Robot brains: which startup is ahead?
In our humanoid robotics market deck, you will find everything you need to understand the market
SUMMARY
FieldAI is currently ahead in the robot brain race, with Sereact close behind and no runaway winner.
The ranking changes depending on what “ahead” means. FieldAI leads overall commercial autonomy, Sereact leads repeated production manipulation, Physical Intelligence has the strongest public technical case, and Skild has built the most powerful route to future scale.
FieldAI’s edge comes from where its systems work, not just how many robots it has shown. Construction sites, mines, energy facilities, data centers, and defense environments produce harder autonomy problems than controlled workstations and give its commercial traction more weight.
Sereact has the deepest disclosed operating history. More than 200 live systems, over one billion production picks, named customers, and a reported remote-intervention rate of roughly 19 per million picks make its maturity difficult to dismiss.
The market is split between mobile autonomy and manipulation. FieldAI’s robots move through changing industrial sites, while Sereact, Dyna, Physical Intelligence, Generalist, and Genesis focus more heavily on arms, hands, picking, folding, and other physical tasks.
Physical Intelligence currently offers the strongest combination of model progression, cross-robot transfer, published research, open tooling, and partner testing. Generalist AI has stronger headline benchmark numbers, but those results still come mainly from its own evaluation setup.
Skild is the best-funded company and has the strongest potential distribution network through major robotics platforms and its acquired Zebra operation. That is a serious advantage, but partnerships and capital are still leading indicators; FieldAI and Sereact have stronger disclosed proof of paid use.
Sereact also gives customers the clearest economics. Throughput, interventions, labor replacement, installation time, and picking costs can be measured directly, while the broader foundation-model companies still require buyers to fund hardware, integration, task data, safety work, and support.
The data moats are different. Sereact owns the clearest production-data loop, Generalist reports the most physical-interaction training hours, and FieldAI gathers rarer failure cases from unstable real-world environments that may be harder to reproduce.
Google DeepMind, NVIDIA, Amazon, and vertically integrated robot manufacturers can pressure every startup in this ranking. The startups with the best defenses are already surrounding their models with deployment data, customer workflows, distribution, or measurable cost advantages.
This market map, featured in our humanoid robotics market deck, highlights top companies and startups in the humanoid robotics market
Which robot brain startups should we actually compare?
Seven startups currently belong in a serious robot brain comparison: Skild AI, Physical Intelligence, FieldAI, Generalist AI, Sereact, Dyna Robotics, and Genesis AI.
We use “robot brain” to mean a learned intelligence layer that turns images, instructions, and sensor data into physical actions. The model should work across several tasks, environments, or robot designs. A conventional warehouse robot programmed for one fixed movement does not qualify.
We also keep the field narrower than the whole physical AI industry. Figure, 1X, Sunday Robotics, and Apptronik develop impressive intelligence, but their models mainly power their own robots. Covariant left the independent field after Amazon hired its founders and licensed its models. Google DeepMind, NVIDIA, and Amazon belong in the competitive-threat discussion rather than the startup ranking.
The seven selected companies fall into three broad groups. Skild, Physical Intelligence, and Generalist are trying to build foundation models that could become the intelligence layer for many robot manufacturers. FieldAI focuses on autonomy across mobile robots in unpredictable industrial environments. Sereact and Dyna began with commercially useful manipulation tasks, while Genesis is combining its model, hands, and robot into one full system.
Funding totals remain imperfect because private companies disclose rounds differently. We count completed financing and exclude Physical Intelligence’s reported discussions for another $1 billion because that deal had not been confirmed as closed.
| Startup | What it is building | Disclosed funding |
|---|---|---|
| Skild AI | An “omni-bodied” brain for arms, humanoids, quadrupeds, mobile robots, and industrial machines | About $1.84B |
| Physical Intelligence | General-purpose models that learn manipulation across tasks, robots, and environments | About $1.07B |
| Generalist AI | Large embodied foundation models trained directly on physical interaction data | More than $500M |
| FieldAI | Risk-aware autonomy for mobile robots operating in construction, energy, mining, inspection, and defense | $405M |
| Dyna Robotics | Foundation-model-powered dual-arm robots for repetitive commercial manipulation | $143.5M |
| Sereact | The Cortex brain for picking, sorting, returns, manufacturing, and related manipulation workflows | More than $140M |
| Genesis AI | The GENE model, dexterous hands, and the Eno general-purpose robot | $105M |
Is one robot brain startup clearly ahead right now?
