Guide: The Great AI Robot Race (2020–2025)

Recently at SiliconSnark, we covered the delightful spectacle of Figure.ai – one of the shiny new humanoid robotics startups – deciding to sue Fortune for allegedly harsh coverage. Yes, nothing says “the robots are coming to save us” quite like the humans behind them reaching for the litigation manual.

Watching Figure lawyer up made me realize that it’s time to step back and take a proper look at the whole AI-powered robot race.

If you zoom out a little, what’s happening isn’t just one startup getting salty over some bad press. It’s a five-year arms race between tech giants, well-funded startups, secretive labs, and questionable PowerPoint decks – all scrambling to build the first truly useful humanoid robot. And along the way, there’s been dancing robots, lawsuits, billion-dollar funding rounds, corporate subtweet wars, vaporware demos, glorious faceplants, and yes, some real progress too.

This guide tracks the chaotic, hilarious, and occasionally impressive evolution of the AI robot race since about 2020 — year by year, feud by feud, and hype cycle by hype cycle — with plenty of SiliconSnark-worthy commentary along the way.
Because if the robots are going to inherit the Earth, the least we can do is roast their startup parents first.

2020: Setting the Stage (and Dancing into 2021)

• Boston Dynamics Steals the Spotlight: Long the reigning champ of legged robots, Boston Dynamics rang in the new year with a viral video of its robots busting out dance moves to The Contours’ “Do You Love Me?”​ The humanoid Atlas and robo-dog Spot showed off parkour-like coordination, prompting reactions ranging from awe to apocalyptic (“We’re doomed”). It was the first big robot spectacle since Hyundai announced it was acquiring Boston Dynamics from SoftBank in December 2020, valuing the firm at about $1.1 billion. Hyundai’s purchase (80% stake for $880 million) signaled serious automotive interest in robots, and perhaps a bid to one-up Elon Musk’s other high-tech ventures. Boston Dynamics, now flush with Hyundai money, continued developing its impressive but pricey machines – and inadvertently set a high bar for coolness that future entrants would struggle to meet.
• The Old Guard Stumbles: Not all robots were having a ball. SoftBank’s friendly humanoid Pepper, once the poster child for social robots, quietly fell from grace. By mid-2021 SoftBank halted Pepper’s production due to weak demand, after only 27,000 units sold​. It turns out over-promising a robo-greeter for every home and shop was, well, awkward. Pepper’s flop foreshadowed a key theme of the coming race: grand visions colliding with harsh reality (and low ROI). Meanwhile, other legacy projects like Honda’s Asimo (retired a few years prior) and Toyota’s prototype humanoids remained mostly science projects. The stage was set for a new generation of ambitious upstarts who smelled a gap in the market – and an opportunity to do it right (or so they thought).

2021: Enter Elon (and the Onstage Dancer)

• Tesla’s Jaw-Dropping, Err, Suit-Wearing “Prototype”: Leave it to Elon Musk to upend the narrative. In August 2021 at Tesla’s AI Day, Musk announced Tesla Bot (codename Optimus), promising a 5’8” humanoid that could automate “boring, repetitious and dangerous” tasks. The catch? Tesla had no robot yet – so a spandex-clad dancer was sent on stage to bust moves in lieu of a prototype. Yes, seriously: instead of an actual robot demo, an actor in a white bodysuit breakdanced to techno beats. Musk quipped a real prototype would arrive “next year.” The tech press and robotics experts had a field day. Was this bold vision or pure vaporware? Musk’s claims that “everyone will want one” and these bots could cost under $30k stoked hype among fans and skepticism among engineers. Boston Dynamics’ CEO Robert Playter dryly noted that Tesla’s entrance “brought a bright light” to humanoids and “ignited a competitive spirit” at his company. In other words: Game on, Elon. Tesla had thrown down the gauntlet (or rather, the dance suit), and suddenly the sleepy field of humanoid robotics got a jolt of Silicon Valley adrenaline – and hubris.
• Stealthy Startups and Lab Projects: While Tesla grabbed headlines, 2021 also saw quiet progress from others. Austin-based Apptronik, a spinout from the University of Texas’s humanoid lab, was working on advanced robotic limbs and exoskeleton tech with NASA roots, laying groundwork for a future humanoid. Sanctuary AI, a Vancouver startup founded in 2018 by D-Wave co-founder Geordie Rose, emerged from stealth with lofty ideals: creating “the world’s first human-like intelligence in general-purpose robots” to tackle labor shortages on Earth (and eventually in space, no less). In practice, Sanctuary was still developing its robot and “cognitive architecture” modeled after the human brain, but it had assembled an impressive team of AI and robotics veterans (including co-founders of Kindred, another robotics firm)​. By late 2021, early demos of Sanctuary’s system showed teleoperated robots doing simple tasks, hinting at a fusion of human pilot and AI control. Meanwhile, 1X Technologies (formerly Halodi Robotics of Norway) was quietly testing humanoid security robots on wheels, and other players across the globe were taking notes. The starting pistol had fired – even if some racers hadn’t publicly revealed their entry yet.

