Inside Ouster: Angus Pacala on LiDAR and Physical AI

Ouster spent more than a decade building technology that allows machines to see the physical world.

Today, its lidar and perception technology is used across robotics, autonomous vehicles, industrial automation, drones, mapping, and intelligent infrastructure.

But the company's path was anything but straightforward.

In San Francisco, Very Frontier sat down with Ouster CEO and co-founder Angus Pacala to understand how the company went from two Stanford engineers working out of a warehouse to surviving one of the most brutal shakeouts in deep tech.

Pacala discussed why Ouster bet on semiconductor-based digital lidar before Physical AI had a name, why so many lidar companies disappeared, how the company survived several moments near bankruptcy, and why he believes the next decade of robotics could be much bigger than the last.

The conversation offers a rare look at what building a real deep-tech hardware company actually looks like when the product has to leave the lab, operate safely in the physical world, and work reliably for thousands of customers.

Ouster was founded in 2015 by Stanford engineers Angus Pacala and Mark Frichtl around the idea that lidar would eventually become a semiconductor technology.

The founders believed autonomy would spread far beyond robotaxis into almost every category of moving machine.

Ouster deliberately built one core technology platform that could serve robotics, automotive, industrial, mapping, and infrastructure customers instead of depending on a single market.

The company came close to bankruptcy multiple times while crossing what Pacala describes as the hardware "chasm" between having a working product and reaching profitable scale.

Its merger with Velodyne became an important part of reaching greater financial and operational scale.

Rev8, Ouster's native color lidar platform, traces back to an idea that appeared in the company's original 2015 pitch deck.

Pacala believes founders building Physical AI cannot market their way around engineering, reliability, and safety.

Ouster was founded in 2015 by Angus Pacala and Mark Frichtl, two engineers who had met at Stanford and had already spent time working around lidar technology.

The original idea did not begin with a specific robot or autonomous vehicle.

It began with a technology thesis.

Pacala and Frichtl would meet in coffee shops around Silicon Valley and ask a more fundamental question: what would lidar look like ten or twenty years into the future?

Their answer came from looking at the history of electronics.

Industries built around analog technologies repeatedly moved toward semiconductor-based systems. Cameras became digital. GPS receivers became semiconductor devices. CPUs, GPUs, inertial sensors, and many other technologies followed similar trajectories.

Pacala and Frichtl believed lidar would eventually do the same.

Instead of building lidar primarily from complex analog architectures, Ouster would attempt to build what it called digital lidar around semiconductor technology.

That idea became the technological foundation of the company.

One of the more interesting parts of Ouster's story is that the company was not created specifically for today's Physical AI boom.

The founders were already seeing the underlying trend a decade earlier.

Around 2015, autonomous driving and robotaxis were absorbing billions of dollars of investment in Silicon Valley. But Pacala believed the significance of that investment extended beyond cars.

Hundreds and eventually thousands of engineers were being trained to build autonomous systems.

Those engineers would not remain inside the robotaxi industry forever.

They would move into drones, delivery robots, agriculture, mining, industrial equipment, logistics, mapping, and entirely new categories of machines.

At the same time, traditional equipment manufacturers were beginning to look for ways to make their machines safer, more automated, and more productive.

The combination created a much larger thesis.

Autonomy would eventually propagate into almost anything capable of moving through the physical world.

Today that idea has a name: Physical AI.

Many lidar companies were built around the assumption that autonomous passenger vehicles would become the dominant market for their technology.

Ouster chose a different strategy.

From the beginning, Pacala says the company wanted to build products that could work across multiple industries.

The same underlying sensing platform could potentially be used by a mapping company, a delivery robot, a drone, a mining vehicle, agricultural equipment, intelligent infrastructure, or an autonomous truck.

That diversity became strategically important.

Instead of requiring one particular autonomous vehicle market to develop on schedule, Ouster could sell similar technology into several different industries.

Today, the company's broader platform includes lidar, cameras, edge compute, perception software, and other technologies designed to allow machines to understand their environments.

The goal is increasingly larger than selling an individual sensor.

Ouster wants to become part of the sensing and perception layer underlying Physical AI.

Deep tech has no shortage of impressive demonstrations.

Pacala argues that demos are a poor substitute for real deployment.

A lidar sensor operating inside a safety-critical system has to continue working across weather, vibration, temperature changes, manufacturing variation, and unpredictable physical environments.

A prototype can work once.

A commercial product has to work repeatedly.

That distinction shaped Ouster's engineering culture.

Pacala says the company consistently sought negative customer feedback instead of trying to protect itself from criticism.

If something failed in the field, the goal was to understand why and engineer the failure out of the product.

That mindset becomes particularly important in Physical AI because the software and hardware can affect real machines moving around real people.

There is eventually nowhere for weak engineering to hide.

Ouster survived a period that eliminated or weakened many of its competitors.

During the autonomous vehicle boom, lidar companies attracted enormous amounts of capital. Expectations for self-driving vehicles accelerated faster than real-world deployment.

When those expectations reset, the lidar market went through a painful correction.

Pacala describes Ouster's journey during much of this period as a grind rather than a conventional Silicon Valley success story.

The company came close to bankruptcy multiple times.

Hardware businesses face a particularly difficult transition.

Once a company begins manufacturing a real product, it takes on the cost of factories, supply chains, engineers, inventory, support, and customers.

But revenue may still be too small to support those costs.

Pacala describes this as a chasm.

A hardware startup has already become expensive to operate, but it has not yet achieved the scale required to sustain itself.

Many companies never make it across.

Ouster did, but Pacala says it required years of pushing, scraping, and continuing to execute even when the financial position was uncomfortable.

A major turning point came when Ouster and Velodyne combined in 2023.

