Apptronik

Snapshot

Apptronik makes Apollo, a 5'8", 160-pound, 55-pound-payload electric humanoid robot designed for factory floors and warehouses. Spun out of the University of Texas at Austin's Human Centered Robotics Lab in 2016 by a team that had built NASA's Valkyrie for the 2013 DARPA Robotics Challenge, the company spent eight years quietly shipping research hardware before unveiling Apollo in August 2023 and signing production pilots with Mercedes-Benz, GXO, and Jabil (The Robot Report, August 23, 2023; TechCrunch, February 13, 2025). A $350M Series A in February 2025 was extended in March 2025 and then again in February 2026 for another $520M, bringing total Series A funding north of $935M at a ~$5.3B post-money valuation (Bloomberg, February 11, 2026; TechCrunch, February 11, 2026). Apollo is a bet on a very specific premise: that electric linear actuators, an academic-engineering pedigree, and a Google DeepMind brain can deliver a useful industrial humanoid at a sub-$50K unit cost before Figure's war chest or Tesla's vertical scale decides the category.

Founding story

The company's roots run deeper than the 2016 incorporation date suggests. Luis Sentis set up the Human Centered Robotics Laboratory at UT Austin in 2004. In 2011 Nick Paine and Sentis joined the NASA Johnson Space Center team building Valkyrie (R5), a 44-degree-of-freedom humanoid that would compete in the 2013 DARPA Robotics Challenge; their linear actuator work on Valkyrie became the starting kernel for everything Apptronik has shipped since (aiwiki.ai; UT McCombs News, Spring 2024).

Jeff Cardenas, who is now CEO, came to robotics by a less obvious path. Son of an Air Force colonel, two Austin degrees (BBA 2008, MSTC 2013), four years at Deloitte Consulting, then a commercialization fellowship at UT's IC2 Institute. At IC2 he met retired Air Force Brig. Gen. Bill Welch — who became Apptronik's initial CEO for its first three and a half years — and through Welch, Paine and Sentis (UT McCombs News, Spring 2024). Cardenas pitched Paine on the commercial vision and argued that Texas, with its manufacturing density and talent pool, was the right place to build it. His defining line, repeated often since, is "this is the PC moment for robotics." The company spun out of UT in March 2016 on a UT Innovation Grant and a NASA SBIR.

How it works

Apollo is a 1.72-meter, 72.5-kg humanoid with a 25-kg payload (The Robot Report, August 23, 2023). The engineering differentiator is below the carbon-fiber skin: Apptronik's own proprietary linear electric actuators — 13+ generations of in-house design — replace both the hydraulics of legacy humanoids like Boston Dynamics' original Atlas and the off-the-shelf rotary motors that most newer entrants use. Torque sensors in every joint give the robot real-time force feedback, which is why Apollo can carry a tote and respond to a human bumping into it without catastrophic motion. Its battery runs ~4 hours and is hot-swappable; with pack rotation Apollo targets ~22 hours/day of uptime.

Perception is a stereo-vision + IMU stack with a safety layer that slows the robot near people and halts on close proximity; in a fall, Apollo curls to protect itself. The initial units ship with simple 0–1 DoF grippers for "gross manipulation" — moving totes, lifting boxes — with dexterous hands coming later, possibly via partnerships (The Robot Report, August 23, 2023). The body is modular: the same torso runs on legs, wheels, or a fixed platform. Apollo 2, unveiled mid-2025, adds next-generation actuation, upgraded batteries, new sensors, and a wheeled variant with a z-axis that extends to seven feet (Automate.org "Robot Park," 2025).

The AI stack is Apollo's second structural bet. In December 2024 Apptronik announced a strategic partnership with Google DeepMind Robotics; Apollo now runs Gemini Robotics and Gemini Robotics 2 vision-language-action models co-developed with DeepMind, and the robot's deployments feed training data back (Apptronik press release, December 19, 2024). This is Apptronik's answer to Figure's decision to ditch OpenAI and build its own proprietary Helix stack — rather than try to build a foundation model, Apptronik rented the best one.

Product and business overview

There is effectively one product: Apollo 2, in a few form factors. The company also sells teleoperation-based skill acquisition — Mercedes-Benz workers at the Berlin-Marienfelde plant have "transferred their knowledge to Apollo using teleoperation processes and augmented reality" (Assembly Magazine, March 21, 2025). Pilots and deployments are the current commercial product; Jabil is the contract manufacturer and co-deployment partner, announced February 2025. The 90,000-square-foot "Robot Park" that Apptronik opened in the former Dell server plant across from HQ in mid-2025 is the testing and burn-in facility.

