Part IThe Story
The Hard Part Is the Laundry
The pants are black, and the robot is losing.
It is late January 2026, inside a concrete box in San Francisco's Mission District. The only sign of the company from the street is a pi symbol on the door, painted a slightly different color from the rest of it, the kind of marking you would notice only if you already knew to look. There is no reception desk and no lit-up logo. There are long blonde-wood tables, some for lunch and some for work, and the line between the two has worn thin. On the lunch tables, a TechCrunch reporter named Connie Loizos counts Girl Scout cookie boxes, small wire baskets overloaded with condiments, and jars of Vegemite. Her parenthetical (someone here is Australian) does more biographical work than most press releases. On the other tables are monitors, tangles of black cable, spare robot parts, and fully assembled robotic arms, each one partway through some task a child could manage.
One arm is trying to fold the pants and failing. Another is trying to turn a shirt inside out, with what Loizos describes as the kind of determination that suggests it will eventually succeed, though not today. A third is peeling a zucchini quickly and well, then dropping the shavings into a separate container. The zucchini, at least, is going fine.
"Think of it like ChatGPT, but for robots," says Sergey Levine. He is an associate professor at UC Berkeley and one of the company's co-founders, a bespectacled man with the patience of someone who has spent years explaining hard ideas to people who don't get them the first time. Every arm in the room is an experiment. The pants-folder is one researcher's test and the shirt-turner is another's. The zucchini-peeler may be checking whether a model that learned to peel one vegetable can peel an apple or a potato it has never seen. Down the hall sits an expensive espresso machine. It is not there for the staff. Any foamed latte it produces is training data.
Then Lachy Groom walks over. He moves, Loizos writes, like someone with half a dozen things going on at once. He is thirty-one and still looks like one of Silicon Valley's boy wonders, a label he earned early and has never fully lost. He sold his first company at thirteen, nine months after starting it. He was the thirtieth employee at Stripe. He spent five years as a solo venture capitalist and backed Figma, Notion, Ramp, and Lattice early. Now he is a co-founder of Physical Intelligence, which investors valued at $5.6 billion in November 2025. The company does not sell a robot. It has no disclosed revenue. Its most visible output, on the day Loizos visits, is a machine that cannot yet fold a pair of pants.
The obvious question about Groom is how he got here. The better one is why someone who spent his entire working life inside software, someone who as a teenager said he liked the internet because it let him hide, would stake the second half of his career on the stubborn, physical, unhideable world. Laundry. Cardboard. Vegetables. Things you cannot do from a laptop in Perth while pretending to be older than you are.
By the Numbers
From Perth to π
13Groom's age when he founded PSDtoWP, sold nine months later to an American buyer
#30His employee number at Stripe, where he stayed roughly six years and ran Issuing
$48M → $250MGrowth from LGF's first fund to its third, raised in 2021
119Portfolio companies tracked by Dealroom, including 24 unicorns
$5.6BPhysical Intelligence's valuation after a $600M round led by CapitalG, November 2025
$3,500Retail price of the company's robot arms, 'with an enormous markup'
Nobody on the Internet Knows You're Thirteen
The stories begin with a grandfather. Depending on which local paper you trust, Groom was ten or eleven when his grandfather taught him HTML. The West Australian says ten. The Sydney Morning Herald says that by eleven he was working in HTML and CSS. One translated tech blog credits his father, and that seems to be an error. Groom himself has told it the same way for more than a decade, from a 2012 newspaper interview to an e-commerce podcast in 2018: "My granddad actually taught me HTML." He picked up CSS on his own, and soon noticed something most people learning a skill don't notice until much later. Both languages were in demand, and people would pay for them.
Before that there was the usual entrepreneurial childhood. His father, Geoff, a nuclear physician working in radiology, summed it up for a reporter: "He took dogs for walks for money. He set up a lemonade stand with a mate. He was always looking for an opportunity to make pocket money." The lemonade stand is a standard opening in founder biographies. What came next is less standard.
In 2008, at thirteen, Groom started PSDtoWP. Designers would mock up a website in Photoshop, and he would turn the layered file into a working WordPress site. The clients were mostly in the United States and Europe. He handled them from South Perth, which meant getting up before dawn to answer Americans and working evenings to reach Europeans, with school in between at Wesley College, a private boys' school on the south side of the Swan River. "That grew very quickly," he said later. "Nine months later when an American company bought it, I had hundreds of clients all over the world." One investor database names the buyer as PSD2HTML.com. The terms were never disclosed.