2022: Funding Frenzy and First Prototypes

• Money Flows to Metal Bodies: 2022 became a fundraising bonanza for humanoid robot ventures. In April, Oregon-based Agility Robotics – makers of the bipedal “Digit” robot – announced a whopping $150 million Series B round​. The round was led by DCVC and Playground Global, with Amazon’s Industrial Innovation Fund hopping onboard, signaling e-commerce giant Amazon’s interest in robots that could one day tote packages in warehouses. Agility, an OSU spinout, had already sold a few of its bird-legged Digit robots to Ford and others, and now it doubled down on scaling production. Not to be outdone, Sanctuary AI secured a hefty C$75.5M (~US$58.5M) Series A in March, backed by names like Bell, Verizon Ventures, and even a Canadian government fund. Sanctuary made waves touting its mission of “human-like intelligence” in a human-shaped body​ – ambitious wording that raised eyebrows but loosened investors’ purse strings. Over in Palo Alto, a tiny startup called Kindred (Kind) Humanoid was born, led by ex-Google researcher Christoph Kohstall. Kind set out to build a low-cost humanoid on a scrappy budget (their makeshift lab famously used an $80 coat rack as a robot gantry)​. And in stealth mode, entrepreneur Brett Adcock (fresh off selling his flying taxi startup) self-funded $100 million to launch Figure in mid-2022​ – recruiting top talent from institutions like Boston Dynamics, Tesla, and Apple to develop a “general purpose” humanoid. In short, 2022 saw a gold rush: everyone with a blueprint (or a cool render) for an AI-driven robot could find a willing VC.
• From PowerPoint to Prototype: Hype began translating into hardware. At Tesla AI Day in September 2022, Elon Musk finally unveiled a real Optimus prototype – a clunky, unpainted robot that managed to walk a few shaky steps and wave to the crowd​. It was far from the sleek dancer envisioned, but it was a tangible robot. Tesla also rolled out a polished (non-walking) concept model, claiming these bots would revolutionize manufacturing soon. Skeptics noted the demo’s “work in progress” nature – the robot needed external cables for support – but Musk beamed about the “a lot of progress” made. Meanwhile, Agility Robotics introduced Digit 2.0 at a trade show, sporting a new animated LED face and rudimentary hands to better fit into warehouse work​. The design was deliberately minimalist (no uncanny valley here) – just enough head and eye motion to signal intentions to nearby humans. Agility promised its first beta Digits to early customers in 2023​, focusing squarely on dull logistics tasks (like moving bins) rather than sci-fi AI smarts. On another front, Xiaomi (yes, the Chinese phone maker) randomly unveiled a full-size humanoid named CyberOne in August 2022, complete with 177 cm height, 52 kg weight, and the ability to slowly shuffle on stage. Xiaomi’s CEO introduced CyberOne as a vision of “AI and mechanical capability,” though observers wondered if it was more publicity stunt than product. (The robot’s awkward gait and Xiaomi’s quick denial of any near-term mass production gave off a whiff of vaporware.) Nonetheless, CyberOne proved the robot race had gone global: even consumer electronics giants and Chinese startups wanted in on the action.

2023: Rise of the Robo-Startups

• New Contenders Emerge: If 2022 was the seeding round, 2023 was the breakout. In early 2023, Brett Adcock’s Figure finally came out of stealth, declaring it was building a general-purpose humanoid called Figure 01 to solve labor shortages. By May, Figure raised a $70 million Series A​ (on top of Adcock’s $100M personal investment), indicating serious confidence. Despite having only rendered images publicly until then, Figure hinted it had a working prototype that “took its first step” within a year – supposedly the fastest timeline in the industry​. The company aggressively hired experts and touted its “AI-first” approach (though skeptics noted that actual timelines and capabilities were still fuzzy). Around the same time, Norway’s 1X (Halodi) grabbed headlines by securing investment from OpenAI’s startup fund and Tiger Global, part of a ~$23 million round in March. By January 2024, 1X had raised a total of $100 million​, with backers including OpenAI itself – a clear sign that the AI crowd sees humanoid robots as the next frontier​. 1X leveraged its experience building a human-sized rolling android (used in security pilot projects) and announced it was developing a bipedal model named NEO. Over in Texas, Apptronik – the quiet veteran – made a splash in August by unveiling Apollo, a 5’8” 160-pound humanoid built for real work. The Austin startup, born out of years of lab research, began pilots with Apollo and boldly targeted commercial availability by end of 2024. Apollo’s reveal showed a sleek blue-and-white robot designed for factory and warehouse tasks, and Apptronik touted how it had built 10+ prior robots (including NASA’s famous Valkyrie) to get here​. In short, 2023 saw a crowded starting line: at least half a dozen serious humanoid bots were now strutting in labs and teaser videos, each claiming to be the one to automate our world.
• Big Tech and Big Bucks: The surge in AI enthusiasm (thanks, ChatGPT) spilled into robotics. Nvidia’s CEO declared intelligent robots would be “the next wave” of AI, and suddenly every venture capitalist wanted a humanoid startup in their portfolio. By late 2023, rumors swirled that Figure was in talks to raise a monster $1.5 billion at a jaw-dropping $39 billion valuation​ – a number that elicited gasps and groans on tech Twitter. (For context, that’s nearly 40× Boston Dynamics’ valuation, for a company that hadn’t shipped a single robot yet.) Another startup, Foundation Labs, launched by a former fintech founder with a turbulent past, raised an $11 million seed to build humanoids with an aim to “automate GDP” itself​. Foundation’s founder, Sankaet Pathak, brazenly claimed his new AI model “exceeds what any autonomous vehicle’s perception can do” – a bold (some said absurd) statement that nonetheless showed the hype-o-meter was cranked to 11. Meanwhile, Sanctuary AI achieved a notable milestone in the summer: their 6th-generation humanoid (named Phoenix) successfully completed a week-long pilot in a retail store, autonomously (under human supervision) performing 110 different tasks like stocking, cleaning, and folding clothes​. Sanctuary proclaimed this the first deployment of a “human-like intelligence” in a commercial setting​ – a claim others in the field might dispute, but it was backed by real results. By year’s end, Sanctuary raised more funds and even earned a spot on TIME’s Best Inventions list, fueling its momentum. As 2023 closed, it became clear that we were at the dawn of a humanoid renaissance: prototypes were walking (or rolling), early pilot programs were underway, and investors were writing eye-popping checks. The challenge ahead would be turning these flashy demos and fat wallets into sustainable businesses – a task easier said than done.

