Shiny Cars AI wants to solve one of the least glamorous but most consequential problems in architecture: what happens to quality data when a human has to touch the production line.
The San Francisco-based manufacturer announced today Hybrid BRC (Shiny Robot Cell) is an extension of the Shining Factory platform that allows human operators to step inside a sensor-controlled robotic cell to perform defined assembly steps without breaking a digital record that tracks every server from its first screw to its shipping label.
Sounds like an incremental hardware update. It is not. Hybrid BRC is a direct response to structural weakness in high-stakes electronics manufacturing — CEO Sviat Dulianinov starkly quantified in an exclusive interview with VentureBeat.
"If you’re building modern AI servers from manual operations, your initial revenue – your first pass revenue – can be as low as 20%," Dulianinov said. "Then you gradually go up and scale up and it can be in the 60s, 65% or more."
When a single AI server costs hundreds of thousands of dollars and hyperscalers burn billions waiting for infrastructure they can’t deploy fast enough, that number is the whole story. The Hybrid BRC It’s Bright Machines’ attempt to keep human hands in the loop without letting human error in the door.
Why do manual assembly steps create a black hole in production data?
Modern automated assembly lines generate a continuous flow of production data – torque values, positioning coordinates, component serial numbers, inspection images. This "data thread" it allows the manufacturer to prove that the server was set up correctly, and when something fails in the field months later, it allows them to trace the failure back to a specific station, step or part.
But automated lines inevitably require manual intervention, and until now, when this happened, manufacturers had two bad options: shut down the line entirely or move the in-process equipment to a separate manual workstation outside the monitored data flow. The first choice kills productivity. Second, it opens a hole in the production record at a time when human error is most likely to occur.
The Hybrid BRC the company says it’s eliminating those exchanges. Cell guarded entrance doors and safety panels are included directly in the production line. When the operator opens the doors, the robotic arm is turned off and on-screen instructions guide the operator through each assembly step, while the cell’s sensor array – cameras, force feedback and tool sensors – continues to monitor for incorrect installations, missed steps and incorrect components, applying the same quality checks used in full automation. A tracking record is kept at the serial number level from start to finish.
Difference in revenue between humans and robots in AI server assembly
The economy driving the design becomes clear when Dulianinov’s manual assembly figures are juxtaposed against what automation provides. "In robotic operations, productivity per station is typically greater than 98% with our technology, and even at the line level, we typically achieve 97.5%, 97.7% or more," he said.
First pass efficiency measures the percentage of units that exit the correction line the first time without rework. The gap between a 20% manual landing and a 98% automated station isn’t a rounding error—it’s the difference between profitability and disaster on such expensive equipment.
This math explains the company’s design philosophy Hybrid BRCit treats the human operator as a rescue valve for exceptions, not a substitute for automation. "The more man stations you introduce, the greater the risk of low output, which lowers overall revenue." Dulianinov said. "So we prefer to start with at least 50% automation and then move to at least 80%." Speed follows a similar pattern: "At the line level, robots can be 50 to 100% faster than humans." he said in terms of throughput.
How server stacking has become the hidden bottleneck of the AI infrastructure race
The AI infrastructure conversation usually revolves around chip supply, power availability, and data center construction. Dulianinov argues that assembly—the invisible work of turning chips and motherboards into racked, tested, deployable computing—is quietly a big drag on deployment time.
"Once you have your chips and motherboards, you want to be as quick as possible to get it into the data center," he described greenfield deployments where energy and buildings already exist. Building, testing, and frequently rebuilding equipment when quality declines "may be moon" he said. "With more technology used for this, as our technology, we believe we can reduce it by at least a third."
On the call, one company executive added an anecdotal but data point: servers Shiny Cars produces "flies into production" rather than sitting in warehouses waiting to be deployed, it proves that buildability, not just chips or power, trumps hyperscale timelines. The stakes are asymmetric, the executive noted, as the biggest hyperscalers lose millions of dollars a day when servers fail or arrive late. That’s why customers are less interested in buying a box than a warranty – and why the uncut data cord has become a product in its own right.
Inside the secret customer base already operating hybrid production lines
The Hybrid BRC not a steamer. Dulianinov said that the company already operates and has a number of hybrid lines in the United States "Built more than 10,000 computing nodes" through new stations. Bright Machines plans to produce this year "more than half a gigawatt of computing power."
