Industry
Samsung and Meta Enter Humanoid Robotics: What Their Moves Actually Mean
Samsung Electronics plans to unveil its first humanoid robot in H2 2026, targeting its own semiconductor and display fabs. Meta completed the acquisition of Assured Robot Intelligence on May 1, adding foundation models for whole-body control to its Superintelligence Labs. Platform companies are now in the humanoid race in force.
Samsung Electronics plans to unveil its first humanoid robot in the second half of 2026, targeting semiconductor and display production lines. Days earlier, Meta completed its acquisition of Assured Robot Intelligence (ARI), a startup building foundation AI models for robots. Two of the world's largest technology companies entering the humanoid market within weeks of each other is not a coincidence — it is a signal of where the industry is heading.
The announcements, both confirmed this week, mark a new phase for physical AI. The first wave of humanoid robotics was driven by pure-play startups: Figure AI, Agility Robotics, 1X Technologies. The second wave now arriving is dominated by trillion-dollar platform companies with manufacturing scale, chip supply chains, and existing enterprise relationships. That changes the competitive math considerably.
Key Stats
H2 2026
Samsung Robot Reveal Target
2028
Samsung Full Commercial Rollout
May 1
Meta / ARI Acquisition Close
$13K
Projected Unit Cost by 2028
Two Different Bets on the Same Outcome
Samsung and Meta are approaching humanoid robotics from opposite directions, which makes the convergence more interesting. Samsung is a hardware company trying to build robots the way it builds everything else: vertically integrated, proprietary, manufacturing-first. Meta is a software and AI company that acquired the intelligence layer first and will worry about the body later.
Samsung's humanoid program is led by Oh Jun-ho, the engineer who built Hubo — Korea's first bipedal robot — at KAIST in the early 2000s. Samsung brought him in to direct its Robot Center, and the company has been quiet about specifics ever since. What is now known: pilot production lines launch in 2026, initial deployment targets Samsung's own semiconductor and display factories, and full commercial rollout is planned for 2028. The H2 2026 reveal will be the first public look at the hardware.
Meta's path differs. On May 1, 2026, Meta completed its acquisition of Assured Robot Intelligence (ARI), founded by NYU roboticist Lerrel Pinto and UC San Diego's Xiaolong Wang. ARI's work focuses on foundation models that let robots read, predict, and adapt to human behavior during physical tasks — the same problem that has stalled humanoid deployment in unstructured environments. The ARI team joined Meta's Superintelligence Labs, not its hardware division, which tells you where Meta thinks the value sits.
🏭 Samsung's Approach
Vertical integration from chips to actuators to robot frame. Manufacturing deployment first, then broader commercial rollout. Targets 2028 for full factory integration, leveraging existing supply chain relationships and defect-prediction AI already running on Samsung production lines.
🧠 Meta's Approach
Acquire the AI brain, share it industry-wide. ARI's foundation models for whole-body control and human behavior prediction join Meta's Superintelligence Labs. Hardware partnerships rather than hardware ownership. Positioning Meta's AI stack as the operating system for third-party robot platforms.
Under the Hood: What Each Company Is Actually Building
Samsung has not released hardware specifications for its humanoid. What has surfaced through analyst notes and Samsung Research publications: the robot will use electric actuators (not hydraulic), prioritize dexterous hand design, and integrate Samsung's own battery technology — with in-house battery production for robotics targeted for H2 2027. Samsung Research papers describe a target architecture called "Samsung Bot" with three layers: robot intelligence (navigation, manipulation, task planning), robot platform (SLAM, APIs, developer tools), and robot mechanisms (actuator design optimized for cost-effective locomotion).
Rainbow Robotics, the Korean robotics firm Samsung acquired a majority stake in during 2024, provides the foundational mechanical engineering. Rainbow's RB series collaborative robots have been shipping in small volumes since 2022, giving Samsung real manufacturing data on actuator wear, joint tolerances, and force control performance. That data is feeding into the humanoid design.
AI-generated image
ARI's foundation models operate on the principle that robots must learn to predict human intent and adapt mid-task, not simply execute pre-programmed sequences.
Meta's ARI acquisition targets a specific technical gap: current robots fail when tasks deviate from training scenarios. ARI's models use human behavior prediction to anticipate what comes next in a task sequence, enabling whole-body control that adjusts in real time. Pinto's lab at NYU published extensively on dexterous manipulation from demonstration, which ARI built into a deployable platform. The self-learning component reduces retraining cycles when the robot encounters variations in object placement, lighting, or task sequence.
The technical comparison table below places the two newcomers against established humanoid platforms, using available public specifications. Samsung's figures are estimates from analyst reports; the company has not published hardware specs.
| Spec | Samsung (est. 2026) | Boston Dynamics Atlas | Figure 03 | Tesla Optimus Gen 3 |
|---|---|---|---|---|
| Height | ~1.70 m (estimated) | 1.50 m | ~1.70 m | 1.73 m |
| Payload | Not disclosed | 50 kg (110 lbs) | ~20 kg | ~20 kg |
| Actuator Type | Electric (in-house) | Electric (proprietary) | Electric | Electric (Tesla) |
| AI Architecture | Samsung Research / agentic | Google DeepMind / fleet learning | OpenAI (reported) | VLA (in-house Tesla) |
| Primary Target | Samsung fabs / displays | Hyundai HMGMA plant | BMW + enterprise | Tesla Gigafactories |
| Commercial Status | Reveal H2 2026 | Trials (deploy 2028) | Deployed (BMW) | Internal (Tesla) |
Meta + ARI: What the Acquisition Actually Covers
• Whole-body control (WBC): Coordinates all limbs simultaneously for dynamic tasks like moving a heavy object while maintaining balance.
