Today on The Robot Beat: a Wall Street Journal investigation reveals the hidden Chinese hardware backbone inside American humanoid robots — and three bills in Congress are now targeting it. AgiBot doubles its production milestone to 10,000 units in just three months, Eclipse VC commits $1.3 billion to physical AI, and Waymo opens its 11th robotaxi city with a new Lyft partnership model. Plus — Jeff Bezos's stealth lab is poaching top AI talent for embodied intelligence, and AgiBot's new open-source datasets are lowering the barrier for every robotics researcher.
A WSJ investigation quantifies what the American Security Robotics Act (covered since March 26) is actually targeting: China controls an estimated 55% of humanoid robot costs through the motion-control layer, with U.S. leaders including Tesla Optimus and Figure AI heavily dependent on Chinese motors, sensors, and precision components. Two additional bills — the Humanoid ROBOT Act and National Commission on Robotics Act — are now advancing alongside the procurement ban, and Morgan Stanley estimates Chinese supply chains could reduce production costs by two-thirds. Rare Earth Exchanges confirms the split: U.S. leads AI and semiconductors, China dominates the physical embodiment layer.
Why it matters
Prior coverage established the legislation; this WSJ investigation provides the specific cost-structure data that explains why it's so difficult to act on. The 55% Chinese cost share in the motion-control layer — not just rare earths — means compliance would restructure pricing across every U.S. humanoid program. The DJI precedent is directly applicable: that ban preceded viable domestic alternatives by years, and the humanoid version involves far larger capital commitments. The three-bill simultaneity suggests restrictions are moving faster than domestic capacity can respond.
Industry observers note even Korean suppliers (LG, Hyundai) are years from volume production of harmonic reducers and rare-earth permanent magnets. Chinese analysts see deepening leverage, not threat. Robotics startup founders privately worry compliance costs will disproportionately burden smaller companies relative to Tesla and Figure.
AgiBot doubled its cumulative output from 5,000 to 10,000 units in just three months, reaching the milestone by late March 2026. Digitimes maps the resulting three-player structure now defining China's competitive landscape: AgiBot (supply-chain integration, industrial focus), Unitree (low-cost platform ecosystem — pending its April 27 Shanghai IPO), and UBTech (ecosystem deployment across major OEMs including the Honda partnership reported yesterday). China accounted for roughly 90% of global humanoid shipments in 2025.
Why it matters
The doubling from 5,000 to 10,000 units in one quarter — not one year — is the signal. Learning-curve and sourcing advantages are now compounding at a pace Western competitors hadn't modeled. The three-player market structure maps to classic industry consolidation, and the open question is whether Western humanoid companies can compete on intelligence and specialization before volume-driven cost advantages extend across all segments.
Guojin Securities flags August 2026 as Tesla's expected large-scale Optimus production commencement. South Korea's KIMM formally designated 2026 as 'Year One of humanoid commercialization,' projecting cost reductions from $35K to $13-17K within five years.
UniX AI announced global deliveries of its Panther home humanoid robot beginning April 8, 2026. Building on the initial deployment covered in earlier briefings, this update provides new production data: the company has achieved 100+ monthly deliveries since 2025 and is targeting 1,000 units per month. The wheeled dual-arm robot features an omnidirectional 4WS+4WD chassis, 8-DoF bionic arms capable of lifting 26 lbs, and is designed for household, commercial service, and industrial scenarios. UniX AI's differentiated architecture prioritizes practical deployment over humanoid aesthetic purity.
Why it matters
The production ramp from 100 to a targeted 1,000 monthly units represents the first serious attempt to scale a multi-purpose home humanoid beyond pilot quantities. UniX AI's deliberate choice of a wheeled base over bipedal locomotion reflects a pragmatic design philosophy — wheels are more reliable, cheaper, and quieter than legs for indoor environments. This positions the Panther as the first genuine test of consumer demand for a general-purpose household robot, distinct from single-function devices like robot vacuums. The production scaling data provides a concrete benchmark for the home humanoid category.
Multiple outlets (PRNewswire, Fox News, Fakta.co) covered the announcement with consistent specs, validating the claims. Reviewers note the 170+ lb weight could limit adoption in multi-story homes without elevators. The wheeled design trades stair-climbing capability for reliability and lower cost — a trade-off that may prove commercially decisive in the near term.
