Sensing, Vision, And Perception Stack covers the hardware and software layer that turns the physical world into machine-readable inputs for robots, automated factories, vehicles, inspection systems, subsea platforms, and defense systems. Cameras, machine-vision systems, lidar, radar, sonar, tactile sensors, environmental sensors, image processors, and edge-AI vision chips capture light, distance, motion, vibration, touch, heat, gas, or acoustic signals; perception software then classifies objects, defects, hazards, positions, and operating states before the robot command layer acts. The current basket read is that CGNX has the cleanest machine-vision economics, HSAI and OUST have the clearest lidar shipment/revenue evidence, TDY and CODA show scaled or profitable high-reliability sensing economics, AMBA is a perception-compute overlap, and the lower-ranked rows remain program-conversion or funding tests.
What the stack is: sensors, optics, image sensors, lidar transmit/receive modules, radar chipsets, sonar arrays, tactile/force sensors, gas sensors, embedded processors, perception software, calibration tooling, and integration services that let machines detect and measure the operating environment.
What it does in robotics: the stack finds parts, reads barcodes, inspects defects, maps distance, detects obstacles, tracks motion, monitors asset condition, guides autonomous movement, and verifies whether a robot or automated cell can act safely.
Main operating pieces: cameras and lenses, machine-vision controllers, lidar sensors, radar processors, sonar heads, edge-AI systems-on-chip, sensor fusion software, factory or field calibration, mounting hardware, and customer-specific non-recurring engineering.
Where it sits: inside robot heads, vehicle roofs or bumpers, factory inspection stations, warehouse portals, subsea equipment, defense platforms, smart infrastructure, and edge compute modules close to the physical task.
How robotics uses it: robot buyers pay for perception when the sensor package reduces labor, improves quality, raises uptime, enables autonomy, or passes safety requirements; suppliers get paid through product shipments, software or NRE attach, repeat orders, backlog conversion, gross margin, and cash flow.
Terms used later: lidar means laser-based distance sensing; ADAS means advanced driver-assistance systems; NRE means customer-funded development work before production revenue; backlog means contracted or expected orders that still need delivery; product gross margin tests whether shipped hardware is profitable before operating expenses.
Report boundary: this node is the robotics route for sensing, vision, and perception economics.
The broader physical-AI and defense framing stays in
Physical AI And Defense Autonomy.
Price and volume context comes from read-only discovery daily_ohlc. Static setup thresholds in this
page are still the 2026-06-05 package; representative discovery coverage now has daily rows through 2026-06-12,
so setup levels must be refreshed before use as current trading evidence.
Current Setup
Perception conversion gateSensor proof must show up as shipments, margin, and cash.
Robotics demand matters only when the sensing layer moves from demos, design wins, and pilots into paid units, repeat orders, product gross margin, and lower reliance on equity financing.
Positive proofCGNX, HSAI, OUST, TDY, and CODA have the clearest current evidence.
They show machine-vision revenue, lidar shipments, high-reliability imaging, or profitable niche sonar economics.
Conversion gateQ2 evidence must connect demand to recognized revenue.
Watch orders, shipments, backlog delivery, product margin, operating cash flow, and share count.
Primary constraintSmall lidar and sensing rows still need funding discipline.
NRE delays, gross losses, customer concentration, ATM use, shelf capacity, and reverse-split risk can dilute the stack read.
Company proofRevenue, margin, cash, lower dilution
Robotics deployment fails if the system cannot see the work area, locate itself, inspect output, identify hazards, or verify a safe action. This node matters because machine vision, lidar, radar, sonar, tactile sensors, environmental sensors, and edge perception sit before the robot command layer. The business mechanism is direct: sensor validation has to become shipped hardware, software or NRE attach, repeat orders, gross margin, cash conversion, and customer diversification.
The strongest rows have maintained knowledge lanes and current official-source support. CGNX gives the basket a scaled machine-vision and factory-inspection case. HSAI reported Q1 2026 revenue of RMB680.6M, 471,723 lidar shipments, 118,282 robotics lidar shipments, 39.1% gross margin, and GAAP net income. OUST reported Q1 revenue of about $48.6M, product revenue of about $48.2M, more than 12,600 sensors shipped, and GAAP gross margin near 42.9%. TDY adds mature high-reliability imaging, infrared, marine, unmanned-system, and defense-electronics economics. CODA adds profitable subsea and defense perception with high Marine Technology margins. The next positive proof is Q2/Q3 orders, shipments, revenue, margin, and cash flow improving together across more than one row.