2024: Clashes, Consolidation, and Coming of Age

• Rivalries Heat Up: With many players in the race, 2024 saw inevitable showdowns and snark. Tesla, after largely going quiet post-Optimus demo, released new footage of its Optimus robots in action – assembling themselves on production lines and doing basic pick-and-place routines. Musk bragged about Tesla’s prowess in scaling hardware, implicitly poking at startups still perfecting one prototype. In response, startups like Figure and Apptronik highlighted their focused progress: Figure’s humanoid achieved untethered walking and began testing tasks in a warehouse setting, and the company frequently posted updates on social media to prove it was more than just CG renders. Agility Robotics, not in the business of trash talk, opened its $20 million RoboFab factory in Oregon to start producing Digits en masse, aiming to deliver units to customers in logistics and e-commerce. The contrast in styles was evident: some, like Agility and Apptronik, were quietly executing (hitting manufacturing milestones, signing pilot deals with the likes of DHL and Mercedes), while others, like Tesla and certain well-funded startups, were talking big about conquering the world. This culture clash led to occasional drama – think sassy tweets and LinkedIn posts subtweeting the competition. When Figure’s CEO posted about surpassing “legacy” robots, Agility’s team subtly reminded folks that Digit had been picking up boxes in the real world while most others were still in CAD designs. In the background, the community of actual roboticists traded good-natured barbs: “Is it finally the year of the humanoid? Or just the year of humanoid PowerPoints?”
• Mergers and Shakeouts: An early sign that the crowded field might streamline came in late 2024. OpenAI-backed 1X announced it would acquire Kind Humanoid, one of the small Silicon Valley outfits building a humanoid​. Kind’s three-person team (and their Yves Béhar-designed robot concepts) folded into 1X​, marking a mini-consolidation in the industry. 1X touted that uniting with Kind would “accelerate humanoid development” and fit its mission to create “an abundance of labor through safe, intelligent humanoids." The move hinted at a future where only those with the strongest tech (or deepest pockets) survive – a common theme in past tech races (remember the dozens of autonomous vehicle startups that merged or died?). Despite consolidation whispers, new entrants still appeared: e.g. Mentee Robotics, an Israeli startup, emerged from stealth to unveil its MenteeBot prototype – a humanoid they claim can understand natural language commands via a built-in AI brain. Mentee touted its use of transformer-based Large Language Models for following spoken instructions and performing complex tasks. It also boasted of integrating locomotion and dexterity, with the bot balancing while carrying weight and using only camera vision (no fancy lidar)​. By Q1 2024, Mentee Robotics had raised $17 million and promised a production-ready bot in early 2025. The message was clear: despite a looming shakeout, the allure of humanoid robots was still attracting fresh blood and capital. The race was entering a new phase where the frontrunners had to prove themselves not just in labs or investor decks, but in the real world.

2025: The Rubber Meets the Road

• Robots in the Real World: Five years in, 2025 is shaping up to be a pivotal year – call it the first true road test for humanoid robots. Early in the year, Apptronik closed a massive $350 million Series A round led by VC firm B Capital, with none other than Alphabet’s Google joining in​. This war chest is earmarked to scale up production of Apollo robots for actual deployment in factories and warehouses. Apptronik’s CEO Jeff Cardenas declared 2025 the inflection point “to demonstrate useful work” with early customers, with full commercialization in 2026+ if all goes well​. Indeed, Apollo is slated for expanded pilots, including in automotive manufacturing and logistics. Figure AI too hit a milestone: in a partnership with BMW, Figure’s second-gen robot was tested on an assembly line in South Carolina, successfully inserting metal parts into car body fixtures as part of a trial. BMW’s production VP praised the “promising developments” and is evaluating how humanoids could safely integrate alongside human workers​. Seeing a humanoid like Figure 02 actually doing real factory tasks is a watershed moment for the industry. Meanwhile, Tesla claims it’s getting Optimus ready for internal use in Tesla’s own factories, though external validation remains scant. Agility Robotics has begun shipping its first Digits to paying customers, focusing on dull, repetitive warehouse jobs. Even Sanctuary AI pivoted from pure R&D to a “robots-as-a-service” model, preparing to lease out humanoids that can be teleoperated to perform various tasks for client companies while the AI learns (a pragmatic step to generate revenue as true autonomy matures). In short, after half a decade of flashy reveals, 2025 is about delivering something useful – however limited – to end users. This is the year the robot race graduates from demos to the real world.
• A Reality Check (and Hope): The frenzy has tempered into focused execution. The industry mood is cautiously optimistic, but also sober. Many challenges have become clear: robots need to handle long tail oddities of unstructured environments, battery life is still only hours, and no amount of AI hype substitutes for reliable hardware. As one CEO noted, making one robot dance or pick up a box is easy; making thousands of them work day-in, day-out is the real challenge. Yet there’s progress: prices for key components (actuators, sensors) are slowly coming down thanks to scale and innovations. AI advances – from vision to language models – are being integrated so robots can perceive and even take high-level instructions better than ever. The fever pitch of funding has cooled slightly (higher interest rates and a few missed milestones saw to that), but serious investors like corporate strategics (Google, Toyota, Amazon, etc.) and industrial automation firms are in it for the long haul. 2025 hasn’t delivered a robot maid for your home just yet, but on factory floors, in warehouses, and in experimental retail pilots, humanoid and AI-powered robots are incrementally proving their worth. The race isn’t over – in fact, the finish line isn’t even visible yet – but the leaders are emerging, the pretenders are shaking out, and we’re finally seeing these ambitious robots step off the PowerPoint slides and into the workforce.