Building on known details about the Giga Texas humanoid facility and Fremont R&D expansion, three new data points converge: Tesla program lead Konstantinos Laskaris revealed a Gen 3 silhouette at ETH Robotics Club showing thicker forearms and smoother design; Tesla AI posted a Weibo teaser of significantly improved hands with better finger articulation; and the Q1 2026 public unveiling has been delayed pending refinements, though the robot is already walking internally.
Why it matters
The hand improvement is the most technically significant update — manipulation dexterity has been the single largest constraint on humanoid utility. The Q1 delay is a departure from Musk's typical pace and suggests the engineering challenges are harder than originally projected. Combined with the Fremont lease expansion (375K+ sq ft), the picture is dual-site production infrastructure being built now alongside deliberate quality gating on the public rollout.
Some roboticists view the delay as evidence that bipedal locomotion and manipulation are harder than Tesla initially projected.
Researchers at Shanghai Engineering Research Center unveiled Moya, a humanoid robot that replaces traditional motors and rigid joints entirely with pneumatic artificial muscles and a flexible spine-like skeletal structure, mimicking human musculoskeletal anatomy. Unlike yesterday's ASU HARP actuators — which addressed soft robotics at the component level — Moya integrates the concept into a complete humanoid system, with embodied AI continuously modeling muscle deformation to adjust posture and force output in real time.
Why it matters
Every major humanoid currently in production — Boston Dynamics, Tesla, Figure, and the Chinese manufacturers — is built on rigid actuator-joint architectures. Moya's full-body artificial muscle approach targets compliance, impact resilience, and adaptive force control that rigid systems handle poorly. If it scales, it could prove decisive for deployment in unstructured environments like homes and disaster zones where rigid robots struggle.
Robotics researchers note well-documented challenges with pneumatic artificial muscles: precision control, energy efficiency, and operating pressure. The Shanghai team claims their embodied AI overcomes precision limitations via continuous deformation modeling. Skeptics argue the lab-to-deployment gap for soft-body robotics remains large; proponents counter that biology proves musculoskeletal actuation works at human scale.
A detailed Automotive Manufacturing Solutions analysis provides the most granular production data yet on humanoids in manufacturing: Figure 02 has handled over 90,000 components in ten-hour shifts at BMW Spartanburg over ten months, with expansion to Leipzig planned. Toyota and Tesla have also launched production deployments. However, two critical unresolved barriers — dexterity and safety — currently limit humanoids to specific retrofit applications. Emergency stop protocol concerns specifically prevent close human-robot collaboration.
Why it matters
This is the clearest production-data reality check yet, relevant context for the Figure AI safety lawsuit (ongoing from prior coverage) and the BMW deployment mentioned in the Q1 funding record. The gap between 'humanoid working in a factory' and 'humanoid replacing a factory worker' remains substantial — deployments are narrowly scoped retrofits, not general automation. The specific identification of emergency stop protocols as the blocking safety issue adds a concrete regulatory dimension to the broader robotics_safety_ops thread.
Safety engineers note the emergency stop problem for a 150+ lb humanoid in motion remains unresolved. Toyota's parallel deployment provides a second independent data point beyond BMW.
Following the Wired review yesterday confirming Mammotion crossed the robot mower viability threshold, the company unveiled the LUBA 3 AWD with Tri-Fusion navigation (LiDAR + Network RTK + AI camera) that eliminates boundary wire installation and roof-mounted antennas entirely. Free lifetime RTK positioning via Wi-Fi or 4G is included in North America. The LUBA Mini 2 AWD uses LiDAR alone for smaller yards.
Why it matters
Installation complexity has been the single largest adoption barrier the consumer_robotics_skepticism thread has tracked. Eliminating wire installation makes these genuinely plug-and-play. The free lifetime RTK service is a structural competitive moat against Husqvarna's subscription model, turning a hardware sale into an ongoing relationship.
Reviewers question whether camera-based obstacle detection performs reliably in low-light conditions during early morning or evening mowing.
Building on the Wisconsin ElliQ companion robot pilot reported yesterday (280 seniors, 88 daily interactions — 2x the national average), South Korea's Ministry of Health and Welfare is going further: directly subsidizing ChatGPT-enabled AI companion dolls (Hyodol) for seniors aged 65+ as a clinical health intervention. MIT Technology Review named AI companions for elderly a 2026 breakthrough technology, and U.S. HHS has launched a $2 million prize for AI caregiver tools.