FieldAI leads the robot brain market today, with Sereact close enough behind that we see a narrow top two rather than one runaway winner.
FieldAI has the best combination of commercial value, customer breadth, robot variety, and difficult real-world deployment. The company has passed $100 million in recognized revenue and customer contracts from more than 30 clients across the United States, Europe, and Asia, according to figures reported by Business Insider. Its models operate on several kinds of mobile robots in construction sites, mines, data centers, energy facilities, and defense environments.
Those environments give FieldAI’s traction extra weight. A robot inside a changing construction site cannot depend on a perfectly arranged workstation, a fixed object list, or an engineer standing nearby to reset the scene. Layouts change, people and machines move, communications fail, and mistakes can damage expensive equipment.
Sereact is ahead wherever repeated production work carries the most weight. Its latest financing announcement reported more than 200 live systems and over one billion production picks for customers including BMW, Mercedes-Benz, Daimler Truck, PepsiCo, Austrian Post, bol., and Active Ants. That is the deepest disclosed operating history among independent robot brain startups.
Sereact’s evidence remains concentrated in warehouses and factories, while FieldAI works across a wider range of industries and robot bodies. That breadth gives FieldAI first place overall.
Physical Intelligence probably has the strongest publicly documented general-purpose manipulation model. Skild has the biggest budget, the widest stated ambition, and the most powerful industrial relationships. Both remain behind FieldAI and Sereact in disclosed paid usage.
The picture is fairly clear. FieldAI leads overall, Sereact leads production manipulation, Physical Intelligence leads transparent generalist-model research, and Skild has built the strongest platform for a future mass rollout.
If you want more recent data on this point, please see our latest humanoid robotics market report.
As this chart shows, and as featured in our humanoid robotics market deck, search interest in where to buy robots has been rising steadily
Who has turned a robot brain into the strongest real business?
FieldAI has built the strongest robot brain business so far, while Sereact has produced the clearest proof that customers keep using its product after installation.
FieldAI’s $100 million milestone combines revenue already recognized with signed customer contracts, so it should not be read as annual recurring revenue. Even with that qualification, no close competitor has disclosed a comparable monetary figure. More than 30 clients also give FieldAI a broader commercial base than one large design partner or a handful of experimental deployments.
The average recognized and contracted value works out to more than $3 million per client, although the actual contracts will vary widely. The useful part is the order of magnitude: FieldAI is signing industrial relationships worth millions rather than running scattered low-cost trials.
The company also says several customers have expanded from initial deployments into longer-term production agreements. Its work with Boston Dynamics has followed the same pattern. The relationship developed from technical integration into joint commercial work around autonomous construction and industrial operations.
Customer-level revenue, renewal rates, and contract lengths are still missing. Even so, FieldAI gives us enough evidence to see a real business forming around its autonomy platform.
Sereact provides less financial disclosure and stronger operating detail. Its named customers span automotive manufacturing, postal logistics, e-commerce, food, and third-party fulfillment. Several have publicly discussed throughput, SKU diversity, or rollout plans.
Active Ants compared multiple suppliers before choosing Sereact. Bol. placed the technology in an operation handling more than 500,000 different products. These are demanding environments where the robot must cope with inventory changes rather than repeat the same rehearsed sequence.
Physical Intelligence is beginning to show a plausible business model through deployment partners. Ultra ran π0.6 for a full customer shift at 96.4% autonomy while packaging real orders. Weave reported that adding its laundry data to pretraining cut missed grasp sequences by 42% and interventions by 50%.
Those deployments show that Physical Intelligence can improve a commercial service. They remain much smaller than the operating bases of FieldAI and Sereact.
Skild’s commercial position is harder to judge. ABB, Universal Robots, NVIDIA, Foxconn, and the acquired Zebra robotics operation could open major distribution channels. Skild has yet to disclose how many end customers pay for its brain, how frequently they use it, or whether early installations are expanding.
Which robot brain product is mature enough for daily work?
Sereact has the most mature robot brain product currently available. Its software already handles large volumes of customer work with a measurable support burden.
Sereact says only about one pick in 53,000 requires remote human help. That works out to roughly 19 remote interventions per million picks. The number comes from Sereact rather than an independent audit, but it is still more informative than a short demo success rate because it covers ongoing operations across many installations.
Warehouse picking sounds simple until the robot encounters crushed packages, transparent plastic, reflective objects, tangled products, unusual shapes, or a badly positioned item. Production maturity means handling these variations for months without turning each exception into an engineering project.