Meet the Players: Who’s Who in Humanoid & AI Robotics

The timeline introduced the major teams in this robotic relay. Now let’s dive deeper into the key companies and projects, complete with their backstories, accomplishments, and occasional faceplants. Think of it as a roster of our robo-contenders:

Figure AI

Founded: 2022 in California by Brett Adcock (after exits in hiring and flying taxis).
Mission: “A general purpose humanoid to free up labor” – essentially, a robot worker that can do a bit of everything. Figure unabashedly positions itself as the Tesla of robotics, aiming for mass-market humanoids that are affordable and useful across industries. Adcock even self-funded an initial $100 million, a sign of his conviction (or ego, depending who you ask).

Notable Moments: Came out of stealth in 2023 with slick renderings of Figure 01, a human-sized biped. Quickly raised $70 million Series A from eager VCs​. Hired dozens of experts from places like Boston Dynamics, Apple, and Google, making rivals nervous about a brain drain​. In 2023, Figure claimed its prototype took first steps faster than any competitor’s timeline​, though the actual performance was still under wraps. By late 2023, leaked news of a potential $1.5 billion raise at $39 billion valuation​ put Figure at the center of both hype and skepticism – was it really worth 15× what Agility Robotics raised, or was the market overeager? In early 2025, Figure’s 2nd-gen robot made headlines for working on a BMW assembly line trial, carefully placing parts in car chassis​. This tangible use-case lent credibility to Figure’s ambitions (and surely helped fundraising conversations).

Reputation: Figure is seen as ambitious and aggressive. Fans laud its speed and Adcock’s willingness to “go big or go home.” Critics point out that until recently Figure had shown more CGI than actual GI (general intelligence). There’s a bit of a rivalry between Figure and more established players – e.g. Apptronik’s CEO implicitly contrasted their decade of experience to Figure’s relative newness​. Figure loves to talk about AI and autonomy, branding itself as solving labor shortages with smart robots. The snarky take? Figure might have “skipped a few steps” in hyping general AI brains before nailing basic locomotion – but with serious money and talent, they are undeniably a frontrunner in the humanoid stakes.

Apptronik

Founded: 2016 in Texas, as a spinout of UT Austin’s Human-Centered Robotics Lab (though its humanoid roots trace back to the DARPA Robotics Challenge in 2013).
Mission: Build practical humanoids that actually do useful work. Apptronik’s mantra is more “robotics first, hype later.” They spent years developing robotic components (arms, leg actuators, exoskeletons) and working with NASA on the Valkyrie humanoid, before unveiling their own full humanoid Apollo in 2023.

Notable Moments: For a long time, Apptronik flew under the radar, content with smaller projects and revenue-generating pilots (they even sold some logistics robot systems and hit break-even on just $28M raised prior to 2025)​. In August 2023, they introduced Apollo to the world – a friendly-looking bipedal robot designed for manufacturing and warehouse tasks. Apollo’s launch was accompanied by news of pilot programs and a goal to be commercially available by end of 2024​. Then came the bombshell in Feb 2025: Apptronik raised $350 million (one of the largest rounds ever in robotics), with investments from Google’s DeepMind (partnering on “embodied AI” for Apollo) and even Mercedes-Benz and ARK Invest joining later, bringing the total Series A to $403M​. CEO Jeff Cardenas has been vocal that Apptronik’s decade of humanoid experience is a key differentiator versus newcomers like Figure, 1X, and Tesla. The company already has agreements with major firms (a deal with Mercedes to explore factory roles, one with GXO Logistics for warehouse trials)​.

Reputation: Apptronik is the quiet workhorse of the group. Robotics insiders respect their technical chops – they know how to build reliable hardware thanks to that Valkyrie lineage and lots of R&D. Apptronik’s approach is incremental: get robots to do something useful in controlled environments now, then scale up. As Cardenas put it, 2025 is about proving utility with early adopters, and full scale comes later​. They aren’t promising a robot in every home tomorrow, which makes them a bit less flashy in headlines, but possibly more sustainable. Their huge funding round in 2025 signaled that even conservative investors trust them to deliver. The snark angle? Some say Apptronik “finally joined the hype club” after raising hundreds of millions – inevitably, expectations will soar. Will their pragmatic engineering culture survive the pressure of big VC money? If they keep executing, Apollo could quietly become the first humanoid that your factory coworker meets on the job.

Agility Robotics

Founded: 2015 in Oregon, as an offshoot of academic research at Oregon State University.
Mission: Create legged robots that work in the real world – specifically bipedal bots for logistics and industrial tasks. Agility’s emphasis has been less on human-like intelligence and more on human-like mobility: their robots navigate human spaces (stairs, narrow halls) with ease.