— Geoff Groom, Lachy's father, to The West Australian, 2012He was a 14-year-old dealing with people 30 years older than him. He was organising projects and getting programmers in Czechoslovakia and Thailand to do work for him.
Czechoslovakia had stopped existing about a year and a half before Lachy was born. The detail is charming because it is wrong. A nuclear physician in Western Australia is describing, in slightly outdated geography, his son's global outsourcing operation run from a bedroom. The father's map was out of date. The son's map was the internet, which had no borders worth mentioning, and on it a fourteen-year-old could hire contractors in Central Europe and Southeast Asia to build sites for American designers who did not know they were working for a child.
That last point mattered most, and Groom understood it clearly. In June 2012, at seventeen, he told StartupSmart the most revealing thing he has said on the record:
"The great thing about the internet is it's not face to face. People don't know my age unless they do a little bit of research, so I can hide behind the internet in a lot of ways."
Keep that sentence in mind. It is the key to much of what follows.
The businesses kept coming. On holiday in Bali he read Tim Ferriss's The 4-Hour Body, a bestseller that recommended a four-supplement regimen it called the PAGG stack (policosanol, alpha-lipoic acid, green tea extract, garlic). According to one account, he registered PAGGStack.com within hours and called suppliers. Within a month of getting home he was packaging and shipping the kits worldwide. That business also sold.
Then there was the iPad. "I'm a huge Apple fan, everything tech I own is Apple," he told the Herald. "I had a US friend ship in the latest iPad, took it to school, and pretty quickly a friend dropped it on concrete. Screen cracked. iPad destroyed." He had no project going at the time, and his father suggested he build something for passive income. So in May 2010 he launched iPadCaseFinder.com, a table of protective cases from various vendors in which every outbound link was an Amazon affiliate link. "Amazon pays generously," he said. The site drew 400,000 unique visitors in its first few weeks and was sold within months.
The pattern is clear enough. A teenager spotted a gap, filled it with software, captured the traffic, and sold to a bigger company before the gap closed. None of it required a body. All of it required a sense of timing, which he later stated as a rule to Anthill magazine: "Start a business in reaction to a trend and you'll ride the tail end of the wave. Anticipate the wave and you'll ride it to the end."
The iPad site, to be fair, was a reaction. It began with a cracked screen on a school courtyard. The maxim was something he was still growing into.
Value Left on the Table
Cardnap was the business he kept the longest, and the one that taught him the most, mainly by being painful.
The premise came from a statistic. The Australian Consumers' Association, now called Choice, had found that one in three Australian gift vouchers expires before anyone uses it. Groom saw that unused value as a market. Cardnap, launched in March of the year he turned fifteen and co-founded with a partner named Josh Davies, let Australians enter their unwanted vouchers and get an instant cash quote, paid by bank transfer or PayPal. They could also swap a card for one from another store, posted free anywhere in Australia. The site covered more than forty-five retailers: Westfield, David Jones, Myer, Woolworths. Users could donate part or all of a card's value to Kids Under Cover, a youth homelessness charity. Cardnap bought cards for up to 90 percent of face value and resold them at discounts of up to 30 percent. The founders had plans for the United Kingdom, New Zealand, Singapore, Hong Kong, and South Korea.
Look at that list of countries. It comes back later.
Running a two-sided marketplace for gift cards meant dealing with money: taking it in, sending it out, holding inventory, checking that a card had the balance its owner said it did. Six years later, as head of a product at Stripe, Groom described that period to the hosts of an e-commerce podcast with feeling: "It was just a nightmare to set up." He had to accept payments, pay people out, manage inventory, and handle verification. Eventually he pivoted Cardnap into a simpler product, an aggregator that searched for discounted gift cards across sellers. Anthill reported that by his own estimate it was the largest of its kind. That version avoided the hard parts. It could run as software alone.
By mid-2012 the Herald credited the teenager with helping create fifteen jobs in the United States at an average salary of $79,000. Cardnap was by then available only to American users, and Groom was already talking about selling it. "I still own this site," he said, "but may be looking to sell it soon so I can concentrate on getting to the US." In another interview he said he would probably move on to "something in education or something else."