Why it matters
The shift from consumer purchase to government-subsidized health intervention is the key new development — it changes the buyer from an individual to a public health system, removing price sensitivity as the primary adoption barrier. Combined with the Wisconsin data validating engagement rates, this signals a cross-national convergence toward robotics-based eldercare as public health infrastructure. The go-to-market implications for robotics founders: demonstrated health outcomes unlock government procurement channels potentially worth orders of magnitude more than consumer sales.
Critics question whether robotic companionship addresses root causes of elderly isolation. Proponents argue any intervention reducing isolation measurably improves health outcomes at a scale human caregivers cannot match.
Jeff Bezos has launched Project Prometheus, a stealth AI lab with approximately $6.2 billion in funding and roughly 100 engineers and researchers, including Kyle Kosic — co-founder of Elon Musk's xAI and former OpenAI infrastructure lead. The Financial Times confirmed the hire. Unlike general-purpose chatbot labs, Prometheus is focused specifically on applying frontier AI to engineering, manufacturing, aerospace, and physical-world systems. The lab is hiring aggressively across robotics, sim-to-real transfer, computer vision, and mechanical engineering — disciplines at the intersection of AI and physical manipulation.
Why it matters
Project Prometheus represents the highest-profile billionaire bet yet on embodied intelligence as a distinct technical frontier. The $6.2 billion war chest and aggressive talent acquisition (poaching from xAI, which itself poached from OpenAI) signals that physical AI is now competing for the same caliber of researchers who built GPT-4 and Grok. The focus on physical-world systems rather than language models suggests Bezos sees embodied AI as the next frontier after the LLM race — and that the talent market for robotics engineers is about to get significantly more competitive.
The FT reports the Kosic hire as a significant talent shift in the AI startup ecosystem, suggesting Prometheus is already pulling top-tier engineers away from Musk's ventures. Industry observers note that Bezos already controls significant physical AI infrastructure through Amazon Robotics and Blue Origin, making Prometheus potentially the connective tissue between those assets. Skeptics question whether a stealth lab can move fast enough to compete with deployed systems from companies like AgiBot and Figure AI.
AgiBot — which just shipped its 10,000th humanoid — is now also opening its data infrastructure: AGIBOT WORLD 2026 is a large-scale heterogeneous dataset combining hundreds of hours of real-world teleoperation data with 1:1 digital twin simulations, including whole-body control, force-feedback capture, and critically, error-recovery trajectories. Genie Sim 3.0 adds text-to-environment generation, RL via RLinf, and compresses environment creation from hours to minutes.
Why it matters
The inclusion of error-recovery trajectories sets this apart from most robot datasets, which only capture successful completions. By open-sourcing these tools, AgiBot simultaneously advances the broader field and positions its data formats as de facto standards — a platform play that complements its hardware volume advantage. This directly attacks the data scarcity bottleneck that Avala Research identified yesterday as the binding constraint on VLA progress.
Some researchers note that AgiBot's open-sourcing strategy also serves its commercial interests by creating ecosystem lock-in around its tools and data formats.
Eclipse, a Palo Alto-based venture firm, announced a $1.3 billion fund ($591 million for early-stage, remainder for growth) focused on backing and building physical AI startups across transportation, energy, infrastructure, and robotics. The portfolio includes Cerebras, Wayve, Bedrock Robotics, Redwood Materials, Arc, and Mind Robotics. Eclipse's explicit strategy is to build an interconnected ecosystem of complementary companies that partner with each other to leverage cross-sector data and build scale. The fund represents one of the largest dedicated capital pools for physical AI to date.
Why it matters
Eclipse's fund validates physical AI as a distinct investable category — not just 'AI applied to hardware' but an integrated thesis requiring specialized capital and portfolio construction. The ecosystem approach (investing in companies that can share data and infrastructure) mirrors the network effects that made software platforms dominant, applied to physical systems. The portfolio composition reveals where sophisticated investors see near-term commercialization: autonomous construction (Bedrock), battery recycling (Redwood), electric marine (Arc), autonomous driving (Wayve), and industrial robotics (Mind). For robotics founders, this signals both available capital and strategic expectations — investors want companies that can plug into broader physical AI infrastructure, not standalone hardware plays.