Notable Moments: Agility made waves back in 2019 by selling a humanoid-ish robot (called Digit) to Ford, which used it in a pilot for last-mile package delivery (Digit could fold up in a self-driving car trunk and then carry a box to your doorstep). Fast-forward to 2022: Agility raised $150 million in one of the first mega-rounds for humanoid robots, bringing in big-name investors and partners (the round included Amazon, indicating a potential use in warehouses). In 2023, Agility unveiled the next-gen Digit with improvements like an LED “face” for better human-robot interaction and gripper hands for handling totes. They also announced the construction of RoboFab, a dedicated factory capable of building over 10,000 Digits per year, aiming to make Digit one of the first mass-produced humanoid robots. By late 2023, Agility had a queue of beta customers in e-commerce and parcel delivery eager to test Digit in sorting centers and warehouses. There were rumors (not confirmed) that Agility was nearing a $300–400M new funding to further scale​, which wouldn’t be surprising given the competitive heat. In 2025, Agility’s focus is delivering their beta units and proving out ROI: can Digit move boxes or unload trucks cheaply and reliably enough to justify its price? They’re not seeking press with sci-fi claims; they’re literally in the backrooms of distribution centers seeing how often their robot drops a package.

Reputation: Among peers, Agility is seen as practical and seasoned. They’ve been at this longer than most (their bipedal research goes back to the Cassie robot, which was effectively Digit’s leggy ancestor). Instead of chasing a humanoid form for its own sake, Agility zeroed in on a specific use-case (moving goods) and optimized Digit for that – notably, Digit doesn’t have an expressive humanoid face or five-fingered hands, because it doesn’t need them for its job. This restraint is admirable, though it means Agility doesn’t always capture the public’s imagination like a shiny new android might. The company tends to let results speak: for instance, demonstrating Digit working an entire shift in a warehouse without a human intervening would be their kind of milestone. Snarky observers might call Digit “the headless delivery drone that walks” – it certainly lacks the humanoid flair of an Atlas or Apollo. But if your online orders get to you faster in the future because a fleet of quiet, bird-legged Digits is hustling in the fulfillment center, Agility Robotics will be having the last laugh (all the way to the bank).

Sanctuary AI

Founded: 2018 in Vancouver, Canada by Geordie Rose (of quantum computing fame) along with colleagues from Kindred (a robotics+AI startup).
Mission: To develop general-purpose humanoid robots with human-like intelligence. If that sounds broad, it is – Sanctuary’s vision is arguably the most grandiose: a robot that can do virtually any work a human can, by deeply integrating cognitive AI with an anthropomorphic body. They’ve even mentioned long-term goals like helping with space exploration.

Notable Moments: Sanctuary operated in semi-stealth for a few years, working on multiple iterations of their robot in the lab. In 2022, they secured a big $58.5M Series A​, and around that time they revealed their 5th-generation humanoid (a system named “Carbon” for the AI control and a physical robot model). Unlike others, Sanctuary initially leveraged teleoperation: human pilots would remotely control the robot to perform tasks, while Sanctuary’s AI software learned from those demonstrations. This approach allowed them to quickly amass “experience” in tasks without the robot failing on its own – essentially, a safety net of human intelligence behind the scenes. By early 2023, the company felt confident enough to attempt a commercial pilot. In January 2023, a Sanctuary humanoid robot was deployed at a Mark’s retail store (owned by Canadian Tire), and over a week it completed 110 different retail tasks (from stocking shelves to cleaning) correctly – all tasks that had only been done in their lab prior. Sanctuary proclaimed this as the first instance of “human-like AI” in a general-purpose robot doing real work​. The feat earned them media attention and even a bit of patriotic pride in Canada’s tech scene. In mid-2023, Sanctuary unveiled Phoenix, their 6th-gen humanoid, which is smaller and sleeker, and said that their AI “Carbon” had achieved a level where it could autonomously handle a broad range of tasks (still often with a safety pilot as backup). They also got a $30M grant from the Canadian government’s Strategic Innovation Fund to continue R&D. By 2025, Sanctuary is moving towards a service model: leasing robots to enterprises with a combination of AI autonomy and remote human operators guaranteeing performance. They’ve effectively acknowledged that pure autonomy is not yet reliable for all tasks, but with clever use of teleoperators (who might oversee multiple robots), they can start generating revenue and gradually decrease the human assist as the AI improves.

Reputation: Sanctuary is viewed with a mix of admiration and skepticism. On one hand, they are among the very few that have demonstrated a humanoid doing lots of different things in a real environment, not just one repetitive task. Their holistic approach – design the “mind” (AI) and body together – is the holy grail of robotics. On the other hand, some peers feel Sanctuary overstates its AI’s abilities; critics suspect much of that retail pilot’s success was due to skilled teleoperators and carefully controlled conditions. The company’s messaging about “human-like intelligence”​ and “AI controlling the robot can work across virtually every industry” comes off as lofty, perhaps excessively so, given no AI today is truly human-equivalent. Sanctuary’s CEO often cites big-picture labor economics (aging populations, labor shortages) as justification for their mission​, which is compelling. The snarky summary: Sanctuary might be “talking to god” a bit – aiming for true AI in a robot – but at least they’re brave (or crazy) enough to try, and they’ve backed it up with one very impressive pilot. If they can gradually reduce the human-in-the-loop and have robots that self-drive through tasks, they could leapfrog the competition. If not, they risk becoming an overpriced remote-controlled robot temp agency. Time (and a lot more training data) will tell.