Business model and pricing

Cardenas has publicly targeted a unit cost below $50,000, versus his earlier framing of "about the same as a new car" (TechCrunch, February 13, 2025; The Robot Report, August 23, 2023). For perspective, Figure has not disclosed pricing; Agility's Digit is reported in the ~$250K range; Unitree's G1 starts near $16K and the R1 at ~$5,900 (Interesting Engineering, 2025).

Mix today is pilot deals plus outright sales. During the pre-Series-A period, Cardenas told TechCrunch Apptronik had generated more revenue than it had raised through pilots and sales of earlier robots (Astra, Draco, Hume, the Sagittarius/Apex exoskeletons for USSOCOM). No formal Robot-as-a-Service pricing has been confirmed; the <$50K figure is a target unit cost, not a lease. "Demonstrating useful work" is Apptronik's 2025 framing; "true commercialization and scaling" is 2026 and beyond (TechCrunch, February 13, 2025; Dallas Innovates, 2026).

Traction over time

Three co-founders, 15+ systems before Apollo. In late 2020 Astra — an upper-body humanoid — became Apptronik's first commercial product. June 2022: $14.6M seed led by Grit Ventures, Apollo development begins (Built In, June 29, 2022). August 2023: Apollo unveiled. March 2024: Mercedes-Benz pilot. June 2024: GXO Logistics multi-phase R&D deal. December 2024: Google DeepMind partnership. February 2025: $350M Series A. March 2025: Series A expanded to $403M with Mercedes-Benz taking an equity stake, ARK Invest and Japan Post Capital joining. Mid-2025: Apollo 2 and Robot Park. February 11, 2026: $520M extension bringing total Series A past $935M at ~$5.3B — with AT&T Ventures, John Deere, and Qatar Investment Authority joining Google, Mercedes, B Capital, and PEAK6 (Bloomberg, February 11, 2026; Dallas Innovates, 2026).

Headcount: just over 170 at the time of the $350M Series A in February 2025; nearly 300 by early 2026 — a ~75% growth plan Cardenas laid out immediately after the Series A. Revenue is not disclosed and the company has explicitly said it is still in "early stages of development" (TechCrunch, February 11, 2026).

Market analysis

The humanoid TAM is the loudest number in venture right now. Goldman Sachs Research pegged it at $38B by 2035 in a January 2024 report — a 6x revision upward — with shipments of 1.4M units by 2035 and 250K industrial units by 2030, and a per-unit manufacturing cost that fell from $50K–$250K to $30K–$150K in a single year (Goldman Sachs Insights, February 27, 2024; Goldman later raised the 2035 unit forecast to 6.5M). Morgan Stanley went further: a "$5 trillion global TAM by 2050" with more than a billion humanoids in use (Morgan Stanley Research, "A $5 Trillion Global Market," May 14, 2025). The labor backdrop — Deloitte and NAM project 2.1M unfilled U.S. manufacturing jobs by 2030 — is the thesis every humanoid CEO invokes.

Competitive intel

Figure AI (Brett Adcock, San Jose) is the direct rival and the fundraising giant. Series C in September 2025 at a $39B valuation on more than $1B in new funding, following $675M at $2.6B in February 2024 (Microsoft, Nvidia, Bezos, Amazon, OpenAI participated) (Wikipedia; SiliconAngle, September 16, 2025). Figure 03 (October 2025) is 1.7m, 60 DoF, 20 DoF per hand, with tactile sensors under 3g and wireless inductive charging. Figure dropped OpenAI in early 2025 and now runs its own Helix vision-language-action model. Figure's edge over Apptronik is fundraise scale — 5x — and its own AI stack. Apptronik's edge over Figure is deeper academic and NASA pedigree, Mercedes as an equity investor, and Google DeepMind models rather than a rolled-your-own foundation bet.

Agility Robotics (Digit, Salem/Pittsburgh) is the warehouse rival and the first humanoid in commercial operations. $150M in 2024 at ~$1B+; reportedly raised $400M in April 2025 (FreightWaves/SiliconAngle); a reported SPAC merger at $2.5B in mid-2026 (AI Weekly, June 24, 2026). Digit is already deployed in Amazon warehouses and the GXO Spanx facility — exactly where Apollo is pointed. Agility's edge: first-mover in commercial warehouse ops. Apptronik's: human-form factor plus live manufacturing pilots.