TechCrunch notes that Eclipse's approach contrasts with generalist AI funds by requiring deep operational expertise in physical industries. CB Insights' Q1 2026 data confirms the broader trend: while deal count contracted 15% to the lowest since Q4 2016, hard tech and AI infrastructure dominate top performers. The zero-shot fund ($100M) from OpenAI alumni, also announced this week, represents a complementary but smaller capital source specifically targeting manufacturing robotics.
China's Spirit AI has raised approximately $420 million across two funding rounds in just 30 days, including a $145 million round co-led by Shunwei Capital (Lei Jun/Xiaomi) and Yunfeng Fund (Jack Ma). The company develops embodied intelligence models for real-world robotic tasks and has accumulated over 200,000 hours of robot interaction data from systems deployed in retail (barista robots) and industrial (battery insertion on CATL production lines) environments. Spirit AI targets 1 million hours of training data by end of 2026.
Why it matters
The speed and scale of this raise — $420M in a month from two of China's most prominent tech investors — reflects the velocity of embodied AI commercialization in China. Spirit AI's data flywheel approach (collecting real-world operational data from deployed robots, then feeding it back into model training) mirrors the strategy that made Tesla's Autopilot data moat formidable. The 200,000+ hours of operational robot data already collected gives Spirit AI a significant head start over labs that rely primarily on simulation. The CATL battery line deployment specifically demonstrates industrial-grade precision in a sector with near-zero tolerance for error.
Chinese venture analysts note the rare joint backing from both Lei Jun and Jack Ma ecosystem funds, suggesting cross-industry conviction that embodied AI will be as transformative as mobile internet. Western competitors counter that data quantity matters less than data quality and curation (per the Avala Research analysis from April 6). The timing — right as U.S.-China supply chain tensions escalate — suggests Chinese investors are betting on domestic capability building.
D-Robotics (also known as Digua Robotics) announced completion of a $150 million Series B2 funding round, bringing total Series B financing to $270 million. Backers include Prosperity7 Ventures (Saudi Aramco's venture arm), Envision Group, and a major retail/supply chain giant. The company reports 180% year-over-year shipment growth and a developer community exceeding 100,000 members. Capital will accelerate global expansion and development of an integrated hardware-software platform with strategic partner Horizon Robotics, which provides foundational embodied AI models.
Why it matters
The $270M total Series B demonstrates that embodied AI platform companies — not just humanoid robot makers — are attracting significant capital. D-Robotics' developer-first ecosystem model (100K+ developers) suggests a platform play where the community builds applications on top of standardized hardware-software stacks. The Horizon Robotics partnership adds a foundation model layer, creating a vertically integrated offering. The involvement of Saudi Aramco's venture arm signals that sovereign wealth capital is diversifying beyond traditional AI into physical AI infrastructure.
The developer ecosystem scale (100K+ members) distinguishes D-Robotics from hardware-only competitors. The dual funding confirmation across Gasgoo Autonews and BigGo Finance validates the round. Some analysts note that the retail/supply chain investor participation suggests near-term deployment in logistics and fulfillment — a potentially large-volume market.
Intel announced a strategic partnership with Elon Musk's Terafab AI chip complex — a joint venture involving Tesla, SpaceX, and xAI — to design and manufacture processors for AI, robotics, and data center applications at the Austin, Texas facility. The project targets 1 terawatt per year of annual compute production, with one of two planned fabs explicitly dedicated to powering Tesla's autonomous vehicles and Optimus humanoid robots. Intel CEO Lip-Bu Tan said the partnership will accelerate Terafab's production timeline. The estimated investment exceeds $20 billion.
Why it matters
This partnership addresses a structural bottleneck in humanoid robotics scaling: the availability of purpose-built silicon for on-device inference and real-time control. A dedicated fab for robotics and AV processors — separate from generic data center chips — validates that the compute demands of embodied AI are large enough to justify specialized manufacturing infrastructure. Intel's involvement also signals that the Terafab project has moved beyond Musk's orbit into a genuine multi-party industrial initiative. The terawatt-scale production target implies confidence that robotics and inference workloads will sustain chip demand independent of the training GPU market.
TechWire Asia notes the vertical integration aspect — chip design, production, and packaging under one roof — as a departure from the distributed supply chains that characterize most semiconductor manufacturing. Market analysts see this as Intel's bid for relevance in the AI chip race after losing ground to NVIDIA and TSMC. Tesla investors responded positively, with shares jumping 4.3% in after-hours trading on the announcement.
