
Capital is still finding its way into Physical AI
The market selloff hasn't stopped companies from investing in technologies designed to make robots more useful.
In fact, several announcements this week suggest competition is expanding beyond the robots themselves.
On November 9, NXP Semiconductors introduced a robotics platform connecting AI processing, sensors, motor control, communications and security through reusable software and hardware designs.
The goal is to reduce the time engineers spend integrating components before a robot can perform useful work.
Teradyne announced Omnyx Edge the same day, extending its automated testing capabilities from data-center hardware to the electronic boards used inside robots, autonomous vehicles and other intelligent machines.
These announcements illustrate an important shift.
As robots become more capable, the businesses supplying their intelligence, sensors, electronics, testing systems and development tools may become increasingly valuable.
But product introductions aren't evidence of significant new revenue. Both companies still need customers to adopt their offerings at scale.
Meanwhile, private investment continues.
On November 10, Path Robotics raised $300 million at a $2.45 billion valuation to expand its autonomous welding technology into additional manufacturing applications.
Path reports that annual recurring revenue has increased seventeenfold over eighteen months, although it hasn't disclosed the starting revenue figure.
CEO Andy Lonsberry summarized the ambition clearly:
"Our goal isn't to automate one step, but to automate the entire factory."
That statement captures a challenge we've discussed repeatedly at Black Scarab.
Automating one task can improve productivity. But if the next stage of production cannot keep up, the bottleneck simply moves elsewhere.
For manufacturers, the value of Physical AI depends not just on what an individual robot can accomplish, but on whether the overall operation becomes more productive.
And that requires integration, maintenance, trained personnel and dependable equipment.
The next major commercial opportunity may come as much from making automation easier to deploy as from making the robots themselves more intelligent.
Inflation makes the productivity question more urgent
The week's economic data added another dimension to the investment picture.
On November 10, the Bureau of Labor Statistics reported that U.S. consumer prices increased 0.4% in October, exceeding economists' expectations of 0.3%.
Persistent inflation is increasing expectations that the Federal Reserve could raise interest rates again.
Higher borrowing costs matter to an industry that depends on capital investment.
A manufacturer considering a new robotic production line must weigh the cost of financing the project against the labor savings, production increases and other benefits it expects to receive.
When financing becomes more expensive, projects with uncertain returns become harder to justify.
But another economic release deserves attention.
On November 5, the government reported that overall U.S. nonfarm business productivity increased at a 3.6% annualized rate in the third quarter.
Manufacturing productivity increased just 0.3% at an annualized rate.
United States · Q3 2026
Productivity growth
Output per hour, seasonally adjusted annualized rates. Sector output definitions differ; these data do not isolate AI’s contribution. Source: BLS, November 5.
The difference is striking.
America's broader economy is becoming more productive, but those gains aren't appearing at the same pace in manufacturing.
A single quarter cannot establish a long-term trend, and these figures don't isolate the contribution of AI or robotics. BLS also notes that the two sectors use different output measures.
Indeed Hiring Lab economist Aubrey Woessner cautioned that a major AI-driven acceleration in efficiency has not yet appeared clearly in the economic data.
For industrial automation companies, this creates both an opportunity and a challenge.
Factories have substantial room to improve productivity. But investors and customers will increasingly want evidence that new technology can deliver those improvements at an acceptable cost.
What we're watching next
Siemens reports fiscal fourth-quarter results on November 12, providing another important test of industrial demand.
We'll be looking for evidence that automation orders are improving beyond semiconductor and AI infrastructure customers, alongside sustained margins and cash generation.
The following day, Federal Reserve Vice Chair Philip Jefferson is scheduled to discuss technology and productivity, a timely subject as markets weigh higher financing costs against the potential economic benefits of AI.
Our outlook for Physical AI remains constructive, particularly for specialized systems that solve expensive problems in existing industrial operations.
But we're becoming more selective about what constitutes meaningful progress.
Growing orders matter. Profitable deployments matter more. And a company that can expand without requiring proportionately more capital, engineering and customer support may ultimately create more value than one growing rapidly without improving its economics.
The AI selloff has made those distinctions harder to ignore.
The opportunity in Physical AI hasn't disappeared. Investors are simply being reminded that building intelligent machines and earning attractive returns from them are two different accomplishments.
Sources and methodology
Coverage: November 4 to 10, 2026. Returns use November 3 to 10 closing prices, excluding dividends and currency movements. Japanese listings start from November 2 because November 3 was a holiday. Company forecasts, claims and operating measures retain their stated scope.
Factory automation faces a more skeptical market
In this edition +
Explore the Physical AI 15 ↗74%.
That's how much Cognex increased its quarterly operating profit from a year earlier.
Revenue reached a record $313 million. Profit margins expanded, cash generation improved, and management raised its full-year outlook.
Its stock still finished the week down 12.6%.
Rockwell Automation told a similar story. Its factory-control software business grew 15%, margins improved, and cash generation increased. Its shares fell 11%.
These weren't struggling businesses announcing deteriorating results. They were established industrial technology companies showing meaningful operational improvements while investors sold their stocks.