The companies had overlapping customers and complementary technologies, but the merger was also about scale and survival.

The combined company could consolidate operations, technology, customer relationships, and financial resources at a time when the lidar industry was undergoing significant pressure.

For Ouster, it was another step toward becoming something larger than a single-product lidar startup.

The company subsequently expanded further into software and perception and acquired Stereolabs, bringing stereo cameras, AI vision, compute, and additional perception capabilities into the business.

The direction is clear.

Ouster increasingly wants to provide the complete perception stack around the machine rather than simply one sensor attached to it.

Perhaps the best example of how long deep-tech development can take is Ouster's Rev8 platform.

Rev8 introduced native color lidar, combining three-dimensional depth information with color information inside the sensing system.

The idea sounds like a logical next step.

For Ouster, it was nearly a decade in the making.

Pacala says the company's first pitch deck from July 2015 already included the goal of producing colorized point clouds.

Getting there required multiple generations of Ouster's underlying chip architecture.

This is a useful reminder of how differently progress happens in hardware compared with many software businesses.

A software product can sometimes move from idea to deployment in weeks.

A sensing architecture that must operate reliably inside vehicles, robots, infrastructure, and heavy machinery may require years of semiconductor development, manufacturing work, testing, and field validation.

The roadmap can be measured in decades rather than quarters.

Humans do not understand the world using geometry alone.

We combine shape, depth, texture, color, motion, and context.

Robots face a similar problem.

Traditional lidar is particularly powerful at understanding three-dimensional structure. Cameras capture rich visual information.

Bringing those sensing modalities closer together can give autonomous systems a more complete representation of their surroundings.

That becomes increasingly important as robots leave controlled environments.

A warehouse robot operates in a relatively structured space.

A delivery robot moving through a city, an autonomous machine working in agriculture, or a drone inspecting infrastructure must understand far more complicated environments.

Better perception becomes a fundamental prerequisite for better autonomy.

Much of the recent AI boom has happened inside computers.

Models generate text, software, images, and video.

Physical AI introduces another constraint: the output eventually has to interact with reality.

A mistake from a chatbot might produce an incorrect sentence.

A mistake from an autonomous vehicle or industrial machine can have physical consequences.

That changes the engineering standard.

Pacala argues that companies operating in these markets cannot rely on hype as a substitute for reliability.

Safety-critical systems require detailed engineering across hardware, software, manufacturing, calibration, testing, and deployment.

Every tolerance matters.

That is one reason the development cycles can be so long.

When asked whether he understood how difficult the journey would be when starting Ouster, Pacala's answer was effectively no.

And that may have been necessary.

Founders attempting to build extremely difficult technologies often begin without fully appreciating the number of problems that will have to be solved.

Pacala describes a certain degree of ignorance or naivety as almost necessary to look at an established technical industry and decide that you are going to rebuild it.

What matters afterward is whether that optimism is accompanied by execution.

For Ouster, that meant years of engineering, listening to customers, recruiting people with expertise the founders did not have, and continuing through periods where the company's survival was uncertain.

Pacala also emphasizes knowing what you do not know.

He and Frichtl came from product and engineering backgrounds.

As Ouster expanded, they needed people with deeper expertise in manufacturing, operations, finance, and other areas required to turn engineering into an industrial business.

Building the technology was only part of building the company.

Despite the enormous amount of capital invested in robotics and autonomy over the past decade, the physical world still looks surprisingly conventional.

There are not millions of autonomous mobile robots moving through every city.

Most vehicles still require drivers.

Most industrial machines are not fully autonomous.

Robotic systems remain relatively rare outside specific industries and geographic areas.

For Pacala, that gap represents the opportunity.

The technology has advanced substantially, but widespread deployment is still early.

Better sensors, lower costs, more powerful edge compute, stronger AI models, and better perception software could finally allow autonomous machines to operate across a much broader range of real-world environments.

Ouster's future is therefore broader than lidar.

The company is expanding across cameras, stereo vision, compute, software, sensor fusion, and perception.

It is also building complete applications in areas such as intelligent transportation infrastructure.

This reflects a larger shift happening across robotics.

The value of an individual sensor matters.

But the larger opportunity is increasingly in combining sensing, computing, and intelligence into a system capable of understanding and acting in the physical world.

For Ouster, the ambition is to sit underneath that transition.

If the next decade really does bring autonomous machines into logistics, agriculture, transportation, construction, cities, and everyday environments, those machines will all need some way to understand what is around them.

The race to build the intelligence of robots is already underway.

Ouster is betting that building their eyes may be just as important.

What does Ouster do?

Ouster develops sensing and perception technology for autonomous systems and Physical AI. Its platform includes digital lidar sensors, cameras, compute, and perception software used across robotics, industrial automation, automotive applications, and smart infrastructure.

Who founded Ouster?

Ouster was founded in 2015 by Stanford engineers Angus Pacala and Mark Frichtl. Pacala serves as CEO and Frichtl as CTO.

What is digital lidar?

Digital lidar is Ouster's approach to building lidar around semiconductor technology. Lidar measures the physical environment using light to create precise three-dimensional information that robots and autonomous systems can use to understand their surroundings.

What is Ouster Rev8?

Rev8 is Ouster's latest generation of digital lidar sensors. It introduced native color lidar, combining 3D depth measurements and color information to provide richer perception data for autonomous systems.

What is Physical AI?

Physical AI refers broadly to AI systems capable of perceiving, reasoning about, and acting within the physical world. Examples include autonomous vehicles, delivery robots, drones, industrial robots, agricultural machinery, and intelligent infrastructure.

Why is lidar important for robots?

Lidar gives machines accurate three-dimensional information about their surroundings. That can help robots understand distance, geometry, obstacles, and movement, particularly in environments where reliable spatial perception is critical.