Tesla released FSD v14.3 featuring a full compiler and runtime rewrite using MLIR, achieving 20% lower inference latency alongside improvements to vision encoding, parking prediction, emergency vehicle response, and edge cases. Chris Lattner — MLIR's creator and former Tesla Autopilot lead — publicly endorsed the switch.
Why it matters
The lesson for robotics: 20% latency gains from compiler-level optimization alone, without changing the neural network, means the same hardware runs faster. As the compute_platforms_robotics thread has tracked, inference efficiency — not raw training compute — is becoming the binding constraint for deployed robots. MLIR is open-source and cross-platform, so these techniques are available to any robotics developer, not just Tesla.
Some autonomous driving engineers question whether 20% latency improvement translates to meaningful safety gains in real-world scenarios.
Cisco surveyed 1,000+ OT decision-makers across 19 countries and 21 sectors: 61% now deploy AI in live physical operations, but network infrastructure readiness, cybersecurity, and IT/OT collaboration determine whether deployments scale. 97% expect AI workloads to impact network requirements; 40% cite cybersecurity as the biggest obstacle. A complementary Deloitte survey finds robots still operate only 30-90 minutes before requiring intervention — far below the 95-99% reliability industrial environments demand.
Why it matters
The South Korean '99% problem' covered yesterday identified edge cases as the algorithmic gap; this survey identifies infrastructure as the equally binding deployment gap. The 61% adoption rate confirms physical AI is mainstream, but the 30-90 minute intervention window directly quantifies why the industrial_vla_deployment thread's deployments remain bounded. For robotics developers: network and security environment design is as critical as model performance.
The International Federation of Robotics released its World Robotics 2025 report showing record robot density globally: Western Europe leads at 267 robots per 10,000 employees, followed by North America at 204 and Asia at 131. China achieved a world record operational stock of 2 million industrial robots, representing 54% of global installations in 2024. The data confirms that automation is accelerating across all major manufacturing regions, with China's installation rate outpacing all competitors combined.
Why it matters
The IFR data provides the authoritative baseline for understanding where industrial automation stands heading into the humanoid era. China's 2 million operational robot stock and 54% global installation share quantify a structural manufacturing advantage that humanoid robots will amplify, not create. The regional density gaps (Europe at 267 vs. Asia at 131 per 10K workers) suggest significant room for growth in Asian markets beyond China. For entrepreneurs, these numbers frame the total addressable market for robotics components, software, and services — and explain why Chinese manufacturers can iterate faster on humanoid production.
IFR notes that density varies dramatically by industry — automotive and electronics lead — suggesting that humanoid robots will follow similar adoption curves into high-value, high-automation sectors first. Some economists argue that robot density correlates with manufacturing competitiveness, not job displacement, as highly automated economies tend to maintain strong employment in manufacturing-adjacent services.
America's largest shipbuilder Huntington Ingalls Industries signed an MOU with GrayMatter Robotics to automate surface preparation, coating, and inspection in shipbuilding. GrayMatter's systems have processed over 30 million square feet of surface area, delivering up to 12x throughput versus skilled labor and 95% rework reduction. HII targets a 15% throughput increase in 2026.
Why it matters
Shipbuilding is among the most complex, high-variability manufacturing environments — unique geometries, safety-critical coatings, multi-decade asset lifecycles. GrayMatter's claimed 12x throughput in this context would demonstrate physical AI handling genuinely specification-heavy operations, not just repetitive tasks. The defense urgency is acute: the U.S. is struggling to build warships fast enough, making this a national security priority rather than a cost play.
Manufacturing Dive notes this is an MOU, not a deployment contract. GrayMatter's 30M+ sq ft track record provides evidence of operational maturity beyond typical startup claims. Splash247 emphasizes that HII builds the majority of U.S. Navy surface combatants and submarines.
Waymo launched fully autonomous ride-hailing in Nashville — its 11th city — covering a 60-square-mile area. The notable new development: Nashville partners with Lyft for fleet management and ride-hailing integration, distinct from the Uber partnerships used in other cities, signaling a shift toward a technology-supplier model. This comes the same week Waymo begins London testing and amid the still-unresolved NYC permit expiration reported yesterday.