The "something else" turned out to be the plumbing he had hated building. Somewhere between Perth and San Francisco, the boy who found payments a nightmare decided to work at the company built to make that nightmare go away.
Aliens of Extraordinary Ability
He finished Year 12 in December 2011, though only just, and only by arrangement. David Gee, the principal at Wesley College, had worked out a deal so Groom could keep his businesses running during school hours: one free period a day for work, and his mobile phone three times a day. "Right from the beginning of Year 11 he had some clear pictures that school was something he had to do," the headmaster told the Herald, describing a negotiation in which his parents wanted him focused on his final two years and he wanted to focus on business. Gee meant it as praise when he said: "How many 15 and 16-year-olds have sold businesses to overseas competitors? I've not seen too many like him."
Soon after graduating, Groom flew to San Francisco on a tourist visa. He had found a San Francisco startup running a pilot program to train people in Ruby on Rails. This was Dev Bootcamp, a nine-week apprenticeship. "I seized the opportunity immediately after graduating and moved out of home to San Francisco on a whim," he told Anthill. He finished the program with several job offers, and the bootcamp itself hired him. The West Australian, which sent a reporter to see him, said he had been offered the head of marketing role. By December, SmartCompany reported that he was handling Dev Bootcamp's PR, marketing, and rebranding.
— Lachy Groom, to StartupSmart, June 2012The only reason to stay in Australia – other than it's a great place to live – is if you're targeting the Australian market. The reason why Australia doesn't work is because the valuations are so much higher in the US.
There was a catch, and it was physical. The internet had let him hide his age, but U.S. immigration law looks at the person, not the website. Groom had interviewed with Twitter and Zynga, and each time, StartupSmart reported, "the interview process always comes to a grinding halt because Groom does not have a work visa." The main visa for skilled workers required a degree. He did not have one, because he had been busy selling companies when other people his age were applying to university. "Ironically," The West Australian wrote, "Mr Groom's youthful achievements are holding him back from his dream to work in Silicon Valley."
He applied for the one category that might fit, the visa for people of "extraordinary ability," which Anthill described as reserved for artists, scientists, and businesspeople. American law used the old term "aliens," and the press used it too. "I have some amazing, successful CEOs writing me letters of recommendation," he said. "I would love to work in the USA."
He was seventeen, and he could neither drink nor vote in the country where he was trying to build a career. He was also self-aware about it: "It's something I'm very aware of and I'm very aware it's held me back in a lot of ways. I feel I need to validate myself to people." He noted, with some irony, that if he had been forty-eight, the Sydney Morning Herald would not have written about him.
The Anthill profile from that year, written in the punchy style of Australian business magazines, opens with a question it asks all its 30under30 applicants: "Have you ever been lucky, punk?" That's Dirty Harry, a San Francisco cop. Groom's favorite quote, printed at the bottom of the piece and attributed to "Unknown," was: "Entrepreneurship is living a few years of your life like most people won't, so that you can spend the rest of your life like most people can't."
What the paperwork asked him to prove, in other words, was that he was extraordinary. Remember that phrase too. The company he would later co-found is betting on the opposite claim, that the ordinary is the hardest thing there is.
The Thirtieth Employee Learns the Plumbing
By December 2012, SmartCompany noted almost in passing that Dev Bootcamp was not his only American job: "He also works at Stripe, an online payments processor."
He was the thirtieth employee. Later profiles disagree on the date. One investor's blog says 2014, a crypto-news write-up repeats that, and both are wrong. Groom's own records and the 2012 coverage put his start in 2012, so he joined Stripe at eighteen.
He stayed about six years, which is a long time in Silicon Valley and an eternity for someone who had sold three companies before finishing school. His work covered product, growth, and operations. He led Stripe's expansion into Singapore, Hong Kong, and New Zealand. That is three of the five markets the fifteen-year-old had listed for Cardnap, now reached with a payments company behind him instead of a teenage gift-card site. The boy who had imagined going international ended up doing it for someone else.
Eventually he ran Stripe Issuing. On the podcast he described it to retail analysts Jason Goldberg and Scot Wingo as "an end-to-end platform for quickly creating, distributing, and managing physical and virtual cards." The job meant managing payment operations and relationships with Visa, Mastercard, and American Express.