The lower half of the basket is still a conversion and financing test. AEVA, INVZ, ARBE, MVIS, LIDR, ODYS, MSAI, and LINK can have real sensing technology, but the current evidence test is recognized revenue, positive or improving gross margin, customer breadth, cash-burn reduction, and controlled share count. HSAI also carries China, ADR, receivable, ASP, and procurement-policy caveats. AMBA is included for edge-vision compute, but broad semiconductor, security-camera, IoT, and automotive demand can dominate robotics attribution. Watch for NRE slips, pilots that do not become orders, backlog without delivery terms, gross losses, ATM or shelf usage, convertible financing, inventory build, or trading moves that cannot be tied to operating evidence.
Basket
This basket is hand-curated from the robotics parent, the expanded physical-AI sensing list, maintained knowledge lanes, official Q1 or fiscal Q1 2026 releases, read-only discovery coverage, and bounded earnings-date checks on 2026-06-13. Ranking uses source-backed node economics before chart timing. Core rows show scaled revenue or direct shipment evidence; option rows have strong exposure but narrower scope or attribution caveats; watch rows require revenue, margin, backlog, customer, or funding proof before they can move higher.
Scaled machine-vision supplier with the cleanest direct factory inspection and guidance economics in the node.
Market cap$10.9B
Next earningsJul 29 est.
Latest qtr revenue$268.4M
Role in stack
Cognex sells machine-vision systems for inspection, identification, measurement, barcode reading, and robot or line guidance in factories, logistics sites, electronics plants, and automotive facilities.
Revenue mix
One operating segment limits robot-only attribution. The node uses CGNX as factory/logistics machine-vision exposure where automation capex becomes systems revenue, gross margin, and cash flow.
Latest qtr revenue
Q1 2026 revenue was $268.4M, up 24.3% year over year, from the cited Q1 release and MarketBeat/Fiscal.ai earnings page checked 2026-06-13.
Scaled lidar shipment row with the clearest disclosed robotics lidar unit volume.
Market cap$3.4B
Next earningsAug 13 est.
Latest qtr revenueRMB680.6M
Role in stack
Hesai supplies lidar sensors for ADAS and robotics customers. The economics convert through shipped sensors, robotics mix, product gross margin, receivable collection, and customer breadth.
Revenue mix
Q1 shipments were 471,723 total lidar units, including 353,441 ADAS units and 118,282 robotics units. ADAS scale supports cost absorption but adds ASP and China policy caveats.
Latest qtr revenue
Q1 2026 net revenue was RMB680.6M from the company Q1 release and local HSAI lane; MarketBeat showed $98.71M reported revenue for the same quarter.
U.S. lidar and perception platform with visible product revenue, sensor shipments, and software optionality.
Market cap$1.6B
Next earningsAug 6 est.
Latest qtr revenue$48.6M
Role in stack
Ouster sells digital lidar, cameras, AI compute, and perception software into robotics, industrial automation, smart infrastructure, automotive, and defense/security deployments.
Revenue mix
Q1 was product-led: about $48.2M of $48.6M revenue was product revenue. Stereolabs and REV8 add platform breadth, but vertical economics are still not separately disclosed.
Latest qtr revenue
Q1 2026 revenue was about $48.6M, with more than 12,600 sensors shipped and GAAP gross margin near 42.9%, from the OUST lane and Q1 release.
Diversified high-reliability imaging, instrumentation, marine, unmanned-system, and defense-electronics anchor.
Market cap$28.8B
Next earningsJul 22 est.
Latest qtr revenue$1.56B
Role in stack
Teledyne supplies Digital Imaging, infrared, X-ray, marine instrumentation, unmanned systems, aerospace/defense electronics, and engineered sensing systems.
Revenue mix
Digital Imaging and Aerospace and Defense Electronics are the most relevant sensing channels, while Instrumentation and Engineered Systems make TDY a diversified industrial technology row.