Dyna is the next most convincing product for a narrow commercial task. DYNA-1 completed a 24-hour napkin-folding test at a reported 99.4% success rate and around 60% of human throughput. The company now says its systems operate in laundromats, restaurants, factories, and other commercial environments.
The missing number is fleet size, and it is a big one. Dyna has shown that the product can work at customer sites, but it has not shown how widely the product has spread.
Physical Intelligence sits one step earlier. Its models are doing real work through Ultra and Weave, although human intervention remains part of the service. That is understandable with deformable clothes, variable packaging, and other tasks where small visual differences can change the correct movement. It also places the product closer to limited production than fully hands-off automation.
FieldAI is mature for mobile autonomy rather than dexterous manipulation. Its robots navigate changing industrial sites without relying on GPS, prior maps, or preplanned routes. Customers use the systems for inspection, site capture, mapping, and documentation, where persistent autonomous coverage can already replace repetitive human work.
Generalist AI has reported approximately 99% success across selected internal tasks, including long sequences of repeated trials. Those results look strong, but no named customer has published equivalent production performance.
Genesis remains earlier. The company plans targeted Eno deployments with customers before the end of the year.
If you want more recent data on this point, please see our latest humanoid robotics market report.
This chart, featured in our humanoid robotics market deck, illustrates yearly venture capital funding for humanoid robotics startups
Which startup has built the smartest general-purpose robot brain?
Physical Intelligence currently has the strongest public case for the smartest general-purpose robot brain, although the industry still lacks a neutral benchmark that can produce an undisputed winner.
Physical Intelligence has released several model generations rather than relying on one polished demonstration. π0 established a generalist vision-language-action architecture. π0.5 improved generalization to unfamiliar environments. π0.6 learned from autonomous experience and human corrections. π0.7 added richer control over language instructions, visual subgoals, speed, and task quality.
π0.7 can combine learned skills, follow new instructions, and transfer behavior across robot embodiments without requiring a separate specialist policy for every task. The company has also published papers, released code through OpenPI, and tested models through operating partners. That makes its technical claims easier to examine than most.
Generalist AI has published the strongest headline benchmark. GEN-1 reportedly increased average success from 64% to 99%, ran up to three times faster than previous methods, and learned each evaluated task from approximately one hour of robot-specific data.
Generalist also showed hundreds or thousands of consecutive attempts on several tasks. That is better evidence than a compilation of successful clips because long runs reveal whether the model fails once every ten attempts or once every thousand.
The company designed the evaluation itself, and customer results have not yet confirmed the same performance outside its test setup. For now, GEN-1 is a serious technical challenger rather than the leader.
Sereact’s Cortex 2.5 offers a more practical form of adaptability. The latest model can learn a nearby task from two or three demonstration videos without changing its model weights. In Sereact’s early tests, this approach reached 92.6% success compared with 97.6% for a fully trained baseline.
Sereact clearly says the tests cover short tasks within a familiar domain. That limitation helps us understand what Cortex 2.5 can do now rather than imagining that it can learn any physical task from three videos.
Genesis has demonstrated a wide collection of visually difficult hand tasks, including cooking a multi-step meal, wire harnessing, laboratory work, solving a Rubik’s Cube, preparing a smoothie, and playing piano. These demonstrations show impressive dexterity.
Genesis has not published standardized success rates, failure counts, or long-duration results. The videos establish range without proving that GENE-26.5 can match PI or GEN-1 in repeatable performance.
Skild may eventually prove broader than all three. Its public evidence currently focuses more on cross-body versatility and industrial demonstrations than directly comparable task-success data.
We rank Physical Intelligence first technically because the company has built the strongest combination of model progression, research detail, cross-embodiment transfer, released tooling, and real partner testing.
Which robot brain works across the widest range of robots?
Skild AI covers the widest claimed range of robot bodies, while FieldAI has stronger evidence across real commercial machines.
Skild says it trained its brain across 100,000 simulated robot designs, covering quadrupeds, humanoids, arms, mobile manipulators, and other machines. Physical Intelligence has shown better-documented transfer across manipulation setups, while FieldAI has deployed one autonomy stack across drones, rovers, quadrupeds, and other mobile systems.
Skild leads on claimed embodiment breadth. FieldAI leads on demonstrated variety in customer environments. Physical Intelligence remains strongest across manipulation behaviors.