And that raises an interesting question for Physical AI.
Are investors discovering problems in factory automation, or are they simply becoming less willing to pay for the future of AI?
This week's earnings, economic data and industry developments suggest the answer is more complicated than either explanation alone.
The AI selloff reached far beyond Silicon Valley
November 5 was one of the most difficult trading days for technology stocks in recent years.
The Nasdaq Composite plunged 5.52%. The Nasdaq 100 fell 6.4%, wiping out approximately $2.5 trillion in market value. The Philadelphia Semiconductor Index fell about 12%, while Nvidia lost more than 9%.
According to Reuters, the selloff followed the U.S. midterm elections, as investors worried that greater political scrutiny could slow data-center construction.
The concerns extended beyond Washington. Electricity costs, local opposition, permitting delays and the enormous financing requirements of AI infrastructure were already making investors question how quickly the industry's spending plans could be realized.
UBS analysts estimated that roughly 30% of planned U.S. data-center capacity additions for 2028 could face risks from construction pauses.
That doesn't mean 30% of projects have been canceled. It illustrates how much uncertainty surrounds an expansion that investors had increasingly treated as inevitable.
For Physical AI, the distinction matters.
A company selling equipment to semiconductor factories is exposed to a different customer budget than one supplying inspection cameras to packaging plants or software to warehouse operators.
But when investors suddenly reduce their expectations for AI growth, many of those companies can fall together.
Our Black Scarab Physical AI 15 reflects that tension.
The index gained just 0.36% between the November 3 and November 10 closes, outperforming the S&P 500's 0.63% decline and the Nasdaq's 1.21% loss.
Only five of its fifteen constituents rose.
And two sensor companies accounted for much of the positive momentum.
The headline index return concealed an increasingly divided market.
Weekly market performance
Cognex and Rockwell delivered better results than their stocks suggested
Last week, we identified Cognex and Rockwell as important tests of whether improving U.S. factory activity was translating into stronger demand for automation.
Both provided encouraging evidence.
Cognex, which supplies machine vision systems that help factories inspect, identify and measure products, reported quarterly revenue growth of 13%. Excluding an unusual payment received in the previous year, growth reached 19%.
More importantly, operating profit grew 74%, and operating margin climbed from 20.9% to 32.2%.
Think about that difference.
For every $100 in sales, Cognex generated roughly $32 in operating profit, compared with $21 a year earlier.
Operating margin
Free cash flow also increased to approximately $104 million.
These aren't projections of what AI might accomplish in five years. They're financial results from equipment being sold into real industrial markets.
Management raised its full-year profitability outlook. Its previously announced agreement to acquire RealSense for approximately $500 million, disclosed on September 22, would expand its reach from traditional machine vision into robotic depth sensing and perception.
Yet the stock declined.
Part of the selling preceded the November 4 earnings announcement, and the following session coincided with the broader technology rout. The earnings themselves were also broadly in line with Wall Street's expectations, meaning strong growth did not necessarily translate into a positive surprise.
Rockwell's results require similar context.
At first glance, quarterly revenue growth of just 1% looked disappointing.
But that figure included an $85 million accrual for refunds to customers affected by earlier tariffs, reducing reported and organic growth by approximately four percentage points.
Beneath that accounting effect, Rockwell's Software & Control division delivered 15% sales growth, with operating margin increasing to 34%.
Quarterly free cash flow climbed to $602 million from $405 million, although tariff refunds also contributed to the increase.
Management expects organic sales growth of 6–9% in fiscal 2027, including roughly one percentage point from the easier comparison created by this year's refund program.
There are reasons for investors to remain cautious. Neither company is insulated from semiconductor investment, both face changing customer budgets, and their future growth still has to justify their valuations.
But the operating improvements are difficult to dismiss.
The lesson isn't that factory automation is immune to an AI correction.
It's that a falling stock price and a deteriorating industrial business are not necessarily the same thing.
The lidar rally revealed another divide
While Cognex and Rockwell struggled in the market, two companies supplying the sensing systems behind autonomous machines delivered remarkable gains.
Ouster shares rose 31.9% during our measurement period. Hesai gained 19.3%.
The companies benefited from encouraging business developments, although the timing matters.
Ouster reported earnings after the November 3 close, just before our performance window began. Quarterly revenue increased 50% to approximately $59 million, supported by industrial and robotics customers.
The company shipped more than 27,000 lidar and camera sensors, with applications spanning warehouses, construction, mapping and autonomous systems.
That is tangible evidence of growing demand for the technologies that help machines understand their surroundings.
But Ouster still reported a $14 million net loss.
Revenue growth is accelerating, yet the company has more work to do before that growth translates into consistent profitability.
Hesai offers an especially revealing comparison.
Its November 10 results showed quarterly revenue growth of 34%, with lidar shipments reaching nearly 730,000 units.
Most of those units still served automotive driver-assistance systems rather than robotics.
The company's established lidar business generated RMB145.5 million in operating profit, an 88% increase from a year earlier.
Its newer Strategic Growth Initiatives division, which includes robotic actuation and spatial-intelligence technologies, generated RMB90.5 million in revenue but lost RMB82.9 million from operations.