1X Technologies (Norway/US, OpenAI-backed) focuses on consumer; its Neo humanoid is in home trials in 2025 (The Robot Report, January 10, 2024). Tesla Optimus has committed up to $25B and targets $20–30K pricing, with Musk calling for mass production in 2026 — a vertical-scale threat Apptronik cannot match on its own. Boston Dynamics, owned by Hyundai, retired hydraulic Atlas in April 2024 and launched electric Atlas aimed at commercial manufacturing; Hyundai in January 2026 announced it will purchase "tens of thousands" of Boston Dynamics robots (Gadget Review; AJC, January 6, 2026) — an in-house order book nobody else has. Sanctuary AI (Vancouver, Phoenix humanoid) competes on dexterous manipulation at smaller scale. Unitree is the Chinese price war — H1 at ~$90K, G1 at $16K, R1 at $5,900 — cheap but unproven in industrial reliability, and the structural threat to the whole Western pricing stack.

History and evolution

2004: HCRL founded at UT Austin. 2011–2013: Paine and Sentis join NASA JSC's Valkyrie team; Valkyrie competes in DARPA Robotics Challenge. March 2016: Apptronik founded; UT Innovation Grant and first NASA SBIR. 2016–2020: exoskeletons (Sagittarius, Apex for USSOCOM) and research humanoids (Dreamer, Draco, Hume, QDA/QDH/QDB). Late 2020: Astra ships commercially. June 2022: $14.6M seed, Apollo program begins. August 2023: Apollo unveiled. 2024: Mercedes pilot (March), GXO deal (June), Google DeepMind (December). February–March 2025: $350M → $403M Series A; Jabil manufacturing deal; Mercedes Berlin deployment goes live. Mid-2025: Apollo 2 and Robot Park. February 11, 2026: $520M extension, $5.3B valuation.

What people say

The case for. Mercedes-Benz board member Jörg Burzer (production & supply chain) publicly framed humanoids as "an exciting new frontier" and described employees transferring know-how to Apollo via teleoperation (Assembly Magazine, March 21, 2025). AT&T Ventures' Vikram Taneja called humanoid "physical AI" "a huge validation" of Apptronik's team and platform (Dallas Innovates, 2026). Trade coverage consistently points to Apptronik's academic pedigree, electric linear actuators, and modular form factor — the torso rebutting the "humanoid form is wrong" critique by running on legs, wheels, or a fixed base.

The complaints. Rodney Brooks, iRobot co-founder and a widely-cited industry voice, argued in September 2025 that humanoid robots are "doomed to burst" because the human form factor is a marketing choice rather than an engineering one, and because humanoids lack the hand dexterity to pay back their cost; he singled out the whole segment — Figure, Apptronik, Tesla — not any one company (TechCrunch, September 26, 2025, "Famed roboticist says humanoid robot bubble is doomed to burst"). Robotics-forum critiques echo Brooks on dexterity. The deeper concern is capital efficiency: Figure raised nearly four years of Apptronik's cumulative capital in one Series C round, and the humanoid category arms race is not won by the most technically elegant actuator.

Outlook: the open question

The answer depends on whether Apollo 2 hits its sub-$50K unit-cost target with meaningful uptime in production — not demo — settings before Figure's capital and Tesla's vertical scale set the category's price and performance floor, and before Unitree's Chinese price curve breaks the industrial economics for everyone. For the Apptronik thesis to be true, the Mercedes/GXO/Jabil pilots must convert to repeat orders at target unit cost, Gemini Robotics 2 must generalise outside the task library Apollo was trained on, and the company must move from "early stages of development" revenue to recurring deployments by 2027 or 2028. For the bear thesis to be true, Figure buys the same AI brain Apptronik rents (Google is an investor in Apptronik but has no exclusive) and outspends them into every pilot, or Tesla Optimus emerges from vertical manufacturing at half the price with good-enough reliability, or Unitree makes the sub-$50K target table stakes.

How to attack it

A new seed-stage team cannot credibly launch a competing humanoid — the fundraise bar is a nine-figure Series A and the IP stack takes a decade of academic work to replicate. The attackable wedges are adjacent components that every humanoid company will need and none of them wants to build from scratch. Dexterous end-effectors — the hand problem that Brooks identifies — is the cleanest: Apollo ships with 0–1 DoF grippers because Apptronik has not solved hands, and Figure's 20-DoF hands are the exception, not the rule. A hand that drops into Apollo or Agility's Digit via a standard flange would be a very valuable company. Humanoid-safe environments — perception, safety mats, soft-barrier hardware — are the second: factory floors are being rebuilt for humanoid co-work and nobody owns that layer. Teleoperation and training stacks — the actual pipeline Mercedes used to train Apollo — are the third; Scale AI has the data labelling lane, but nobody owns the teleop UX layer.