Latest qtr revenue
Q1 2026 net sales were about $1.560B from Teledyne's Q1 release and local TDY lane.
Edge-AI vision compute overlap where robotics relevance must be separated from security, auto, IoT, and semiconductor cycles.
Market cap$3.3B
Next earningsAug 27 est.
Latest qtr revenue$100.4M
Role in stack
Ambarella sells low-power edge-AI vision SoCs for cameras, physical security, automotive, robotics, drones, and edge infrastructure.
Revenue mix
IoT and automotive dominate disclosed end-market commentary; robotics is an application inside the edge-AI product cycle rather than a separately reported segment.
Latest qtr revenue
Q1 FY2027 revenue was $100.4M with GAAP gross margin of 58.4%, from the AMBA lane and company Q1 FY2027 release.
FMCW 4D lidar option with named customer validation but pre-scale product economics.
Market cap$855.3M
Next earningsNot confirmed
Latest qtr revenue$6.26M
Role in stack
Aeva supplies FMCW 4D lidar-on-chip sensing and perception for automotive, industrial precision, defense, infrastructure, and physical-AI deployments.
Revenue mix
Q1 revenue was still development-heavy: product revenue was $2.43M and professional services/NRE was $3.84M. Product mix and production milestones are the conversion gate.
Latest qtr revenue
Q1 2026 revenue was $6.262M from the AEVA lane and company Q1 release; current next earnings date was not confirmed in the bounded check.
Automotive and Physical AI lidar program row that needs NRE catch-up and positive margin evidence.
Market cap$156.9M
Next earningsAug 12 est.
Latest qtr revenue$7.13M
Role in stack
Innoviz supplies automotive-grade lidar hardware and perception software for OEM, trucking, defense/security, and non-auto Physical AI programs.
Revenue mix
Revenue depends on NRE, lidar unit revenue, start-of-production volume, and customer program cadence. Q1 weakness came from NRE milestone timing and negative gross margin.
Latest qtr revenue
Q1 2026 revenue was $7.13M from the INVZ lane and Q1 release; management retained FY2026 revenue guidance of $67M-$73M.
Tactile, force, HMI, printed-electronics, and environmental-sensor row with small-scale customer concentration.
Market cap$56.1M
Next earningsAug 6 est.
Latest qtr revenue$3.07M
Role in stack
Interlink provides force/touch sensors, HMI, printed electronics, membrane keypads, and gas/environmental sensors for medical, industrial, automotive, IoT, and specialty customers.
Revenue mix
Q1 2026 revenue mix included medical, industrial, automotive, and standard products, but one-segment reporting and customer concentration limit product-family economics.
Latest qtr revenue
Q1 2026 revenue was $3.074M with gross margin of 43.5%, from the LINK lane and Q1 release.
High-resolution radar-perception option with program evidence but tiny current revenue.
Market cap$98.2M
Next earningsAug 12 est.
Latest qtr revenue$0.46M
Role in stack
Arbe sells high-resolution radar chipsets and Phoenix radar systems for L4, robotaxi, robotruck, defense/security, perimeter, and physical-AI applications.
Revenue mix
Current sales are pre-scale. The guide depends on backlog, Hirain/China L4, robotaxi, security, physical AI, and complete radar-system conversion.
Latest qtr revenue
Q1 2026 revenue was $0.461M and backlog was about $1.0M, from the ARBE lane and Q1 release.
Condition-intelligence row where software growth and deployments need to outrun burn and dilution.
Market cap$12.0M
Next earningsAug 12 est.
Latest qtr revenue$1.61M
Role in stack
MSAI sells multi-sensor monitoring systems and MSAI Connect software for industrial reliability, data centers, distribution facilities, manufacturing, and early threat detection.
Revenue mix
Q1 revenue included hardware, software, and service. Software grew quickly from a small base, but three customers represented 77% of Q1 revenue.
Latest qtr revenue
Q1 2026 revenue was $1.614M, with software revenue of $0.675M and gross margin near 57%, from the MSAI lane and Q1 release.
Visual predictive-maintenance row with defense/aerospace validation but almost no current recognized revenue.
Market cap$74.8M
Next earningsNot confirmed
Latest qtr revenue$0.08M
Role in stack
Odysight sells AI-powered visual sensing and predictive-maintenance systems for aerospace, defense, transportation, and industrial assets.