This chart, featured in our humanoid robotics market deck, shows how Agility Robotics is capturing share in humanoid robotics
Who has the best robot data advantage?
Sereact has the clearest production-data advantage, although Generalist and FieldAI own different kinds of valuable data.
Sereact’s later financing announcement reported more than one billion production picks, while its homepage still displayed more than 500 million. We use the later figure cautiously because the company’s own pages are inconsistent.
Even the lower number represents a powerful data loop. Sereact records real objects, failed grasps, recoveries, throughput pressure, and unusual packaging across paying operations. Generalist reports 500,000 hours of physical interaction data, while FieldAI collects rarer examples from unstable industrial environments.
Sereact currently owns the most proven production dataset. Generalist leads on disclosed training hours, and FieldAI’s data may be harder for a competitor to reproduce.
Which robot brain gives customers the clearest financial return?
Sereact currently gives customers the clearest robot brain economics because it connects throughput, labor replacement, deployment speed, and cost savings in one measurable product.
Sereact says one system can replace between three and 4.5 full-time roles depending on shift patterns. It claims savings of up to €111,000 per month, typical throughput of 300 to 450 units per hour, and peaks above 600 units per hour with optimized feeding.
The company also says installations can begin producing within a day and reduce picking costs by as much as 77%. These are vendor claims, and the actual result will vary with local wages, facility design, product mix, and utilization.
Those claims are more useful than broad promises about future robotic intelligence. A warehouse operator can measure units per hour, interventions, errors, uptime, and labor cost before and after installation.
Dyna has a plausible economic story in laundries, restaurants, and hospitality. Repetitive fabric handling is difficult to staff, and a machine working longer shifts can spread its hardware cost across more productive hours. Dyna currently publishes reliability and customer stories without providing prices, payback periods, or labor savings.
FieldAI creates value differently. Its robots can inspect sites, capture images, update digital models, and document progress during hours when human teams are unavailable. The size of its signed business suggests that customers see a financial return, but FieldAI does not publish enough pricing or productivity data for us to calculate it.
Physical Intelligence, Skild, Generalist, and Genesis are selling a broader future. Their models may eventually lower the cost of building many robot applications. Today, the customer still needs hardware, integration, task data, safety work, and technical support.
That puts Sereact first here: a buyer can connect the robot’s actions to a cost line immediately.
If you want more recent data on this point, please see our latest humanoid robotics market report.
This chart, featured in our humanoid robotics market deck, illustrates yearly funding for humanoid robotics startups
Which robot brain startup can scale to thousands of deployments?
Skild has assembled the strongest route toward thousands of deployments, while Sereact has already scaled further in systems we can count.
Skild can distribute its model through ABB, Universal Robots, Mobile Industrial Robots, and the former Zebra robotics business. Sereact uses standard industrial hardware and already supports more than 200 live systems. FieldAI can expand through existing platforms such as Boston Dynamics’ Spot without manufacturing complete robots.
Skild has the best potential distribution network. Sereact still owns the strongest proven rollout system.
Is the best-funded robot brain startup also the leader?
Skild is the best-funded robot brain startup by a wide margin, but its commercial evidence still trails FieldAI and Sereact.
Skild has raised about 4.5 times as much capital as FieldAI and roughly 13 times as much as either Sereact or Dyna. That money has produced major partnerships, training infrastructure, industrial integrations, international expansion, and the acquisition of Zebra’s robotics operation.
The company has yet to connect that spending to disclosed revenue, paying-customer counts, or a large active fleet running Skild Brain. Funding gives Skild the most options, but not first place today.
If you want more recent data on this point, please see our latest humanoid robotics market report.
This chart, featured in our humanoid robotics market deck, compares the main business model options for humanoid robot manufacturers
Which robot brain startup has gained the most ground lately?
Sereact and Skild have gained the most ground lately, with Sereact reducing commercial risk and Skild radically expanding its route to market.
Sereact raised $110 million, expanded into the United States, developed Cortex 2.0 around planning and imagined future outcomes, and then released Cortex 2.5. The latest version teaches nearby tasks from a handful of videos without requiring a full new training run.
The pieces reinforce each other: funding supports expansion, deployments create data, and the data improves the model. Sereact is strengthening several parts of the business at once rather than producing disconnected announcements.
Skild’s progress has been more corporate and industrial. The company closed a $1.4 billion round, announced integrations with three major robotics platforms, demonstrated a Blackwell assembly workflow with NVIDIA and Foxconn, and acquired Zebra’s robotics division.
