# Figure AI vs Agibot

> Figure AI vs AgiBot: Compare these leading humanoid robotics companies on hardware, AI, pricing, deployment scale, and industrial capabilities in 2026.

Source: https://metavert.io/compare/figure-ai-vs-agibot  
Published: 2026-03-24  
Updated: 2026-03-24

Comparison

The humanoid robotics race has two clear frontrunners heading into 2026: [Figure AI](https://metavert.io/figure-ai), the Silicon Valley startup valued at $39 billion after its blockbuster Series C, and [AgiBot](https://metavert.io/agibot), the Shanghai-based manufacturer that shipped more humanoid robots in 2025 than any company on earth. Both are pursuing the vision of general-purpose humanoid robots — but from radically different starting points and with divergent strategies.

[Figure AI](https://metavert.io/figure-ai) has bet on deep vertical integration, building its own AI foundation model (Helix) and a dedicated manufacturing facility (BotQ) capable of producing 12,000 units per year. Its Figure 03, unveiled in October 2025, represents a ground-up redesign optimized for home environments. Meanwhile, [AgiBot](https://metavert.io/agibot) has taken a portfolio approach, offering full-sized humanoids (A2), compact humanoids (X2), industrial wheeled humanoids (G2), and quadrupeds (D1) — and has already crossed 5,000 units shipped globally with over RMB 1 billion in revenue.

This comparison breaks down where each company leads, where they lag, and which is the better fit for specific use cases — from factory floors to consumer homes to research labs.

## Feature Comparison

| Dimension | Figure AI | AgiBot |
| --- | --- | --- |
| Headquarters | San Jose, California, USA | Shanghai, China |
| Valuation (2025-2026) | $39B post-money (Series C, Sep 2025) | ~$2B; targeting $5-6B at planned HK IPO in 2026 |
| Total Funding | $1.9B+ across Series A-C | ~$83M across multiple rounds |
| Units Shipped (2025) | Limited pilot deployments (BMW, etc.) | 5,168 units — #1 globally |
| Flagship Humanoid | Figure 03 (Oct 2025): 1.7m, lighter than predecessor, 10 Gbps mmWave data offload | A2 Series: 1.75m, 55 kg, 49+ DoF, 200 TOPS AI computing |
| AI Foundation Model | Helix — proprietary Vision-Language-Action model (replaced OpenAI dependency) | GO-1 foundation model + WorkGPT multimodal AI (96% accuracy across text/audio/visual) |
| Industrial Precision | General-purpose manipulation; BMW deployment focused on part loading | G2: sub-millimeter accuracy, 0.5N force-controlled 7-DOF arms, demonstrated RAM insertion at CES 2026 |
| Dexterous Manipulation | Custom tactile sensors detecting forces as small as 3 grams; palm cameras per hand | OmniHand robotic limb; A2 visual fingertip sensors capable of threading needles |
| Product Breadth | Single humanoid platform (Figure 03) | Full portfolio: A2 humanoid, X2 compact, G2 industrial, D1 quadruped, C5 cleaning robot |
| Software Platform | Helix VLA model, fleet learning via mmWave offload | AimRT (custom C++20 runtime), Genie Sim 3.0 (NVIDIA Isaac Sim), AgiBot World open dataset |
| Manufacturing Capacity | BotQ facility: 12,000/year initial, targeting 100,000 over 4 years | Already producing at scale; 5,000+ shipped in 2025 alone |
| Business Model | Direct sales and enterprise pilot programs | Direct sales + Robot-as-a-Service (RaaS) leasing model |

## Detailed Analysis

### AI and Autonomy: Proprietary Models vs. Open Ecosystems

The most consequential divergence between Figure AI and AgiBot is their approach to [artificial intelligence](https://metavert.io/artificial-intelligence). Figure AI made a pivotal strategic decision in 2025 by moving away from its partnership with OpenAI and developing Helix, a proprietary Vision-Language-Action (VLA) model. Helix allows Figure 03 to perceive its environment, reason about tasks in natural language, and execute physical actions — all within a single integrated model. This vertical integration gives Figure tighter control over the full AI stack but means the company bears the full R&D burden.

AgiBot has taken a more distributed approach, layering multiple specialized AI systems: WorkGPT handles multimodal interaction, the GO-1 foundation model drives embodied reasoning, and Genie Sim 3.0 provides a simulation environment built on [NVIDIA](https://metavert.io/nvidia) Isaac Sim for training. AgiBot has also open-sourced its AgiBot World dataset — a move that positions the company as a platform player and earned an IROS 2025 Best Paper Award finalist nomination. Their AimRT middleware, a custom C++20 runtime, outperforms ROS2 and reflects serious investment in software infrastructure.

For enterprises evaluating these systems, Figure's single-model approach may offer more coherent behavior in novel situations, while AgiBot's modular stack provides flexibility and a growing open-source ecosystem that researchers and integrators can build upon.