From lidar profit to group profit
In other words, Hesai's profitable sensing business is helping finance its expansion into new Physical AI markets.
That is not necessarily a problem. Companies often invest heavily before emerging businesses reach scale.
But it is an important distinction for shareholders.
Hesai's established business is already generating operating profits. Its newer ventures still need to demonstrate that revenue growth can eventually cover their development and operating costs.
We've previously examined one of those ventures, Kosmo, which aims to turn captured physical environments into reusable training data for robots.
The opportunity is intriguing. The financial returns remain to be demonstrated.
And although Hesai reported strong results on November 10, its shares had already gained substantially before the announcement. The entire weekly rally cannot be attributed to those earnings.
Capital is still finding its way into Physical AI
The market selloff hasn't stopped companies from investing in technologies designed to make robots more useful.
In fact, several announcements this week suggest competition is expanding beyond the robots themselves.
On November 9, NXP Semiconductors introduced a robotics platform connecting AI processing, sensors, motor control, communications and security through reusable software and hardware designs.
The goal is to reduce the time engineers spend integrating components before a robot can perform useful work.
Teradyne announced Omnyx Edge the same day, extending its automated testing capabilities from data-center hardware to the electronic boards used inside robots, autonomous vehicles and other intelligent machines.
These announcements illustrate an important shift.
As robots become more capable, the businesses supplying their intelligence, sensors, electronics, testing systems and development tools may become increasingly valuable.
But product introductions aren't evidence of significant new revenue. Both companies still need customers to adopt their offerings at scale.
Meanwhile, private investment continues.
On November 10, Path Robotics raised $300 million at a $2.45 billion valuation to expand its autonomous welding technology into additional manufacturing applications.
Path reports that annual recurring revenue has increased seventeenfold over eighteen months, although it hasn't disclosed the starting revenue figure.
CEO Andy Lonsberry summarized the ambition clearly:
That statement captures a challenge we've discussed repeatedly at Black Scarab.
Automating one task can improve productivity. But if the next stage of production cannot keep up, the bottleneck simply moves elsewhere.
For manufacturers, the value of Physical AI depends not just on what an individual robot can accomplish, but on whether the overall operation becomes more productive.
And that requires integration, maintenance, trained personnel and dependable equipment.
The next major commercial opportunity may come as much from making automation easier to deploy as from making the robots themselves more intelligent.
Inflation makes the productivity question more urgent
The week's economic data added another dimension to the investment picture.
On November 10, the Bureau of Labor Statistics reported that U.S. consumer prices increased 0.4% in October, exceeding economists' expectations of 0.3%.
Persistent inflation is increasing expectations that the Federal Reserve could raise interest rates again.
Higher borrowing costs matter to an industry that depends on capital investment.
A manufacturer considering a new robotic production line must weigh the cost of financing the project against the labor savings, production increases and other benefits it expects to receive.
When financing becomes more expensive, projects with uncertain returns become harder to justify.
But another economic release deserves attention.
On November 5, the government reported that overall U.S. nonfarm business productivity increased at a 3.6% annualized rate in the third quarter.
Manufacturing productivity increased just 0.3% at an annualized rate.
Productivity growth
The difference is striking.
America's broader economy is becoming more productive, but those gains aren't appearing at the same pace in manufacturing.
A single quarter cannot establish a long-term trend, and these figures don't isolate the contribution of AI or robotics. BLS also notes that the two sectors use different output measures.
Indeed Hiring Lab economist Aubrey Woessner cautioned that a major AI-driven acceleration in efficiency has not yet appeared clearly in the economic data.
For industrial automation companies, this creates both an opportunity and a challenge.
Factories have substantial room to improve productivity. But investors and customers will increasingly want evidence that new technology can deliver those improvements at an acceptable cost.
What we're watching next
Siemens reports fiscal fourth-quarter results on November 12, providing another important test of industrial demand.
We'll be looking for evidence that automation orders are improving beyond semiconductor and AI infrastructure customers, alongside sustained margins and cash generation.
The following day, Federal Reserve Vice Chair Philip Jefferson is scheduled to discuss technology and productivity, a timely subject as markets weigh higher financing costs against the potential economic benefits of AI.
Our outlook for Physical AI remains constructive, particularly for specialized systems that solve expensive problems in existing industrial operations.
But we're becoming more selective about what constitutes meaningful progress.
Growing orders matter. Profitable deployments matter more. And a company that can expand without requiring proportionately more capital, engineering and customer support may ultimately create more value than one growing rapidly without improving its economics.
The AI selloff has made those distinctions harder to ignore.
The opportunity in Physical AI hasn't disappeared. Investors are simply being reminded that building intelligent machines and earning attractive returns from them are two different accomplishments.
Sources and methodology
Coverage: November 4 to 10, 2026. Returns use November 3 to 10 closing prices, excluding dividends and currency movements. Japanese listings start from November 2 because November 3 was a holiday. Company forecasts, claims and operating measures retain their stated scope.
About the author
Rodolfo Garcia Calderoni, CFAFounder, Black Scarab