RoboSense’s Peacock Mass Production Is Technology News With a Harder Test Ahead

RoboSense has moved its Peacock chip into volume shipments, turning a four-month-old silicon announcement into technology news with measurable commercial stakes. The company says its new E2 lidar entered volume production during the third quarter of 2026. It projects more than 200,000 Peacock-based product deliveries during the second half.

That transition matters because Peacock is not another component inside an established sensor. It combines photon detection and signal processing in a company-designed SPAD-SoC, or single-photon avalanche diode system-on-chip. RoboSense is betting that this integration can make solid-state lidar smaller, denser, and easier to manufacture for robots.

The more difficult comparison is now commercial, not conceptual. Rival Hesai shipped more lidar units during the first half and remained profitable in the second quarter. RoboSense, meanwhile, increased shipments sharply but reported a wider net loss and lower gross margin. Peacock must therefore prove that proprietary silicon improves both product performance and business economics.

RoboSense Has Turned Peacock Into a Shipping Product

The important change is that Peacock has crossed from a chip roadmap into a volume-produced lidar with a stated delivery target.

RoboSense first introduced the Peacock chip at its Shenzhen technology event on April 21, 2026. The company presented it alongside Phoenix, a separate SPAD-SoC intended primarily for automotive lidar. Peacock targeted robots, industrial systems, low-speed vehicles, and short-range automotive sensing.

At launch, the company said Peacock would reach mass production during the third quarter. Initial batch deliveries were already underway, according to its chip architecture announcement. That wording described an expected production transition, rather than confirming sustained commercial output.

The confirmation arrived with RoboSense’s interim results, released through the Hong Kong Stock Exchange on August 26. The filing states that E2 commenced volume shipments during the third quarter. It also projects more than 200,000 Peacock-based deliveries during the second half of 2026.

That date resolves an important verification gap in the original hot-list item. Peacock was unveiled in April, while the underlying production milestone was formally reported on August 26. The event is therefore a late-August commercialization update, not a new chip launch in September.

E2 is the first mass-produced RoboSense lidar built around Peacock. The product debuted publicly at the World Artificial Intelligence Conference in July. It is designed for humanoid robots, quadrupeds, robotic lawn mowers, drones, and spatial-data collection systems.

The company describes E2 as a fully solid-state lidar. This means the sensor scans its environment electronically, without the rotating assemblies found in traditional mechanical lidar. Fewer moving parts can support smaller enclosures and longer operating lives, although field reliability still requires customer validation.

Peacock contains a 640 by 480 SPAD array. Each SPAD can detect extremely weak returning light at the photon level. Integrating the detector array and processing functions into one chip lets RoboSense produce dense depth information without cascading several receiver chips.

RoboSense says E2 offers a field of view reaching 180 degrees horizontally and 135 degrees vertically. Its published point rate reaches 1.5 million points per second, while its highest stated precision is one centimeter. These remain manufacturer specifications, not results from an independent comparative benchmark.

The E2 specifications also list typical power consumption of no more than five watts. That characteristic matters for battery-powered machines, where every sensor competes with motors and onboard computing for limited energy.

Mass production changes how those specifications should be evaluated. A prototype can show an impressive point cloud during a controlled demonstration. A shipping sensor must maintain performance across manufacturing variation, temperature changes, vibration, dust, rain, and extended operation.

The reported delivery target creates a clearer test. If RoboSense ships more than 200,000 Peacock-based units before year-end, the chip will have moved beyond limited qualification batches. It would become a meaningful part of the company’s robotics hardware business.

The company has not disclosed the initial customers, order value, or exact product mix behind that projection. It also has not separated firm purchase commitments from internal shipment planning. The milestone is real, but its commercial depth remains only partly visible.

Why This Technology News Centers on Custom Silicon

Peacock matters because lidar competition is moving from assembled sensors toward control of the semiconductor architecture underneath them.