### Hardware Design Philosophy: Home-Ready vs. Industrial-Grade

Figure 03 was explicitly designed with consumer and home environments in mind. It is 9% lighter than its predecessor with reduced volume, features soft-touch materials, wireless charging, and an improved audio system for voice-based interaction. The embedded palm cameras and tactile sensors capable of detecting 3-gram forces suggest a robot built for the delicate, unstructured tasks of domestic life — folding laundry, loading dishwashers, organizing shelves.

AgiBot's hardware strategy is broader and more commercially pragmatic. The A2 is a full-sized humanoid with 49+ degrees of freedom suited for reception, guided tours, and general service tasks. The G2 is purpose-built for [smart manufacturing](https://metavert.io/smart-manufacturing), with IP42 rating, wheeled omnidirectional mobility, and 7-DOF force-controlled arms achieving sub-millimeter precision — it demonstrated live electronics assembly at CES 2026. This portfolio approach means AgiBot can address manufacturing, education, logistics, and consumer markets simultaneously.

The tradeoff is clear: Figure is optimizing a single platform for maximum general-purpose capability, while AgiBot offers specialized form factors for specific verticals. Neither approach is inherently superior — it depends on whether a buyer needs a Swiss Army knife or a purpose-built tool.

### Scale and Manufacturing: Ambition vs. Execution

This is where AgiBot holds an unambiguous lead. The company shipped 5,168 humanoid robots in 2025, ranking first globally in both units and revenue (exceeding RMB 1 billion). Its A2 won multiple Best of CES 2026 awards and even earned a Guinness World Record for a 106 km autonomous walk. AgiBot's Robot-as-a-Service model further lowers adoption barriers, letting customers lease rather than buy.

Figure AI's BotQ manufacturing facility is designed for 12,000 units per year initially, with a 100,000-unit target over four years. But as of early 2026, Figure's real-world deployments remain limited to pilot programs — most notably an 11-month stint at BMW's Spartanburg plant where Figure 02 loaded over 90,000 parts across 1,250+ runtime hours. The BMW deployment is impressive as a proof of concept, but it's a single site running a single task.

China's [manufacturing](https://metavert.io/manufacturing) ecosystem gives AgiBot structural advantages in scaling production quickly and cost-effectively — a dynamic that mirrors the country's dominance in EV production. Figure will need to execute flawlessly on BotQ to close the gap.

### Funding and Financial Position

Figure AI is one of the most well-capitalized private robotics companies in history, with over $1.9 billion raised and a $39 billion post-money valuation after its September 2025 Series C. Investors include [NVIDIA](https://metavert.io/nvidia), Intel Capital, Salesforce, and Qualcomm Ventures — a strategic roster that provides not just capital but semiconductor and enterprise distribution partnerships.

AgiBot has raised a comparatively modest $83 million but is planning a Hong Kong IPO in 2026 targeting a $5-6 billion valuation, which would raise over $1 billion. The company's path to profitability looks shorter given its existing revenue base (RMB 1B+), while Figure's massive valuation carries higher expectations that will take longer to justify through revenue.

The financial contrast illustrates a broader pattern: Figure is a venture-backed moonshot with enormous resources but pre-revenue economics, while AgiBot is a commercially validated business scaling toward public markets.

### Ecosystem and Developer Access

AgiBot has made deliberate moves to build an open ecosystem. The AgiBot World dataset (hosted on GitHub) provides large-scale manipulation data for [embodied AI](https://metavert.io/embodied-ai) research. Their AimRT middleware is designed as a developer-friendly runtime, and the X2 compact humanoid is explicitly marketed for education and research applications. This openness accelerates third-party development and positions AgiBot as infrastructure for the broader robotics community.

Figure AI's approach is more closed. Helix is proprietary, and the company's fleet learning system — where robots upload terabytes of data via 10 Gbps mmWave for centralized model improvement — is designed as a walled garden. This may yield faster iteration on Figure's own models but limits external developer participation. For research institutions and universities, AgiBot is currently the more accessible platform.

### Global Strategy and Market Access

AgiBot made its U.S. market debut at CES 2026, showcasing its full portfolio and launching a global online store. However, the company faces potential headwinds from U.S.-China trade tensions and technology export controls that could limit its access to Western markets, particularly for defense-adjacent or critical infrastructure applications.

Figure AI, as a U.S.-based company with government-friendly investors, faces fewer geopolitical barriers in Western markets. CEO Brett Adcock has publicly projected that humanoid robots will perform unsupervised multi-day tasks in unfamiliar homes by 2026 — positioning Figure squarely in the consumer and enterprise markets where trust and regulatory compliance matter most. For buyers in regulated industries or government-adjacent sectors, Figure's U.S. origin is a meaningful advantage.