• Human behavior prediction: Models trained on human demonstration data anticipate task sequences, reducing reliance on pre-scripted motion plans.
• Self-learning loop: Robots update their internal models from each new task attempt, narrowing the sim-to-real gap over time without requiring full retraining.
• Cross-platform deployment: ARI's stack was designed to be hardware-agnostic — applicable to multiple robot form factors rather than one proprietary system.
Who's Entering the Race Now
The humanoid market in 2024 and 2025 was venture-backed: Figure AI raised $675 million at a valuation reported near $2.6 billion; Agility Robotics secured Toyota investment; 1X Technologies pulled funding from OpenAI. In 2026, the character of investment is changing. Platform companies with existing market positions are moving in, and they are not writing checks to startups — they are building or buying directly.
• Samsung Electronics: Majority stake in Rainbow Robotics (2024). Internal humanoid program led by Oh Jun-ho (Hubo creator). First reveal H2 2026, factory pilots in 2026, full commercial rollout 2028. CFO Park Soon-cheol: "Through the development of humanoid robots, we aim to innovate manufacturing productivity and daily experiences."
• Meta: Acquired Assured Robot Intelligence (May 1, 2026, undisclosed terms). ARI founders Lerrel Pinto and Xiaolong Wang join Meta Superintelligence Labs. Strategy: build the AI stack, partner on hardware rather than own it. Part of a broader pivot away from the metaverse toward physical AI.
• Hyundai / Boston Dynamics: New electric Atlas unveiled at CES January 2026. 56 degrees of freedom, 50 kg payload, fleet-learning AI in partnership with Google DeepMind. Robotics Metaplant Application Center (RMAC) opens in the U.S. for training. All 2026 Atlas units already committed to RMAC and initial partners including Google DeepMind.
• Tesla: Optimus Gen 3 operating inside Tesla Gigafactories on battery handling tasks. Internal deployment only as of May 2026; no external commercial availability announced. CEO Elon Musk has stated Optimus production scale is Tesla's primary long-term value driver.
• Figure AI: Figure 03 deployed at BMW plant, highest-funded pure-play humanoid startup at approximately $3.9 billion valuation as of Q1 2026. Expanding from BMW pilot to broader enterprise sales.
• AGIBOT: Chinese firm debuted its A2 humanoid model at the Met Gala on May 5, 2026, alongside designer Alexander Wang. Embodied AI platform targeting household, commercial, and industrial sectors. Backed by Chinese state-adjacent investment.
What Platform Entry Means for the Humanoid Market
AI-generated image
Warehouse and factory automation represents the near-term deployment opportunity for humanoid robots, where tasks are repetitive, environments are structured, and the cost calculus is clearest.
Platform companies entering a market do not simply add competition — they change the structure of the market. Samsung entering humanoid robotics means Samsung's downstream customers (consumer electronics brands, automotive OEMs, appliance manufacturers) get a credible internal option. It also means Samsung has an incentive to drive component costs down in a way that benefits its own robot first and competitors second.
Meta's move is different in kind. By acquiring an AI stack rather than building a robot, Meta is positioning itself as infrastructure. If ARI's models become the standard approach to whole-body control and human behavior prediction, every hardware company that adopts them becomes, in some sense, a Meta partner. That is a familiar playbook from Meta's core advertising and social platform business: own the intelligence layer, let others compete on the hardware.
The economic pressure this creates on pure-play startups is real. Figure AI, Agility Robotics, and 1X Technologies built their companies on the assumption that neither Samsung nor Meta would move quickly into this space. That assumption expired in May 2026. Startups that have locked in long-term enterprise contracts (Figure's BMW deal, Agility's Toyota relationship) have a buffer. Startups still in R&D mode face a tighter fundraising environment as platform credibility shifts toward incumbents.
$35K
Avg. humanoid unit cost in early 2026
$13K
Projected unit cost by 2028 at scale
1M+
Annual units projected by 2030–2035
Key Insight
Samsung's core advantage over every humanoid startup: it already builds the components. Actuators, batteries, sensors, compute chips — Samsung either makes them or has supply contracts at a scale that lets it undercut rivals on cost by design. A $35,000 humanoid robot is a bold product for a startup. For Samsung, it is a product line that can leverage existing manufacturing infrastructure running at volumes humanoid-only companies cannot match.
The 12-Month Outlook
The second half of 2026 will test Samsung's timeline. A public reveal of the humanoid hardware — even a controlled demo — would validate that the program is on schedule. What analysts and journalists are watching for: payload specs, the actuator configuration, and whether Samsung shows the robot performing actual production tasks or just walking across a stage. A CES 2027 moment (full demo + deployment announcement) is a credible scenario if H2 2026 reveal is a softer preview.