A lidar sends laser light into an environment and measures how long reflected photons take to return. Those measurements produce a three-dimensional representation known as a point cloud. Robots use that data for localization, mapping, obstacle avoidance, and object measurement.

Traditional lidar suppliers often combine externally sourced chips with proprietary optics, scanning systems, firmware, and calibration. That approach can shorten development time, but it leaves the supplier exposed to component availability and vendor pricing. It can also limit how tightly the detector and optical system are optimized together.

RoboSense’s strategy is deeper vertical integration. Its EOCENE architecture provides a common foundation for several internally designed SPAD-SoC products. Peacock serves wide-angle robotics applications, while Phoenix targets longer-range automotive perception.

This division matters because robots and cars impose different requirements. A road vehicle needs to identify objects at long distances and highway speeds. A lawn mower or humanoid needs detailed coverage across a wider nearby area, including objects close to its body.

Peacock concentrates on that near-field challenge. Its stated 180-degree horizontal view lets one sensor cover a broad section around a machine. The 135-degree vertical view can help detect low obstacles, raised objects, and changing terrain.

The detector density also supports more detailed spatial output. RoboSense says the 640 by 480 array moves solid-state lidar toward image-like depth sensing. Dense depth maps can give perception software more information about object contours than sparse ranging points provide.

A higher point rate does not automatically produce better robotic behavior. Algorithms must process the added data within available computing and power limits. Calibration, timing, noise rejection, and sensor fusion can matter as much as nominal resolution.

However, a tightly integrated chip can improve several parts of that pipeline together. The receiver, signal processing, and optical design can be tuned as one system. RoboSense can also revise future sensors around the same chip platform instead of rebuilding each design from disconnected parts.

Manufacturing is the second mechanism. A solid-state electronic scan can reduce the number of precision mechanical components inside the sensor. Integrating functions into one chip can also lower component count and simplify assembly once production yields stabilize.

Those qualifications are important. Custom silicon requires substantial research spending, qualification work, packaging expertise, and foundry coordination. An integrated design only produces a cost advantage when fabrication yields and shipment volumes justify those fixed investments.

RoboSense spent RMB346.5 million on research and development during the first half of 2026, up 12.2 percent year over year. Its filing attributes part of that increase to semiconductor chips and perception sensors for automotive and robotics applications.

The company is therefore paying for vertical integration before its full economic benefit becomes visible. Peacock’s production ramp is the first opportunity to test whether that spending can support scale. Phoenix will provide a separate automotive test when its first vehicle program begins production.

This is what makes the RoboSense Peacock chip more significant than a routine product refresh. It ties sensor performance, supply-chain control, and manufacturing economics to the same proprietary architecture. Each advantage depends on successful production rather than specifications alone.

Robotics Growth Gives Peacock Its Opening

RoboSense is introducing Peacock while robots are becoming a larger and faster-growing part of its actual sales mix.

The company sold approximately 719,200 lidar units during the first half of 2026. That was a 169.6 percent increase from the same period in 2025, according to its interim results.

Products classified as robotics and other applications accounted for 282,600 units. That total rose 510.4 percent year over year. The category represented about 39 percent of company-wide lidar shipments during the period.

The wording deserves attention. RoboSense reports “robotics and others” as one category, so the 282,600 units should not be treated as exclusively robotic. Still, the increase shows that demand outside passenger-vehicle driver assistance has become material.

The revenue mix has also changed. RoboSense said automotive and robotics revenue stood at roughly nine to one during 2024. By the first half of 2026, robotics generated nearly half of product-sales revenue.

That shift gives E2 a larger commercial base than an experimental robotics sensor would have had two years earlier. RoboSense already sells lidar into robotic lawn mowers, logistics equipment, humanoids, quadrupeds, and other mobile platforms. Peacock can enter customer programs through those existing relationships.