## Best For

#### Precision Manufacturing & Assembly

AgiBot

AgiBot's G2 offers sub-millimeter accuracy with 0.5N force-controlled arms, purpose-built for industrial assembly. Its CES 2026 RAM insertion demo proved real-world manufacturing viability that Figure hasn't yet matched.

#### Home Assistance & Domestic Tasks

Figure AI

Figure 03 was designed from the ground up for home environments — lighter, softer materials, wireless charging, 3-gram tactile sensitivity. Helix's VLA model is optimized for the unstructured chaos of residential spaces.

#### Large-Scale Commercial Deployment (Today)

AgiBot

With 5,000+ units shipped and a RaaS leasing model, AgiBot is the only company that can deliver humanoid robots at scale right now. Figure's BotQ is promising but not yet producing at volume.

#### Automotive & Heavy Industry Pilots

Figure AI

Figure's 11-month BMW deployment — 10-hour daily shifts, 90,000+ parts loaded — is the most rigorous automotive proof point in humanoid robotics. Western automakers will prefer a U.S.-based partner for supply chain and IP reasons.

#### Research & Academic Use

AgiBot

AgiBot's open-source AgiBot World dataset, AimRT middleware, and the affordable X2 research platform make it the better choice for universities and labs. Figure's closed ecosystem limits external experimentation.

#### Hospitality & Customer-Facing Service

AgiBot

The A2's proven deployment in reception and guided interaction scenarios, combined with WorkGPT's 96% multimodal accuracy for voice/visual interaction, gives AgiBot a clear edge in service environments.

#### Regulated Industries & Government

Figure AI

For defense-adjacent, healthcare, or critical infrastructure applications in Western markets, Figure's U.S. headquarters and strategic investor base (Intel, Qualcomm, NVIDIA) make it the safer procurement choice.

#### Multi-Environment Fleet Operations

Tie

Figure's mmWave fleet learning system enables rapid cross-robot improvement, while AgiBot's diverse product portfolio (humanoids, quadrupeds, cleaning bots) covers more physical environments. The best choice depends on whether you need depth or breadth.

## The Bottom Line

AgiBot and Figure AI represent two credible but fundamentally different bets on the future of humanoid robotics. If you need robots deployed at scale today — in factories, hotels, retail spaces, or research labs — AgiBot is the pragmatic choice. No other humanoid company has matched its 5,000+ unit shipments, its breadth of form factors, or its commercial revenue. The Robot-as-a-Service model and open-source developer tools further reduce adoption risk. AgiBot is executing the China manufacturing playbook that dominated EVs, and it's working.

Figure AI is the higher-ceiling bet. Its $39 billion valuation reflects investor conviction that Helix — a proprietary, vertically integrated VLA model — will eventually produce the most capable general-purpose humanoid on the planet. The Figure 03 hardware is the most refined consumer-oriented humanoid design available, and the BMW deployment proves the technology works in demanding industrial settings. But Figure is still pre-scale, and the gap between a $39B valuation and actual revenue is vast. If BotQ delivers on its 100,000-unit roadmap and Helix continues improving through fleet learning, Figure could leapfrog the competition by 2027-2028.

Our recommendation: enterprises with immediate deployment needs — especially in Asia-Pacific markets — should evaluate AgiBot's portfolio seriously. Western companies in regulated industries or those betting on long-term home robotics should partner with Figure AI now to secure early access. The smartest large organizations will pilot both.

## Related Topics

- [Figure AI](https://metavert.io/figure-ai)
- [AgiBot](https://metavert.io/agibot)
- [Humanoid Robots](https://metavert.io/humanoid-robots)
- [Embodied AI](https://metavert.io/embodied-ai)
- [NVIDIA](https://metavert.io/nvidia)
- [Smart Manufacturing](https://metavert.io/smart-manufacturing)
- [Robot-as-a-Service (RaaS)](https://metavert.io/robot-as-a-service)
- [Tesla Optimus](https://metavert.io/tesla-optimus)

## Further Reading

- [Introducing Figure 03 — Figure AI Official Announcement](https://www.figure.ai/news/introducing-figure-03)
- [AgiBot Makes Its U.S. Market Debut at CES 2026](https://www.prnewswire.com/news-releases/agibot-makes-its-us-market-debut-at-ces-2026-with-its-full-humanoid-robot-portfolio-302652403.html)
- [China Is Running the EV Playbook on Humanoid Robots — Rest of World](https://restofworld.org/2026/china-humanoid-robots-unitree-agibot-tesla-optimus/)
- [AgiBot World: Large-Scale Manipulation Platform (GitHub)](https://github.com/OpenDriveLab/AgiBot-World)
- [Figure Exceeds $1B in Series C Funding at $39B Valuation](https://www.figure.ai/news/series-c)