1X (Halodi Robotics)

Founded: 2014 in Norway (originally as Halodi Robotics, rebranded to 1X Technologies in 2023).
Mission: Create safe, affordable humanoid robots to “augment” the workforce. 1X’s philosophy emphasizes working with humans – their slogan involves humanoids that live and learn among us​. They’re also notable for explicitly focusing on safety (both physical safety and broader AI safety concerns given OpenAI’s involvement).

Notable Moments: For years, Halodi Robotics toiled in relative obscurity, developing a humanoid torso on wheels known as EVE. EVE had dexterous arms and a head, but instead of legs it balanced on a self-balancing wheelbase (imagine a mannequin on a Segway). This simpler design let them test in real environments sooner. They landed a partnership with security firm ADT to pilot EVE robots as nighttime security guards in commercial buildings – patrolling hallways, checking doors – a task that doesn’t necessarily need legs. In 2023, rebranded as 1X, the company caught attention by securing funding from the OpenAI Startup Fund (yes, that OpenAI). This was significant: the leading AI research outfit betting on a humanoid robot venture, suggesting a convergence of AI brains with robot brawn. 1X raised about $23.5M in a round that included OpenAI and Tiger Global in March 2023. They then announced their plan to build a bipedal humanoid named NEO, leveraging their experience from EVE. By mid-2023, 1X reportedly had NEO prototypes in development while also delivering on the first batch of EVE wheeled humanoids for real-world trials. In early 2025, 1X made a strategic move by acquiring Kind Humanoid, a tiny Silicon Valley startup with big ideas (and stylish design) for humanoids. This acquisition gave 1X a foothold in the U.S. talent pool and combined two teams with “aligned visions” for humanoid tech​. 1X stated this union would help it accelerate toward its goal of robotic labor augmentation. They also highlighted the role of generative AI in their approach – hinting that large language models and other AI tech from OpenAI could be integrated into their robots’ control systems​.

Reputation: 1X is often described as a dark horse. They aren’t as flashy as some peers (their robots at first glance are less humanoid without legs, and they spent a long time under the radar), but they have a method to their madness. Their focus on practical deployment (even if with a wheeled variant first) is seen as savvy by industry watchers: get something working and making money, then tackle harder problems like bipedal walking. Also, being based in Norway and having a more European approach to collaboration and safety sets them apart from the Silicon Valley bravado of some competitors. OpenAI’s backing gave them a serious credibility bump – it implies 1X’s team has AI expertise that impressed the likes of Sam Altman. Internally, 1X talks a lot about ethics and ensuring their humanoids are “safe” and earn human trust (perhaps a subtle jibe at the trope of evil Terminator robots). The humorous take: 1X might just be building “mall cops on wheels” today, but don’t underestimate them – once those mall cops get legs and a GPT-4 brain, they could surprise the flashier California robots. In a race crowded with egos, 1X’s cooperative and safety-conscious vibe is almost refreshing – though we’ll see if it holds up as the competitive pressure mounts.

Others to Watch

• Xiaomi / Chinese Entrants: As noted, China is not sitting this out. Xiaomi’s CyberOne was a splashy demo in 2022. More substantially, Fourier Intelligence (China) unveiled the humanoid GR-1 in 2023, claiming it would be mass-produced (100 units) by year’s end and targeting healthcare applications (like rehabilitation support)​. Chinese firms often emphasize cost reduction and scale – if they crack a recipe for a cheap humanoid, it could flood the market (albeit with unknown quality). Keep an eye on Shenzhen’s robotics scene for surprises.
• Boston Dynamics (and Friends): The granddaddy of legged robots is officially not chasing general-purpose humanoids for commerce – BD’s focus under Hyundai has been their quadruped Spot (now deployed with police and utility companies) and their moving-arm robot Stretch for warehouses. However, BD’s Atlas continues to be a research platform pushing the limits of agility. In 2023 Atlas learned to grasp and manipulate objects, not just parkour – a hint that BD could pivot to practical humanoid tasks in the future. Also, rumors persist of Toyota and Honda rekindling humanoid programs (Honda showed prototype legs and arms in recent years, and Toyota has a new humanoid avatar robot). These giants have decades of experience but have been quiet lately; if the startup brigade proves there’s money to be made, the sleeping giants could reawaken.
• Laboratory Projects and Open-Source Efforts: Beyond companies, several well-funded research initiatives are feeding the humanoid boom. The AI Institute for Robotics (a $100M NSF-funded collaboration in the U.S.) is working on advanced robot dexterity and autonomy algorithms that many startups will likely adopt. Open-source robot software like ROS 2 and simulation tools (Gazebo, Isaac Sim) have improved drastically, enabling even small teams to do big things by building on common platforms. In short, the rising tide of robotics R&D lifts all boats (or bots). Today’s startups benefit from standing on the shoulders of these open efforts – something they sometimes forget to acknowledge while chest-thumping on Twitter.