Revenue mix
Revenue is pre-scale and backlog-led. Defense pilots, the U.S. Navy CRADA, GACI channel work, and aerospace validation matter only when they convert to funded revenue.
Latest qtr revenue
Q1 2026 revenue was $82 thousand against about $14.0M of backlog, from the ODYS lane and Q1 filing; no credible next earnings date was found in the bounded check.
Lidar turnaround row after Luminar and Scantinel assets broadened the portfolio.
Market cap$218.6M
Next earningsAug 12 est.
Latest qtr revenue$0.94M
Role in stack
MicroVision sells lidar and perception products for industrial, security/defense, automotive, and physical-AI bridge markets after adding Luminar and Scantinel assets.
Revenue mix
The acquired portfolio adds short-, mid-, long-, and ultra-long-range lidar plus software, but revenue remains small and back-half weighted versus cash use.
Latest qtr revenue
Q1 2026 revenue was about $0.94M, from the MVIS lane and Q1 release; management maintained a $10M-$15M 2026 revenue guide.
Early-stage software-defined lidar pipeline row with liquidity but the thinnest current revenue conversion.
Market cap$90.2M
Next earningsJul 30 est.
Latest qtr revenue$0.10M
Role in stack
AEye supplies software-defined long-range lidar for auto, trucking, defense, rail, intelligent transportation systems, and infrastructure channels.
Revenue mix
Current revenue is product-shipment revenue from a tiny base. The pipeline mentions 21 revenue-generating shipment customers, but repeat orders and positive unit economics are still unproven.
Latest qtr revenue
Q1 2026 revenue was $101 thousand from the LIDR lane and Q1 release; liquidity supports runway but not product-market proof.
Market caps use read-only discovery instruments.market_cap values queried on 2026-06-13. Next-earnings dates are MarketBeat/Fiscal.ai estimates checked on 2026-06-13 unless labeled Not confirmed. MarketBeat pages state the listed estimates are based on prior reporting schedules; AEVA and ODYS did not show a current credible next-date estimate in the bounded check.
What Confirms Or Weakens
AreaWhat confirmsWhat weakens or invalidatesWatch next
01Node thesisPerception becomes paid deployment
What confirms
Sensor shipments, machine-vision orders, subsea/defense sensing awards, edge-vision silicon revenue, and backlog conversion improve with gross margin and cash flow.
What weakens or invalidates
Design wins, pilots, CRADAs, and customer references do not become recognized revenue, repeat orders, product margin, or operating cash-flow improvement.
Watch next
Q2/Q3 revenue
Orders and shipments
Product margin
Share count
02Economics mechanismHardware margin and cash
What confirms
CGNX, HSAI, OUST, TDY, CODA, and AMBA show that revenue growth carries high gross margin, adjusted EBITDA or operating income, and cash conversion.
What weakens or invalidates
Growth is NRE-heavy, gross margin turns negative, receivables stretch, inventory builds, or revenue remains too small to absorb engineering and public-company costs.
Watch next
CGNX gross margin
HSAI receivables
OUST EBITDA
AMBA inventory
03Customer conversionPrograms become repeat orders
What confirms
ADAS and robotics lidar shipments, factory vision orders, DAVD and defense awards, data-center monitoring pilots, and industrial sensor projects move into repeat paid deployments.
What weakens or invalidates
NRE slips, pilots stay unpaid, backlog terms remain thin, customer concentration rises, or customer references do not specify purchase orders, delivery timing, or revenue value.
Watch next
HSAI robotics mix
AEVA programs
INVZ NRE
ODYS backlog
04Funding and dilutionRunway supports per-share value
What confirms
Cash burn declines as revenue rises, ATM usage remains limited, share count is controlled, and financing instruments do not outrank common shareholders before product proof improves.
What weakens or invalidates
ATM, shelf, convert, reverse split, preferred financing, or warrant activity becomes the main way to fund operations while revenue stays subscale.
Watch next
AEVA facility use
INVZ ATM
MSAI ATM
LIDR shelf
05Policy and supplyAccess and cost stay manageable
What confirms
Export-control, China, Section 1260H, tariff, component, and advanced-node constraints remain bounded enough that customers keep ordering and margins hold.