This is the most neatly rhyming fact in his biography, and it should be stated plainly. At fifteen, Groom built a business trading other people's plastic cards. At twenty-four, he ran the product that let other businesses create their own. He went from the secondary market for cards to the primary infrastructure behind them. The marketplace that had been "a nightmare to set up" became a problem he solved at scale for thousands of companies. Stripe's customers at the time, his hosts noted, included Warby Parker, Wish, and Target.
The podcast was recorded on Friday, August 3, 2018. Before talking about payments, the hosts asked him to recount his career. He started with his grandfather and HTML and moved on to Cardnap. The transcript is rough. It renders his name as "laki groom" and the business as "card nap." But the order of his story is revealing. Asked how he got to Stripe, he did not begin with Stripe. He began with the bad experience that made Stripe make sense to him.
By the end of that year he had left.
A Venture Firm of One
Venture capital has a conventional structure: a partnership, several general partners, a brand that outlasts any one person, investment committees, associates, Monday meetings. Groom built none of that. LGF (Lachy Groom Fund, as some databases call it) had one general partner. According to securities filings, it was initially based in West Hollywood, which is an odd choice for a San Francisco person and suggests that, at first, the firm was simply wherever he was.
An SEC filing in February 2020 showed the first fund at roughly $48.8 million. On May 19, 2020, The Wall Street Journal reported he was raising about $100 million for a second. In 2021, The Information reported a third fund of $250 million. That fund's first investment was Blues Wireless, a Boston company connecting cellular devices to the cloud, in a $22 million round announced in March 2022 that he co-led with Sequoia Capital. A lone investor sharing the lead with one of the industry's most established firms said something about where solo investing had gotten to. That February, Mario Gabriele published "The Future of Solo Capitalists" in The Generalist, observing that solo general partners now managed more money than many multi-partner funds. The Information ran its own version: "Solo Venture Capitalists Go Big."
The portfolio reads like a list of the decade's breakout software companies. Figma. Notion. Ramp. Lattice. Deel. WorkOS, the enterprise identity company, and Daily, the video and audio developer platform, both led from Fund II. Anduril, the defense company. Zepto, the Indian quick-commerce company, where LGF joined a $200 million Series D in May 2022 at a $900 million valuation. In August 2023 he led a $16 million round for Speak, the AI language-learning startup. Dealroom now counts 119 portfolio companies, including twenty-four unicorns and six decacorns. Using disclosed ownership and latest valuations, it estimates his named stakes at about $1 billion, the largest being roughly 6.5 percent of Zepto. NFX's Signal lists his check range as $100,000 to $10 million with a typical check around $1 million.
The approach was coherent, though he rarely laid it out. Brian Nichols of Hustle Fund summarized it as backing "tools that developers and operators love first, then monetize later." Figma spread among designers before companies bought it. Notion spread among individuals before teams adopted it. Groom had seen this pattern up close at Stripe, which grew through developers who integrated it on their own initiative well before any procurement department approved it. When he explained his 2025 Series A investment in the sales-software company Nooks, he used the same vocabulary as the thesis: "They have this unique combination of insane customer love & engagement, a world-class team shipping product at breakneck pace, and market timing."
Market timing again. The wave, anticipated.
Speed was the main advantage of having no partners. In spring 2026, TechCrunch ran a headline that summed up the model: "A 20-minute pitch wins Indian startup Pronto backing from Lachy Groom." The deal was a $20 million Series B extension that roughly doubled the company's valuation to about $200 million. A twenty-minute meeting can turn into a term sheet when only one person has to agree. In September 2026 he led a $2.5 million pre-seed for Alteon, an Indian aerospace startup trying to keep an aircraft aloft for a year. In July 2026, Axios reported he had raised two new funds.
What he gave away was as telling as what he bought. On May 1, 2019, shortly after leaving Stripe, he posted on X: "Since leaving Stripe I've been spending a lot of time helping friends, and friends of friends think through career decisions (and very often comp negotiations). It's something I really enjoy doing—if you know anyone who needs help, DMs" are open. His profile states that his name is pronounced "locky" and lists his personal Gmail address. In September 2025 he posted one line that drew about 380,000 views: "the opportunity cost of a bad job has never been higher." Coming from someone who had spent five years as an investor while, by his own later account, searching for something to build, it reads partly as advice to others and partly as a note to himself.