Why it matters
The Lyft partnership structure is the new information here — it suggests Waymo is testing multiple commercialization models simultaneously rather than standardizing on a single operator relationship. The contrast with the NYC political blockage reported yesterday is sharp: Waymo's geographic expansion strategy is routing around hostile regulatory environments rather than fighting them.
The remote operator transparency issues raised by Senator Markey's investigation remain unaddressed alongside this expansion.
Also, spun off from Rivian in 2025, has partnered with DoorDash to develop purpose-built autonomous electric vehicles for last-mile delivery in dense urban environments. DoorDash invested $200 million in Also's Series C and placed co-founder Stanley Tang on the board. Also's Commercial TM-Q quadricycle is launching in 2026 and has already attracted Amazon's interest for European and U.S. deployments.
Why it matters
This partnership signals that major logistics platforms are moving from working with third-party autonomous delivery providers to co-developing purpose-built vehicles. The $200M DoorDash investment with board representation indicates deep strategic commitment, not a pilot program. The Rivian lineage suggests strong EV engineering DNA, while the quadricycle form factor targets the regulatory sweet spot between sidewalk robots and full-size autonomous vans. Amazon's reported interest for European deployments suggests the design may be particularly suited to the narrow-street environments common in European cities.
Electrive notes the quadricycle category operates under lighter regulatory requirements than full-size autonomous vehicles in many jurisdictions. The DoorDash partnership provides immediate demand and delivery data. Manna Air's separate $50M raise for drone delivery this week suggests that multiple autonomous last-mile modalities are being funded simultaneously — the market may ultimately segment by density, distance, and payload rather than converging on a single form factor.
Supply Chain Geopolitics Is the New Bottleneck for Humanoid Robots Multiple stories this cycle converge on a single structural tension: U.S. humanoid robots depend on Chinese motors, magnets, and actuators for 55%+ of cost, while Congress advances procurement bans modeled on the DJI drone playbook. Korean and Japanese suppliers (LG, Hyundai) are positioning as alternatives, but verified domestic supply chains are years away. This is the DJI drone story replaying in slow motion with higher stakes.
Chinese Humanoid Production Has Crossed the Mass-Manufacturing Threshold AgiBot hitting 10,000 units, Unitree preparing a profitable IPO, and the Foshan automated production line all confirm that Chinese manufacturers have moved decisively from prototyping to factory-scale output. South Korea's KIMM formally designating 2026 as 'Year One of Humanoid Commercialization' adds a policy dimension. Cost curves are bending faster than Western competitors anticipated.
Mega-Capital Is Concentrating Around Physical AI Infrastructure Eclipse's $1.3B fund, Spirit AI's $420M in 30 days, D-Robotics' $270M Series B, Bezos's ~$6.2B Project Prometheus, and Intel joining Terafab collectively signal that institutional capital is treating physical AI as a distinct, fundable category — not a subcategory of software AI. The capital is flowing toward integrated hardware-software platforms, not point solutions.
Open-Source Tools Are Democratizing Embodied AI Development AgiBot's open-source WORLD 2026 dataset and Genie Sim 3.0 platform, Hugging Face's LeRobot v0.5.1 with Unitree fixes, and OpenClaw's ROS2 integration are collectively lowering the barrier to entry for robotics AI research. The data bottleneck that has historically favored well-funded labs is eroding as high-quality datasets and simulation infrastructure become publicly available.
Inference Economics Are Reshaping Robot Hardware Decisions Tesla's MLIR compiler rewrite yielding 20% latency gains, the training-to-inference chip market shift, and Eclipse's bet on edge AI all point to inference optimization — not raw training compute — as the binding constraint for deployed robots. Specialized inference chips delivering 4.7x better price-performance than NVIDIA GPUs for inference workloads suggest the Jetson-for-everything assumption is under serious pressure.
What to Expect
2026-04-10—IHMC public debut of Alex humanoid robot at Pensacola open house — first live demonstration of the all-electric untethered humanoid.
2026-04-13—MODEX 2026 opens in Atlanta — Delta Electronics, Hai Robotics, Piaggio Fast Forward, and Attabotics/SAVOYE showcasing warehouse automation and autonomous logistics platforms.
2026-04-27—Unitree Robotics IPO on Shanghai Stock Exchange expected — first profitable humanoid robot company to go public, targeting ~$610M raise.