Skild now has hardware access, customer channels, deployment staff, and enough capital to subsidize early adoption. Its next proof point needs to come from measured customer operations. Another partnership would add less information than an active-fleet number or disclosed repeat usage.
Generalist made the fastest research leap. GEN-0 established the company’s scaling argument, GEN-1 reported a sharp improvement in reliability and speed, and a new $400 million round brought total financing above $500 million. External production evidence remains the missing piece.
Genesis also moved quickly from stealth to a model, dexterous hand, general-purpose robot, senior operations hires, and a partnership with LG CNS. The company is still approaching its first targeted deployments, which keeps it below companies already operating at customer sites.
FieldAI publishes fewer product releases and customer names, so its recent momentum is less visible. We see FieldAI as the current leader holding its position, while Sereact and Skild are the challengers changing their competitive positions fastest.
Can Google, NVIDIA, and robot manufacturers crush these startups?
Google and NVIDIA can put serious pressure on robot brain startups because both already offer technology that overlaps with the startups’ core products.
Google DeepMind’s Gemini Robotics family combines spatial reasoning, task planning, multimodal understanding, and robot control. Gemini Robotics-ER 1.6 can act as a high-level brain that breaks down jobs, checks whether actions succeeded, and decides when a robot should retry.
Google can connect these capabilities to Gemini’s broader research, software, compute, and developer ecosystem. It also works with robot manufacturers including Apptronik, Agility Robotics, Boston Dynamics, Agile Robots, and Enchanted Tools.
NVIDIA covers even more of the development stack. Isaac GR00T includes foundation models, training data, simulation, middleware, runtime libraries, and on-robot computing. GR00T 1.7 is openly available with commercial licensing, giving robot manufacturers a credible starting point without building a foundation model from scratch.
Vertically integrated robot companies create another threat. Figure’s Helix 02 controls walking, balance, perception, and manipulation through one neural system. Figure can collect data directly from its own humanoids and optimize the model around one hardware platform.
Amazon has another unusual advantage: millions of hours of fulfillment activity and one of the world’s largest robot fleets. Its acquisition of Covariant talent and models gave it additional expertise in learned manipulation.
Independent startups can still win because many robot manufacturers will hesitate to let Google or NVIDIA own the most valuable layer of their product. A neutral provider such as Physical Intelligence or Skild can serve several hardware companies without competing against them in robot manufacturing.
FieldAI and Sereact have another defense. Both understand specific customer environments deeply and already sit inside operational workflows. Replacing them would involve more than swapping one model API for another.
The most vulnerable startup would offer a closed general model with no proprietary deployment data, no distribution, and no measurable cost advantage. The strongest companies are surrounding their models with assets that are harder to copy.
FieldAI has industrial deployments and risk-aware autonomy. Sereact has production data and workflow integration. Physical Intelligence has research credibility and an open developer ecosystem. Skild has capital, OEM access, and acquired distribution.
This chart, featured in our humanoid robotics market deck, shows the revenue mix across customer segments in the humanoid robotics market
Which robot brain startups are actually ahead?
FieldAI is ahead overall now, Sereact is the closest challenger, and the gap after those two remains meaningful.
We give FieldAI first place because the company has moved across more industries, robot types, countries, and difficult operating environments while also producing the strongest disclosed commercial value.
Its lead over Sereact is narrow. A large Sereact expansion beyond warehouse and factory manipulation could reverse the order.
Sereact takes second because its evidence is unusually difficult to dismiss. Robots run daily, customers are named, intervention frequency is measurable, and buyers can calculate throughput and labor savings. We place Sereact below FieldAI because its operating history remains concentrated in a narrower family of tasks.
Physical Intelligence ranks third. π0.7 currently offers the best blend of technical depth, generalist manipulation, published research, cross-embodiment transfer, and early commercial use.
Physical Intelligence can move into the top two when partners scale from small fleets to hundreds or thousands of robots without losing reliability.
Skild ranks fourth. Its long-term ceiling may be the highest in the group, and no competitor has built a comparable combination of funding, industrial partners, hardware access, and acquisition capacity.
Today, Skild asks us to infer commercial leadership from infrastructure and partnerships. The company needs active-fleet numbers, paying end customers, repeat usage, and revenue before it can rank higher.
Dyna ranks fifth because it has chosen real production before broad claims. Its robot works on selected commercial tasks with high reported reliability, although the disclosed fleet and customer base remain small.