Lawn equipment offers a practical example. A mower must map outdoor boundaries, detect people and objects, and handle uneven terrain. A wide viewing angle can reduce blind areas, while solid-state construction avoids an exposed rotating sensor.

Humanoid robots present a different problem. They require near-field perception for navigation and manipulation around people, furniture, and tools. A compact sensor can fit into the head, torso, or limbs without dominating the machine’s industrial design.

Quadruped and inspection robots add shock and vibration. They may operate near machinery, construction sites, or uneven surfaces. A sensor without a mechanical scanning assembly can offer an architectural advantage, although long-duration field data must validate that expectation.

Drones impose strict limits on size, weight, and electrical consumption. E2’s stated power figure and broad view are relevant here. Yet airborne performance also depends on sunlight rejection, motion compensation, and processing latency, none of which a point-rate figure fully captures.

These scenarios explain why Peacock reaches production through E2 before Phoenix reaches automotive production. Robotics programs can often adopt new components on shorter cycles. Vehicle programs require longer validation, functional-safety work, and production commitments extending across several years.

RoboSense also has an unusually large manufacturing base for this market. It said in March that annual lidar capacity had reached four million units. Existing automotive manufacturing systems can support component traceability, calibration, and quality controls for newer robotics products.

Capacity does not equal demand, however. A factory capable of producing four million sensors can still operate below efficient utilization. The more useful signal is whether customers repeatedly order Peacock-based products after their first deployments.

The second-half target would place E2 among the company’s meaningful product ramps, assuming most projected units reach customers. It would also give RoboSense a growing installed base for measuring failures and collecting application feedback.

That feedback can shape later products using the same architecture. A SPAD-SoC platform becomes more valuable when lessons from one sensor improve other sensors. This reuse is central to the company’s argument for designing silicon internally.

Peacock chip explained simply, the bet is not just that a robot sees more points. RoboSense wants one semiconductor foundation to produce many specialized sensing products. Volume shipments are the first evidence that customers will accept that foundation.

Hesai Keeps the Pressure on RoboSense

RoboSense has established a chip-to-product production path, but Hesai currently sets the harder benchmark for shipment scale and profitability.

Hesai reported 1,099,998 total lidar shipments during the first half of 2026. Of those, 260,653 were robotics lidar units. RoboSense reported fewer total units but slightly more units in its broader robotics-and-other category.

The categories are not perfectly comparable. Hesai identifies robotics lidar separately, while RoboSense combines robotics with other applications. Still, the figures show two scaled suppliers competing for many of the same robot and automotive programs.

Hesai shipped 628,275 lidar units during the second quarter alone. Its robotics volume reached 142,371 units, an increase of 193.4 percent from the previous year. The company reported RMB860.8 million in quarterly revenue and RMB70.6 million in net income.

Those competitor results create a clear standard. A lidar supplier must convert expanding volume into dependable earnings, not simply report faster shipment growth. Hesai has already demonstrated that combination across its broader portfolio.

Hesai is also expanding production capacity beyond four million annual units. It has announced large robotics programs, including a supply arrangement covering robotic lawn mowers within the Dreame ecosystem. Such agreements can help spread manufacturing costs across higher output.

RoboSense brings a different point of leverage. Peacock integrates a dense SPAD array for wide-angle solid-state sensing, while its EOCENE architecture is intended to support a family of chips. The company argues this structure will produce cost and development advantages over time.

The primary contest is therefore proprietary integration against already profitable scale. RoboSense wants Peacock to improve product differentiation and component economics. Hesai can respond through its own chip development, customer reach, capacity, and existing operating leverage.

Neither side wins based on a single specification. Robot manufacturers evaluate detection quality, environmental reliability, unit consistency, software integration, delivery security, and total system cost. A sensor that performs well in one category may be unsuitable for another.

Customer concentration also shapes the contest. Large programs can produce rapid shipment growth, but they create exposure when one robot model misses its sales target. A diverse base of medium-sized programs can be steadier, though harder to support efficiently.