Drama, Feuds, and Faceplants

No cutting-edge tech race is complete without rivalries and juicy stories. The humanoid robot scene, despite a lot of polite academic veneer, has had its share of drama. Here are some notable highlights that SiliconSnark finds too delicious not to mention:

• Musk vs. The World: From the moment Elon Musk announced Tesla Bot with a dancer, the culture clash was set: Silicon Valley bravado meets robotics realism. Musk’s grand claims (millions of bots, everyone having one, solving labor shortages overnight) drew eyerolls from veteran roboticists who’ve eaten humble pie building robots that struggle to open doors. A tongue-in-cheek “feud” played out on social media and podcasts – e.g., Boston Dynamics folks diplomatically saying Tesla’s entrance was good for the field while subtly pointing out that making one-off demo robots is easy, making thousands of reliable robots is hard. Musk later doubled down, stating Optimus would be more significant than Tesla’s car business, sending Tesla fanboys into raptures and putting pressure on actual robot engineers to make good on those words. As of 2025, Tesla has released slick videos of Optimus working in a factory mock-up, but some rivals snark that “it’s amazing what you can do with video editing.” Whether Musk will have the last laugh or eat crow remains to be seen – meanwhile he’s certainly kept competitors on their toes (and perhaps annoyed, since every investor meeting involves answering “How will you compete with Tesla?”).
• Tweetstorms and Subtweets: Humanoid startup CEOs seem to enjoy a bit of online sparring. In mid-2023, Figure’s Brett Adcock tweeted about how certain legacy robot companies were stuck in R&D and too slow to commercialize – a not-so-veiled reference to firms like Agility or Boston Dynamics. This prompted Agility’s CEO to post a humble-brag of Digit working at a customer site, captioned “Real robots, real work 💪.” Another example: when Bloomberg reported the eye-popping Figure valuation attempt, many in the field privately scoffed. Apptronik’s Jeff Cardenas (usually mild-mannered) commented in an interview that “some companies talk about scale, others actually have a production line” – clearly contrasting Apptronik’s factory plan with rivals raising money on promise alone. Even usually neutral observers like IEEE Spectrum editors weighed in, cautioning that the “race” framing can be harmful – cooperation could benefit everyone. But who can resist a bit of competition? It’s clear these companies are watching each other like hawks. Expect more cheeky one-upmanship as they each try to claim “firsts” (first to market, first to raise X amount, first to achieve Y task, etc.). In the end, it might be mostly bluster – the robots themselves don’t care about Twitter.
• Hype vs. Reality (a continuing saga): The entire AI-powered robot race has been a case study in managing expectations. For every genuine breakthrough, there’s been an instance of vaporware or exaggerated demos. We saw Xiaomi trot out a humanoid that could barely wobble – likely years away from utility. We saw some startups showcase CGI concept videos that looked amazing, only to quietly push actual prototypes out by a year or two. There was even a well-publicized moment when one startup’s “leaked demo” turned out to be an animation (cue the collective groan of the community). One particularly cringey episode: a lesser-known firm claimed to have a robot with an “EMOTION engine” that could feel 45 human emotions – which turned out to be marketing fluff about sensor signals, promptly roasted by experts. The hype cycle is a double-edged sword: it attracts talent and capital, but it can also set unrealistic timelines. As a result, the industry has experienced some embarrassing resets – e.g., Google’s Everyday Robots project folded in early 2023 after failing to translate its impressive demos (robot arms cleaning cafeterias) into a viable product, reminding everyone that even tech giants struggle here. The feisty culture of Silicon Valley collided with the methodical ethos of robotics, and sparks flew. The smart players have learned to under-promise and over-deliver (or at least try); the less savvy are still making bold claims that might haunt them. In the words of one roboticist, “We already had an AI winter before – we don’t need a ‘robot winter’ because of overhype.” Consider that a plea as much as a warning.
• Consolidation and Breakups: On the corporate drama front, we’ve seen some relationships bloom and others die. OpenAI’s love affair with 1X was notable – it signaled AI companies and robot companies courting each other (for better or worse). Toyota partnering with Boston Dynamics (via a research project) caused a bit of a stir, as some framed it as “Japan’s revenge” against Tesla’s brashness (though in truth it’s just collaboration on tough problems). And of course, 1X’s acquisition of Kind Humanoid in 2025​ hints that smaller fish may have to join forces to survive the big pond. We’d be remiss not to mention the curious case of Foundation. Here’s a mini-soap opera: Foundation was started by Sankaet Pathak fresh off a fintech failure that left a lot of controversy in its wake​. Many eyebrows were raised that someone could lose tens of millions of customer dollars in a banking fiasco and then almost immediately get millions to do robots. Pathak’s habit of airing grievances about the past on X (Twitter) even as he pitched his grand new venture didn’t inspire confidence in everyone. One venture capitalist anonymously quipped, “I’m all for second chances, but I prefer founders who haven’t misplaced $160 million first.”Ouch. So far Foundation has kept a low profile after the initial splash – it remains to be seen if they can deliver or if investors will become skittish about the founder’s baggage. This just shows the humanoid race isn’t just tech – it’s personalities, egos, and trust at play too.

In summary, the drama in this industry has been both entertaining and instructive. It reminds us that these companies, though working on robots, are run by humans – fallible, competitive, occasionally prone to hubris humans. There’s a famous saying in tech: “Hardware is hard.” The humanoid race proves it, and we might add: “Humility is healthy.” Those who have balanced vision with realism (and maybe a dash of humor about the challenges) seem to fare best. Those who flew too close to the sun might end up like poor Pepper, gathering dust.

The Road Ahead

So, where is this AI-powered robot race headed? Are we on the cusp of a Jetsons-esque future, or a replay of past tech bubbles? SiliconSnark’s take: the truth is somewhere in between, with a side of wit.

First, expect a shakeout in the next couple of years. The sheer number of humanoid projects can’t all survive when the going gets tough (and the VC cash isn’t infinite). We’ll likely see 2–3 leaders pull ahead – those with real deployments, revenue, and sustainable tech – and some others pivot or perish. The good news: unlike the great AI winter of the 1980s, today there is genuine demand driving this field. Labor shortages in certain sectors are very real, and companies are willing to pay if a robot can reliably scrub toilets or load pallets. The bad news: reliability is still the elephant in the room. One robot breaking down is fine; a whole fleet breaking down is a disaster. The pioneers who prove their bots can run 24/7 with minimal babysitting will capture the market (and honestly, none are there yet as of 2025).