The exploitable weaknesses in Apptronik's position are specific. The hand gap — a four-finger gripper with 0–1 DoF is a wedge for anyone who ships a better hand compatible with Apollo. The dependency on Google DeepMind — not an exclusive partnership, so every competitor with cash can rent the same brain. The Austin recruiting base — strong but narrow versus Figure's Bay Area footprint. The absence of a Chinese supply chain at scale — if Unitree's price curve hits industrial reliability, Apollo's BOM math breaks.

Adjacent-segment play

Apollo's hardware stack — linear electric actuators, torque-feedback joints, modular torso — is the strongest asset to repackage outside the humanoid form. Exoskeletons, where Apptronik started (Sagittarius and Apex for USSOCOM) are a natural adjacency and Ekso Bionics has shown there is a real market in industrial lifting and medical rehab; Sarcos tried and failed, but with different economics and a better actuator the category is live. Fixed industrial manipulator arms priced below the $35–50K Universal Robots cobot range are the second: the same actuator stack with no legs, no vision, sold as a hardware component to a system integrator, would be a profitable standalone product — Apptronik effectively walks past this market because humanoids pay the venture math, but a competitor could occupy it fast.

Prosthetics, where Esper Bionics and Open Bionics have validated the demand, is a third — Apptronik's torque-feedback joint IP and hand development would land directly in a market with established insurance reimbursement. The thesis generalises because the underlying hardware is general-purpose; it fails when the integrated humanoid-plus-AI story is the only one Apptronik has the people and the channels to tell, and splitting attention across three adjacencies is exactly how good hardware companies fail.

Sources and further reading

• TechCrunch, "Apptronik raises $350M to build humanoid robots with help from Google," February 13, 2025 — https://techcrunch.com/2025/02/13/apptronik-raises-350m-to-build-humanoid-robots-with-help-from-google/
• TechCrunch, "Humanoid robot startup Apptronik has now raised $935M at a $5B valuation," February 11, 2026 — https://techcrunch.com/2026/02/11/humanoid-robot-startup-apptronik-has-now-raised-935m-at-a-5b-valuation/
• Bloomberg, "Humanoid Maker Apptronik Triples Valuation to Over $5.5 Billion With New Funding," February 11, 2026 — https://bloomberg.com/news/articles/2026-02-11/apptronik-raises-520-million-in-new-funding-to-build-more-humanoids
• The Robot Report, "Apptronik unveils Apollo humanoid robot," August 23, 2023 — https://www.therobotreport.com/apptronik-unveils-apollo-humanoid-robot/
• Assembly Magazine, "Mercedes-Benz Deploys Humanoids at Berlin Assembly Plant," March 21, 2025 — https://assemblymag.com/articles/99135-mercedes-benz-deploys-humanoids-at-berlin-assembly-plant
• UT McCombs News, "Apptronik Co-Founder Jeff Cardenas, BBA '08, MSTC '13, is Building Humanoid Robots," Spring 2024 — https://news.mccombs.utexas.edu/magazine/spring-2024-issue/year-of-ai-ai-in-action/
• Goldman Sachs Insights, "The global market for robots could reach $38 billion by 2035," February 27, 2024 — https://goldmansachs.com/insights/articles/the-global-market-for-robots-could-reach-38-billion-by-2035
• Morgan Stanley Insights, "Humanoids: A $5 Trillion Market," May 14, 2025 — https://www.morganstanley.com/insights/articles/humanoid-robot-market-5-trillion-by-2050
• Apptronik press release, "Apptronik Partners with Google DeepMind Robotics," December 19, 2024 — https://apptronik.com/news-collection/apptronik-partners-with-google-deepmind-robotics
• DC Velocity, "GXO begins R&D trial of humanoid robot from Apptronik," June 20, 2024 — https://www.dcvelocity.com/articles/61111-gxo-begins-r-and-d-trial-of-humanoid-robot-from-apptronik
• TechCrunch, "Famed roboticist says humanoid robot bubble is doomed to burst," September 26, 2025 — https://techcrunch.com/2025/09/26/rodney-brooks-humanoid-robots-doomed/