For Meta, the 12-month question is what Pinto and Wang publish or present under the Meta banner. ARI's work was previously in academic and startup mode; Meta's resources will accelerate development significantly. Watch for papers, API announcements, or hardware partnerships that signal Meta intends to license its physical AI stack rather than keep it internal. A public developer platform for robot AI — something like a Meta Embodied AI SDK — would be the logical next step in the platform strategy.
Milestones to Watch
• Q3 2026: Samsung humanoid public reveal — hardware specs and form factor confirmed.
• Q4 2026: Samsung pilot production line launch at semiconductor or display facility.
• H2 2026: Meta / ARI first public output from Superintelligence Labs (paper, SDK, or partner announcement).
• Q1 2027: Samsung battery division integration milestone — in-house robot battery production confirmation.
• 2028: Samsung full commercial rollout target. Hyundai HMGMA Atlas deployment begins.
Frequently Asked Questions
What is Assured Robot Intelligence (ARI) and why did Meta buy it?
ARI built foundation AI models for robots — specifically models that allow a robot to predict human behavior and adapt its physical responses during complex tasks. Co-founders Lerrel Pinto (NYU) and Xiaolong Wang (UC San Diego) had already published influential research on dexterous manipulation from demonstration. Meta acquired ARI on May 1, 2026, to gain this team and their models, integrating them into Meta's Superintelligence Labs. The acquisition price was not disclosed. Meta's goal is to build the intelligence stack for physical AI, similar to how its open-source Llama models positioned it in the language AI space.
When will Samsung's humanoid robot be available commercially?
Samsung has confirmed a public reveal in H2 2026 (second half of the year) and pilot production lines starting in 2026. Full commercial rollout — meaning external sales beyond Samsung's own factories — is targeted for 2028. Initial deployment will focus on Samsung's semiconductor and display manufacturing plants, where the company already has operational data on automation requirements. Broader applications including consumer and retail environments are not expected before 2030.
How does Samsung's approach differ from Tesla or Figure AI?
Tesla and Figure AI built robots primarily to deploy them — Tesla in its own factories, Figure in enterprise partner facilities. Samsung's angle is different: it is building a robot to deploy in its own fabs while also positioning for external commercial sales. Crucially, Samsung already manufactures the components that go into robots — actuators, batteries, sensors, semiconductors. That vertical integration allows cost reduction paths that pure-play humanoid startups cannot replicate. The risk is that Samsung's manufacturing culture is optimized for high-volume consumer electronics, not low-volume precision robotics, which are very different engineering disciplines.
Does Meta plan to build its own robot hardware?
Not immediately, based on available information. Meta's stated strategy is to develop and share AI and software industry-wide, with hardware partnerships rather than proprietary hardware. The ARI team joining Superintelligence Labs (not a hardware division) confirms this orientation. Meta may eventually release open-weight models for robot AI the way it has with Llama, or it may offer an embodied AI developer platform as a commercial product. A Meta-branded humanoid robot is not part of any announced roadmap as of May 2026.
What happens to startups like Figure AI and Agility Robotics now that Samsung and Meta have entered?
Companies with locked-in enterprise contracts have time. Figure AI's BMW deal and Agility's Toyota relationship give both firms revenue and operational data that Samsung does not have yet. Samsung's first commercial humanoid is still roughly two years away. The more vulnerable position is held by humanoid startups still in pre-deployment R&D with no customer commitments — they face a fundraising environment that will grow more skeptical of pure-play humanoid bets as platform companies demonstrate capability. Consolidation through acquisition is a plausible outcome for several second-tier players by 2027.
The Platform Era Begins
The humanoid robotics market is about two years into its commercial phase, and the structure is already shifting. The 2024–2025 window belonged to well-funded startups racing to prove that humanoid robots could do real work in real factories. That proof has been delivered — Figure at BMW, Agility at Toyota, Boston Dynamics at Hyundai — and the investors who funded that proof are now watching platform companies walk in behind them.
Samsung and Meta bring different strengths: Samsung has manufacturing infrastructure and supply chain leverage; Meta has AI research talent and a platform distribution model. Neither moves at startup speed, but both move at a scale that compresses the market's long-term structure. The companies that will survive this transition are the ones that have either locked in durable enterprise relationships or developed capabilities — manipulation dexterity, specific domain expertise, deployment software — that the platforms cannot simply acquire or replicate quickly.
The Bottom Line: Samsung entering with hardware and Meta entering with AI intelligence in the same week is the clearest signal yet that humanoid robotics has moved from a venture-backed experiment to a strategic priority for platform-scale companies, and the competitive dynamics will not look the same in 2027 as they did in 2025.
The H2 2026 Samsung reveal will set the tone for the second half of the year. If Samsung shows a production-ready robot performing actual factory tasks, it will accelerate investment in the sector and validate Hyundai's similar timeline. If it shows a demo-only prototype, the story stays in the "coming soon" column for another 12 months. Either way, the company's entry has already changed the framing.