Rivalries will evolve: Ironic as it sounds, Tesla might actually become a follower rather than a leader here – they have the resources to catch up, but startups have seized a head start in focus. Don’t be surprised if Tesla acquires one of these humanoid startups by 2026 to bolster Optimus, especially if internal progress lags. Also, watch for cross-pollination: an AI software company might merge with a robotics hardware company to create a true powerhouse. We already saw glimpses with OpenAI+1X. Perhaps a Figure+Sanctuary alliance could happen – pairing one’s hardware focus with the other’s cognitive AI – if they don’t prefer to duke it out independently.

On the technology front, integration of AI advances will be key. The next generation of humanoids will almost certainly have large language models (LLMs) in their brains, so you might actually converse with your robot colleague in plain English and have it somewhat understand. (Whether it responds with delightful wit or becomes a verbose Clippy remains to be seen.) Vision and manipulation will improve as models trained on millions of images help robots identify objects and handle them delicately. We already see Mentee Robotics boasting about NeRF (neural radiance fields) and transformer models in their bot’s control system​– that’s likely the start of a trend where cutting-edge AI research is quickly absorbed into these robots. It’s a virtuous cycle: better AI makes robots more useful, more robots in the world create more data to train better AI.

But let’s not drink all the Kool-Aid just yet. There are some storm clouds to navigate. One is public acceptance. It’s one thing to deploy robots in closed warehouses; it’s another to put humanoids in retail stores or sidewalks where they interact with everyday people. There will be stumbles – literally and figuratively. A few faceplant videos on YouTube, or a robot bumping into a customer, and the memes will flow. Companies will need savvy PR and community engagement to avoid a backlash (the last thing the industry needs is a “robots are taking our jobs!” scare or safety incident). To their credit, many firms emphasize how robots can do jobs people don’t want or fill gaps rather than replace – but skepticism persists.

Another challenge: economics. The current humanoids probably cost low-to-mid six figures each to produce. Even with projections of $100k or $50k unit cost in the future, it’s not pocket change. For widespread adoption, these bots must either get much cheaper or convincingly outperform human labor enough to justify their price. High interest rates and cautious investors mean the era of “grow at all costs” is over; startups will be pressured to show a path to profitability. That could mean focusing on one vertical to dominate (like logistics or eldercare) rather than boiling the ocean of “general purpose” too soon. The ones who find a strong product-market fit will survive the thinning of the herd.

Looking further out, say 5–10 years, humanoids and AI-powered robots will likely be part of everyday life, albeit in niches at first. In warehouses, it wouldn’t be surprising if a quarter of the staff are robots by 2030, working alongside humans. In eldercare, you might see a humanoid helper in some assisted living facilities, gently reminding Grandpa to take his pills or helping Grandma fetch items (this could be hugely impactful given aging societies in Japan, Europe, etc.). In domestic settings, the timeline is fuzzier – doing “everything” a home requires is astoundingly complex (today’s bots still struggle with folding laundry or cooking a full meal autonomously). We at SiliconSnark predict the home humanoid is last to arrive in full form; we’ll get specialized home robots (like vacuuming, lawn mowing, maybe a robo-butler that only operates in your kitchen) before we get a Rosie that does it all.

One thing is sure: the race is marathon, not a sprint. And it’s not strictly winner-take-all – different robots will serve different purposes. A future Tesla Optimus might excel in repetitive factory work with tight integration to Tesla’s manufacturing systems, while a Sanctuary robot might be the preferred choice for dynamic environments where adaptability is key. An Agility Digit might be ubiquitous in every FedEx depot moving boxes, whereas an Apollo could be standard in automotive assembly lines. There may also be surprise players – perhaps Amazon will buy Agility or Apptronik and suddenly corner the warehouse market; perhaps Apple will enter the fray with a beautifully designed home robot that makes toast and plays music (if anyone can make robots “cool” for consumers, it’s Apple – but they’ve shown zero public interest so far).

Finally, a bit of snarky optimism: Past tech hype cycles (VR, autonomous cars, etc.) taught us that breakthroughs take longer than expected, but often we overestimate the short term and underestimate the long term. The humanoid race has had its share of absurdity (we’ll never forget the spandex dancer, thanks Elon), but it has also advanced the state of the art dramatically in just five years. The fact that multiple humanoid robots can now walk untethered, use their hands, see and respond to environments – things that were lab miracles a decade ago – is genuinely impressive. The competition, funding, and yes, even the feuds, have accelerated innovation.

So in the grand tradition of SiliconSnark, we’ll say: Don’t bet against the humanoids – but don’t bet on the hype, bet on the hard work. The robot revolution won’t happen overnight in a viral video; it’ll creep in year by year, task by task. One day you’ll look around and realize your mail was sorted by a robot, or your hotel room was cleaned by one, and it’ll seem normal. And then, perhaps, we’ll reminisce about the early 2020s humanoid race with a chuckle – “remember when they argued on Twitter about who’s robot was less vaporware?” – as we sip tea served by our attentive robot butler.

In the meantime, keep an eye on those funding rounds and pilot results, and enjoy the show. The AI-powered robot race is one heck of a ride, and it’s only just beginning. On your marks, get set, code!