RoboSense says it served more than 3,400 robotics customers when it introduced Peacock. That number signals a broad funnel, but it does not reveal how many customers produce commercial robots at scale. Development kits and evaluation units carry different economic value from recurring production orders.

Its automotive base remains relevant even though E2 targets robotics. By late August, RoboSense had accumulated mass-production design wins covering 194 vehicle models. Automotive programs support manufacturing discipline and purchasing scale that can transfer to robot sensors.

Hesai has strong automotive reach of its own. It reported 839,345 ADAS lidar shipments during the first half. Its larger total output gives it more purchasing leverage and production experience across high-volume programs.

This pressure prevents Peacock from being judged in isolation. RoboSense must show why its integrated SPAD design produces a result customers cannot obtain more cheaply elsewhere. That result might be better near-field data, a smaller enclosure, lower power, or reduced system complexity.

The company has published favorable specifications, but customers will determine which advantages matter. An impressive field of view has little value if glare or multipath reflections reduce usable data. A dense point cloud can become a burden if a robot’s processor cannot consume it efficiently.

Peacock’s first production programs will reveal the balance. Strong repeat orders would support RoboSense’s custom-silicon strategy. Slow adoption would suggest that established sensors remain sufficient for many robotic tasks.

Production Does Not Yet Prove Better Economics

The main uncertainty is whether volume shipments can reverse the margin pressure visible in RoboSense’s first-half financial results.

RoboSense generated RMB1.02 billion in revenue during the first half of 2026, up 30.2 percent year over year. Gross profit rose 9.4 percent to RMB222.2 million.

However, overall gross margin fell from 25.9 percent to 21.8 percent. The company attributed the decline mainly to lower average selling prices and higher raw-material procurement costs.

Its net loss widened to RMB159.9 million from RMB148.6 million. Revenue and shipments increased, but the company did not yet convert that scale into profitability. That gap is the central risk behind the upbeat Peacock narrative.

Custom chips can eventually reduce dependence on externally sourced components. They can also eliminate redundant processing and simplify a bill of materials. Neither benefit appears instantly when a new chip reaches production.

Early production frequently carries lower yields and higher testing costs. Packaging and calibration processes need refinement. Manufacturers may also hold additional inventory to protect customer launches from unexpected quality problems.

RoboSense faces those normal ramp costs while lidar selling prices remain competitive. Buyers of lawn mowers and consumer-oriented robots cannot absorb automotive sensor economics. Suppliers must reduce cost without weakening range, reliability, or environmental tolerance.

The company also cited supply-chain pressure linked to broad demand for AI-related components. Strong demand can tighten access to wafers, memory, and other electronics. An internally designed chip still depends on external fabrication, packaging, and supporting components.

DBS analysts described the proprietary SPAD roadmap as supportive of longer-term margins. However, their margin analysis remained cautious about near-term automotive profitability. External chips, ramp costs, and market-share expansion continued to weigh on that business.

Peacock targets robotics rather than RoboSense’s core long-range automotive segment. That distinction can help because the company says robotics carries a stronger gross-margin profile. It also means E2 alone cannot repair every source of margin pressure.

Phoenix is the more direct test for automotive economics. RoboSense expects the first Phoenix-powered vehicle program to begin production during the fourth quarter. Its results will show whether the EOCENE architecture can support both robotics and automotive volume.

E2 still provides an earlier production laboratory. The company can refine detector yields, packaging, firmware, calibration, and testing before a wider family of chips scales. Improvements learned through Peacock can reduce execution risk across the architecture.

Yet investors and customers need evidence beyond a production declaration. RoboSense has not published Peacock wafer yields, unit manufacturing costs, field return rates, or customer retention data. It has also not identified the buyers responsible for the projected second-half volume.

The 200,000-unit forecast should therefore be treated as management guidance. It is specific enough to evaluate after year-end, but it has not been independently confirmed by customer disclosures.