What weakens or invalidates
Procurement restrictions, customer trust issues, component cost pressure, advanced-node costs, or trade controls delay programs or reduce gross margin.
Watch next
HSAI policy status
TDY export controls
AMBA Asia exposure
Lidar components
06Operating constraintDeployment works in real sites
What confirms
Sensors work in factories, warehouses, vehicles, subsea environments, defense platforms, and data centers with acceptable reliability, calibration cost, and integration time.
What weakens or invalidates
Field failures, excessive customization, long validation loops, warranty issues, or deployment delays prevent repeat orders and margin absorption.
Watch next
Customer acceptance
Repeat orders
Warranty cost
Integration cycle
07Stale conditionRefresh trigger
What confirms
Knowledge lanes, official releases, earnings-date checks, and discovery rows remain current enough to support the ranking and setup labels.
What weakens or invalidates
A new earnings release, filing, financing, policy action, order, cancellation, or trading week arrives before the page is refreshed.
MarketBeat/Fiscal.ai earnings pages were checked on 2026-06-13 for CGNX, HSAI, OUST, TDY, CODA, AMBA, INVZ, LINK, ARBE, MSAI, MVIS, and LIDR. Listed next dates were treated as estimates because the pages describe them as based on past reporting schedules.
Estimated next earnings used in cards: CGNX Jul 29, 2026; HSAI Aug 13, 2026; OUST Aug 6, 2026; TDY Jul 22, 2026; CODA Jun 15, 2026; AMBA Aug 27, 2026; INVZ Aug 12, 2026; LINK Aug 6, 2026; ARBE Aug 12, 2026; MSAI Aug 12, 2026; MVIS Aug 12, 2026; LIDR Jul 30, 2026.
AEVA and ODYS are labeled Not confirmed. AEVA's MarketBeat page was stale to Q3 2025 and did not provide a current next date; ODYS showed recent history but no credible current next-date estimate in the bounded check.
Latest-quarter revenue sources are the linked security lanes plus official releases already cited in node report.json: CGNX Q1 2026 $268.4M; HSAI Q1 2026 RMB680.6M; OUST Q1 2026 about $48.6M; TDY Q1 2026 $1.560B; CODA fiscal Q1 2026 $6.71M; AMBA Q1 FY2027 $100.4M; AEVA Q1 2026 $6.262M; INVZ Q1 2026 $7.13M; LINK Q1 2026 $3.074M; ARBE Q1 2026 $0.461M; MSAI Q1 2026 $1.614M; ODYS Q1 2026 $82 thousand; MVIS Q1 2026 about $0.94M; LIDR Q1 2026 $101 thousand.
Discovery And Chart Provenance
Read-only DuckDB checks on 2026-06-13 queried daily_ohlc coverage and instruments.market_cap for all 14 basket tickers. All 14 had daily rows through 2026-06-12.
The selected-security right rail uses /api/securities/{ticker}/chart?frequency=weekly&window=3y&as_of=latest. Static HTML renders without the API; live charts require the local report API.
Chart assumptions: source table daily_ohlc; weekly bars use first open, maximum high, minimum low, final close, and summed volume; visible horizon is three years; indicators are 20-week EMA, 100-week EMA where history allows, weekly volume, and 20-week average volume.
Static setup thresholds and chart-artifact values in the sidecar metadata remain from the 2026-06-05 node package. Discovery has newer daily rows through 2026-06-12, so the static setup thresholds are stale and must be refreshed before they are used as current trading evidence.
Known Gaps
The page summarizes and routes to knowledge lanes; it does not replace canonical company, sector, or theme research.
Robotics revenue is not isolated for CGNX, TDY, AMBA, LINK, MSAI, or several lidar rows. Their inclusion is based on product role, official filings, and knowledge-lane evidence rather than audited robot-only segment revenue.
Market caps are local discovery values, not live quote-page values. Earnings dates are estimates except where future company confirmation later appears.
Review validation on 2026-06-13 started the local API process, but a separate loopback curl to 127.0.0.1:8765 still returned connection refused in the sandbox. Static API attributes, chart route metadata, chart tests, and absence of embedded OHLC payloads were verified instead.