Generalist ranks sixth despite excellent technical results. GEN-1 could be one of the strongest models in the group, but internal tests cannot carry the same weight as customer production. Generalist could climb quickly once outside companies publish deployment results.
Genesis ranks seventh for now. The company has assembled an ambitious full stack and shown remarkable dexterity, but meaningful production use is still beginning. Its position reflects maturity rather than weak technology.
So the market has a narrow top two, two powerful frontier challengers, and three earlier companies with credible ways to move up. FieldAI leads the whole robot brain race. Sereact leads repeated paid manipulation. Physical Intelligence leads open technical evidence. Skild is best positioned to turn capital and distribution into a much larger platform.
| Rank | Startup | Why it is here |
|---|---|---|
| 1 | FieldAI | Broadest mix of commercial traction, robot coverage, global customers, and operation in difficult environments |
| 2 | Sereact | Strongest production history, customer proof, reliability evidence, and measurable economics |
| 3 | Physical Intelligence | Best publicly documented generalist manipulation technology and strongest open research position |
| 4 | Skild AI | Most powerful funding and distribution position, with operational evidence still catching up |
| 5 | Dyna Robotics | Real customer work and strong narrow-task reliability, but limited disclosed scale |
| 6 | Generalist AI | Excellent internal model results and data scale, with customer proof still missing |
| 7 | Genesis AI | Impressive full-stack technology that remains before meaningful production deployment |
If you want more recent data on this point, please see our latest humanoid robotics market report.
OUR METHODOLOGY
This analysis asks which independent robot brain startup is ahead based on the evidence available now. We separate the question into commercial adoption, deployment maturity, technical performance, robot coverage, data advantage, customer economics, funding, distribution, and competitive risk.
We use “robot brain” to mean a learned intelligence layer that turns images, instructions, and sensor data into physical actions across several tasks, environments, or robot designs. We compare Skild AI, Physical Intelligence, FieldAI, Generalist AI, Sereact, Dyna Robotics, and Genesis AI because each is building an intelligence layer intended to work beyond one fixed robot task.
We kept vertically integrated robot manufacturers such as Figure, 1X, Sunday Robotics, and Apptronik outside the startup ranking because their models primarily power their own machines. Google DeepMind, NVIDIA, and Amazon are treated as competitive threats rather than ranked startups.
We weighted evidence closest to sustained real-world use most heavily: live systems, named customers, repeated operations, measurable intervention rates, throughput, contract value, external deployment partners, and performance observed over meaningful periods. A polished demo, one internal benchmark, one partnership, or one funding round could strengthen a case, but it could not decide the ranking by itself.
Commercial figures are interpreted according to what they actually measure. FieldAI’s reported $100 million milestone combines recognized revenue and signed contracts rather than annual recurring revenue. Sereact’s throughput, savings, intervention, and production-pick figures are treated as company-reported operating evidence rather than independently audited results.
For technical comparisons, we prioritized model progression, published methods, repeatable evaluations, cross-embodiment transfer, released tooling, long-run trials, and results from outside deployment partners. This is why Physical Intelligence ranks ahead technically despite Generalist AI’s stronger headline benchmark and Genesis AI’s wider set of dexterity demonstrations.
Funding totals include completed financing that the companies had disclosed or formally announced. We excluded reported fundraising discussions that had not closed. Distribution potential is assessed separately from proven deployment scale, which prevents Skild’s capital and partnerships from being treated as equivalent to an active customer fleet.
The final ranking aggregates the dimensions rather than averaging them mechanically. FieldAI ranks first because it combines the broadest difficult deployments with the strongest disclosed commercial value; Sereact ranks second because it has the deepest production history and clearest customer economics. The ranking distinguishes leadership today from the capacity to move ahead later.
Key sources include FieldAI’s company overview and news archive; Sereact’s Series B announcement, product-performance page, and Kardex rollout announcement; Skild AI’s Series C announcement, industrial-partnership update, Zebra robotics acquisition announcement, and omni-bodied model description; Physical Intelligence’s OpenPI repository; Generalist AI’s GEN-1 results and data description; Genesis AI’s GENE-26.5 technical overview and Eno deployment announcement; Google DeepMind’s Gemini Robotics documentation; and NVIDIA’s Isaac GR00T platform.
This chart, featured in our humanoid robotics market deck, shows how factory humanoid robot technology has evolved over time
Related blog posts
- Which robot hand startup is making the most progress?
- Which startup is making humanoid robots truly affordable?
- Who has the most advanced humanoid robot?