His investing did not stay narrow. Starting with software and fintech, it moved into the physical world: Anduril, Hadrian (the factory-automation company), Pacific Fusion, Peak Energy, Standard Bots, Mytra. Dealroom tags six of his portfolio companies as robotics. Over time his investments were drifting toward atoms.
Five Years of Waiting for the Right Room
"I was looking for five years for the company to go start post-Stripe," Groom told Loizos in the Mission District lab. "Good ideas at a good time with a good team — that's extremely rare."
That may seem surprising coming from someone whose public story is a string of businesses started almost on a whim: a domain registered within hours in Bali, a site built in an afternoon after a friend broke an iPad. The adult Groom was much more patient than the teenager. He spent five years and three funds watching other founders build, and he did not start a company of his own.
What changed was some academic papers. Groom described it for an Australian YouTube documentary made by Sachin Shah and Adam, two creators who flew to San Francisco to ask how someone from Perth ended up co-founding one of the most hyped companies in the world:
"I've always been somewhat obsessed with robotics and frankly somewhat disappointed with how little progress we've made in robotics. It feels like we've made so much progress on the Internet on software, and so little progress on technology in the physical world."
Around 2021 and 2022, he began reading research from two labs, Sergey Levine's at Berkeley and Chelsea Finn's at Stanford, and followed everything they published. "I mentioned to a handful of friends, if you ever hear of them thinking of starting something, let me know."
It's an unusual move for a venture capitalist. The standard approach is to back a founder once they've decided to start something. Groom set up an early-warning system for when certain researchers might decide to leave academia, so he could be in the room as a co-founder, not just a funder. In 2023, a friend passed along a rumor that they might be thinking about it.
The people involved fit the pattern. Each had spent years in one environment waiting for a hard problem to become solvable:
Sergey Levine, associate professor at Berkeley, helped pioneer deep reinforcement learning for robots. His work starts from the premise that machines should learn movement through trial and error rather than being programmed. He appeared on Lex Fridman's podcast to talk about robotics and machine learning years before the field got the attention money brings.
Chelsea Finn, a Stanford professor working on robotic learning, ran the other lab Groom had been watching.
Karol Hausman had been a staff research scientist at Google Brain and led robotic manipulation at DeepMind, the team responsible for teaching machines to use their hands, the hardest part of the problem. Groom was introduced to him first. "We got to know each other quite well."
Brian Ichter, with a Stanford PhD in aeronautics and astronautics, had been a research scientist at Google Brain and DeepMind working on getting mobile robots through complex tasks. Quan Vuong came from the same world.
Adnan Esmail was the hardware person: formerly SVP of engineering at Anduril and before that at Tesla. Anduril was also in Groom's portfolio, so the two had connected through capital before they connected as co-founders.
As Groom tells it, the DeepMind group had been ready for some time. "They finally saw the light at the end of the tunnel, like this is maybe finally a solvable problem," he said, "and we decided to join forces." The company launched in March 2024 with a $70 million seed round. Sequoia, which had co-led Blues Wireless with him two years earlier, backed it through Alfred Lin.
Accounts of his title vary. A professional-network summary calls him CEO. Robotics trade publications list Hausman as chief executive. What is clear is that he is not just the company's business manager. His name appears among the authors of π0: A Vision-Language-Action Flow Model for General Robot Control, the October 2024 paper introducing the company's first generalist model. It appears again on π0.5: A Vision-Language-Action Model with Open-World Generalization, presented at CoRL 2025, and on Hi Robot: Open-Ended Instruction Following with Hierarchical Vision-Language-Action Models at ICML 2025. The kid who learned HTML from his grandfather, skipped university, and was nearly kept out of the country for lacking a degree is now a co-author of machine learning papers alongside Berkeley and Stanford professors.
Cheap Arms, Expensive Minds
The bet is easy to state and hard to execute. Physical Intelligence does not make robots. It makes the models that control them: foundation models that combine camera input, natural-language instructions, and motor control into one system meant to work across many robot bodies. Its own description of the goal is a model that will control any robot to do any task. In industry terms it is a supplier to the hardware companies, which makes those companies its competitors and its potential customers at once.