Who buys? don’t ask "Unfortunately, we cannot name the customers. This is the hardest part of our job." Dulianinov said. "They’re pretty secretive because, as you can imagine, everything related to a data center is IP-related."
He offered growth figures: customers increased "more than 3 times this year" against the previous one to manage what he called intersection "physical AI, building AI infrastructure and on land." Demand is spilling over into real estate — the company is moving this fall from its 16th Street office in San Francisco to a Burlingame location that executives describe as three to four times larger. In total, the company said it has deployed more than 130 microfactories in more than 10 countries, serving more than 60 customers and producing more than 300,000 servers.
What sets Bright Machines apart from Tulip, Instrumental and contract manufacturing giants
It was asked how Hybrid BRC’s tracking claims stack up against operator management and inspection software manufacturers. Lala and InstrumentalDulianinov drew a sharp line around business models.
"Tulip is just an interface company for operators. Instrumental, they focus on inspection. It’s just pieces of the puzzle," he said. "We, as a technology-enabled manufacturer, actually run this whole operation… We put our lines in, we put in our software, we put in the data, we put in our people, and we run it end-to-end."
According to him, the right comparison set is the contract manufacturing giants Flex, Jabaland Foxconn — companies that have a complete manufacturing process, but have historically built it on manual labor that generates little information. What’s different about Bright Machines is that it has robot data, sensor data, and now human station data all flowing through a single orchestration layer into a unified environment that the company calls Bright Insights.
This position is remarkable considering the company’s background. Bright Machines was spun off from contract manufacturer Flex eight years ago and has had a turbulent history: the company planned to go public in 2021 in a $1.6 billion SPAC merger. The Wall Street Journal and CFO Divebefore the contract failed. It rose again in June 2024 with a $126 million Series C — managed by funds managed by BlackRock with the participation of $106 million of capital Nvidia, Microsoft, Eclipse, Jabaland Shinhan Securitiesplus $20 million in venture debt from JP Morgan—bringing the total to more than $400 million, the company announced at the time.
Who owns production data and how employees feel about being monitored
Two management questions arise over any system that so closely instruments human work for technical decision makers, and Dulianinov addressed both directly.
In information ownership, he drew a clean line: "Everything related to the inspection of the customer and their devices and parts shall be expressly the protection and property of the customer." The process and robotics data remains with Bright Machines to power continuous improvement on its platform, he said.
Under worker control, he pushed the frame back. He noted that high-IP electronics floors — especially those that touch aerospace, defense or government workloads — already prohibit workers from carrying personal electronics. "People who know those floors know that it’s part of the game," added the workers "actually appreciate it" traceability because it is at the heart of the security mission: "If you build a data center for the government and then build servers somewhere in China, you can’t guarantee how it’s built and what component is put in there." Monitoring, he said, isn’t about tracking employees — it’s about proving, component by component, that America’s built AI infrastructure is what it claims to be.
A bet on land: Rebuilding American manufacturing without 3 million workers
The Hybrid BRCs modular design carries strategic weight beyond quality assurance. Because the cells are defined by software and put together like building blocks, Bright Machines says it can build lines for new generations of hardware in days or weeks, rather than months. "we can deliver within one day" For minor design changes in the product family, the transition from air cooling to liquid cooling remains, Dulianinov said. "big jump." In an industry where new chip architectures now arrive at a roughly annual pace, transition speed is arguably as valuable as revenue; a production line that takes six months to rebuild is obsolete before being depreciated.
But Dulianov’s final argument was not about the car, but about the labor bill. "We have to build in the US and you don’t have 3 million people to do manufacturing in the US." he said, referring to the massive workforce of Shenzhen-scale electronics factories. "So you have to solve it with AI software and robots, and that’s our thesis… It’s not just robots on the ground, it’s creating jobs. All robots and some people on earth."
Lior Susan, founder and CEO of Eclipse and chairman and co-founder of Bright Machines, made the announcement in the same terms: "The future of manufacturing isn’t about choosing between automation and flexibility—it’s combining both in the same digital manufacturing environment."
For all the talk of gigawatts and yield curves, Hybrid BRC amounts to acceptance wrapped in innovation: even in the most automated factories on Earth, people still have to open the door and walk in. Bright Machines’ bet is that the winners of the AI infrastructure race will not be the manufacturers that eliminate the human hand, but the ones that never lose sight of it.