Performance claims require similar caution. RoboSense says E2 triples precision and point rate compared with its previous generation. The comparison comes from the company, and the operating conditions behind it are not fully described publicly.

Independent testing should examine bright sunlight, dark surfaces, reflective materials, rain, dust, vibration, and overlapping lidar signals. These conditions can expose differences that do not appear in a controlled exhibition.

Software integration presents another risk. Denser data can improve perception, but developers must update drivers, calibration procedures, and inference pipelines. Customers may favor a familiar sensor if switching costs outweigh the expected benefit.

This does not invalidate the production milestone. It defines the evidence still missing. Peacock has passed the first test by entering a shipping product. It has not yet passed the tests of field reliability, repeat demand, and profitable scale.

Three Signals Will Decide Whether Peacock Matters

The next three checkpoints are E2 delivery volume, customer validation, and evidence that proprietary silicon improves company economics.

The first signal is RoboSense’s year-end shipment disclosure. The company projects more than 200,000 Peacock-based product deliveries during the second half of 2026. Meeting that target would support its claim that E2 has moved beyond pilot production.

The composition of those deliveries matters. Units shipped into several independent commercial programs would provide stronger evidence than one concentrated launch. Repeat orders would be more meaningful than inventory delivered ahead of uncertain robot demand.

A miss would not necessarily indicate a technical failure. Customer schedules, production yields, and broader robotics demand can all delay a ramp. However, a large unexplained shortfall would weaken the argument that Peacock is ready for broad commercial adoption.

The second signal is customer evidence. RoboSense has described deliveries but has not publicly named the main E2 buyers. Announcements from robot manufacturers would verify deployments and reveal which applications value the sensor’s wide view.

Field reports should address more than mapping quality. Buyers will watch reliability, sunlight performance, contamination tolerance, power use, calibration stability, and software integration. These measures decide whether a sensor survives beyond an initial design win.

Named customer programs would also clarify demand quality. A production robot with recurring sales provides a stronger foundation than demonstrations or research platforms. Multiple customers across lawn care, humanoids, and industrial systems would reduce concentration risk.

The third signal is financial. Peacock should eventually help RoboSense protect gross margin by replacing externally sourced functions and simplifying sensor construction. That thesis needs to appear in reported results rather than management presentations.

Investors should watch product gross margin, average selling-price changes, inventory growth, and research spending. Improving margin alongside rising Peacock shipments would strengthen the vertical-integration case. Falling margin would suggest that competition or ramp costs are absorbing the savings.

Phoenix adds another part of this signal. Its planned automotive production start during the fourth quarter will test whether the same architecture transfers into a stricter safety and reliability environment. A successful launch would make EOCENE a platform rather than a single-product project.

Readers following technology news should also separate chip capability from robot intelligence. Peacock supplies spatial measurements. It does not choose actions, plan tasks, or guarantee safe behavior. Those outcomes depend on perception software, models, controls, and the complete machine.

For developers, the practical question is whether E2 produces cleaner usable data within a manageable power and compute budget. For enterprise buyers, the important questions concern reliability, supplier capacity, support, and long-term availability.

Knowledge workers tracking this market need a disciplined way to connect specifications, filings, customer announcements, and later results. A searchable AI knowledge base can preserve those claims and show when the evidence changes.

Peacock has already cleared a meaningful boundary. RoboSense designed the SPAD-SoC, placed it inside E2, started volume shipments, and attached a six-figure delivery projection. Few semiconductor announcements reach that stage so quickly.

The harder judgment comes next. Does custom silicon give RoboSense a durable advantage, or does it merely help the company keep pace in an aggressive lidar market? The answer will emerge through delivered units, named deployments, and margins.

Watch those three signals through the end of 2026. If shipments exceed guidance, customers confirm deployments, and gross margin stabilizes, Peacock will become consequential technology news. If those indicators diverge, its production launch will remain an engineering milestone awaiting commercial proof.