Figure AI vs Apptronik

AI Architecture: Vertical Integration vs. Strategic Partnerships

The most consequential difference between these companies is how they source intelligence. Figure AI builds its entire AI stack in-house using the Helix dual-system architecture: a vision-language model (VLM) handles scene understanding while a separate vision-language-action model (VLA) generates 200 Hz motor commands. After exiting its collaboration with OpenAI in late 2025, Figure doubled down on proprietary models trained through teleoperation data collection, sim-to-real transfer, and reinforcement learning. This vertical integration gives Figure full control over its training pipeline and model iteration speed — critical advantages when the bottleneck is data flywheel velocity. Apptronik takes the opposite approach: its partnership with Google DeepMind gives Apollo access to Gemini Robotics, one of the most capable VLA model families in existence. This means Apptronik can benefit from Google's massive research investment without bearing the cost, but it also creates dependency on an external partner whose robotics priorities may shift. The trade-off is autonomy vs. access: Figure owns its AI destiny; Apptronik rents frontier capability from the world's largest AI lab.

Hardware Philosophy: AI-Native vs. Enterprise-Grade

Figure treats its robots as inference platforms — the hardware exists to give AI models a body. The Figure 03, unveiled in early 2026, reflects this philosophy: 9% lighter than Figure 02, covered in soft textiles rather than hard panels, with embedded palm cameras and tactile sensors detecting forces as small as three grams. It is designed for environments beyond the factory floor, including household spaces. Apollo takes the opposite tack. With 71 degrees of freedom — significantly more than most competitors — and hot-swappable battery packs enabling zero-downtime operation, Apollo is engineered for the brutal realities of warehouse logistics: truck unloading, shelf picking, and machine tending. Where Figure optimizes for AI capability expression, Apptronik optimizes for operational reliability in existing industrial infrastructure.

Commercial Traction: BMW vs. Mercedes-Benz

Both companies have validated their robots with flagship automotive partners, creating an interesting parallel. Figure AI's deployment at BMW's Spartanburg plant is the more mature: 11 months of daily 10-hour shifts, over 1,250 runtime hours, 90,000+ parts loaded, contributing to more than 30,000 X3 vehicles. This is arguably the most significant commercial humanoid robot deployment to date. Apptronik's Mercedes-Benz partnership, initiated in March 2024, is earlier-stage but strategically significant — especially given Mercedes-Benz's dual role as both deployment partner and investor. Apptronik's additional partnership with GXO Logistics opens the warehouse vertical, which represents a larger total addressable market than automotive manufacturing. The comparison echoes the broader Boston Dynamics vs. Figure AI dynamic: proven deployment depth vs. broader market positioning.

Manufacturing Strategy and Scale Ambitions

Figure AI has announced plans to deploy 100,000 humanoid robots over the next four years through its BotQ program, relying on in-house manufacturing. This is an extraordinarily ambitious target — for context, the entire global industrial robot installed base is approximately 4 million units accumulated over decades. Apptronik's partnership with Jabil, one of the world's largest contract manufacturers, provides a potentially faster path to scale without the capital expenditure of building proprietary factories. Jabil's existing manufacturing infrastructure, supply chain expertise, and quality systems could give Apollo a production cost advantage, especially as humanoid robots are projected to reach $13,000 per unit by 2035 according to recent industry forecasts.

Ecosystem and Geographic Position

Figure AI operates from Sunnyvale, California, embedded in the Bay Area's AI talent ecosystem — proximity to the foundation model labs (OpenAI, Anthropic, Google DeepMind, Meta FAIR) that are driving the broader AI revolution. Apptronik is headquartered in Austin, Texas, at the center of an emerging robotics cluster that includes Tesla's Optimus program, NVIDIA's robotics division, and UT Austin's Human Centered Robotics Laboratory (from which Apptronik spun out). Austin's combination of lower operating costs, manufacturing infrastructure, and growing corporate R&D presence is making it a serious rival to the Bay Area and Boston as a robotics hub. The geographic difference also signals cultural priorities: Bay Area AI-native ambition vs. Austin's hardware-manufacturing pragmatism.

Investor Thesis and Valuation Gap

The 7x valuation gap ($39B vs. $5.3B) reflects different investor theses. Figure AI's backers — Microsoft, NVIDIA, OpenAI, Intel, Jeff Bezos — are betting on the embodied AI thesis: that whoever solves general-purpose humanoid intelligence captures one of the largest addressable markets in history. The valuation prices in optionality across every sector that uses human labor. Apptronik's investors — Google, Mercedes-Benz, John Deere, AT&T Ventures, Qatar Investment Authority — represent a more grounded thesis: that enterprise customers want reliable, deployable robots now, and that the AI layer can be sourced from partners like Google DeepMind rather than built from scratch. Both theses have merit; the question is whether the humanoid robotics market rewards platform plays or point solutions first.