That is why the hardware in the lab is so plain. The arms sell for about $3,500, and Levine told Loizos that price includes "an enormous markup" from the vendor. Built in-house, the materials would cost under $1,000. A few years earlier, he said, a roboticist would have been surprised these arms could do anything useful. Loizos put the logic simply: good intelligence compensates for bad hardware.
— Sergey Levine, co-founder, Physical Intelligence, to TechCrunch, January 2026Think of it like ChatGPT, but for robots.
The problem has a name, Moravec's paradox, from the roboticist Hans Moravec's observation in the late 1980s: it is fairly easy to get a computer to perform like an adult on intelligence tests or at checkers, and very hard to give it the perception and movement of a toddler. Machines beat grandmasters at chess decades ago. They still cannot reliably fold a shirt. The skills humans think of as impressive turn out to be the easy ones to automate. The skills humans don't think about at all, like how hard to grip a paper cup compared with a lead weight, rest on millions of years of evolution that nobody has managed to write down as code.
Groom's whole career had been built on the first kind of skill, the kind you can perform remotely and in text, where nobody knows you are fourteen. Physical Intelligence is a bet on the second kind.
The results have come in stages. π0 was published on October 31, 2024, and its weights and code were later released publicly. Five days later, on November 4, the company announced $400 million in new funding from Jeff Bezos, OpenAI, Thrive Capital, Lux Capital, and Bond Capital, at a valuation of about $2 billion (some accounts say $2.4 billion). In April 2025, Groom and Hausman appeared on TBPN to discuss π0.5. Their argument was that simulation works for locomotion, but manipulation requires real data from real environments: actual kitchens, actual laundry, actual homes. The company's data loop is the same one Loizos saw. Data is collected at robot stations in the lab and in warehouses and homes, used to train new models, and those models come back to the stations for evaluation.
In November 2025 the company published π*0.6 with an announcement titled, in effect, "RL is back." It was trained with a method mixing human demonstrations, corrections from teleoperators, and robots practicing on their own. The same month brought a $600 million round led by Alphabet's CapitalG, with Lux, Thrive, Bezos, Index, Sequoia, and T. Rowe Price participating, at $5.6 billion. Total disclosed funding reached about $1.1 billion. In December the company reported top-tier results on a set of household tasks it called the "Robot Olympics." Early 2026 brought a memory architecture that let policies track tasks lasting about fifteen minutes, and in April came π0.7, which the company described as able to combine known skills on hardware it had not been trained on. Bloomberg reported talks to raise about $1 billion at more than $11 billion, with Founders Fund and Lightspeed in the discussions and Nvidia mentioned among investors. Humanoids Daily noted carefully that it found no primary confirmation that the round had closed.
A Korean trade publication wrote in March 2026 that the company had no commercial products and no revenue. That was true in the narrow sense. The company had placed robots in short-term rentals to fold laundry, and in the back of a San Francisco chocolate shop. It was running trials, not selling anything.
Groom told the Australian documentarians that useful robots might be three years away, or might never arrive. He presented both possibilities at once, and the documentary's own summary repeated the hedge. For someone at the center of one of the most heavily funded bets in robotics, it is an honest thing to say.
The Physical World Comes to the Door
Groom's house is on Dorland Street, in the same neighborhood as his company, a few blocks from where Mark Zuckerberg once owned a $31 million property. According to property records cited by the New York Post, Groom bought it in 2021 for $1.8 million from Sam Altman's brother. Press coverage now values it at $4.4 million.
On a Saturday evening in November 2025, a man came to the door holding a white box. He said he had a delivery from a UPS affiliate for "Joshua." Joshua, a tech investor who lived in the house and whom later coverage described as Groom's roommate and business associate, opened the door. The man asked him to sign and asked for a pen. He followed Joshua inside. Witnesses heard a loud bang.
Inside, according to police sources cited by the San Francisco Chronicle, the intruder pulled a gun, bound Joshua with duct tape, and held him for about ninety minutes. He beat him. He held up a phone on speaker while voices with foreign accents read out personal information. He poured liquid over him. Then he emptied Joshua's cryptocurrency wallets, roughly $11 million in Bitcoin and Ethereum, and took his phone and laptop. Police arrived at 6:45 p.m. and found the victim with minor bruises. Investigators said they believed the robbery was a targeted operation by an organized crime group. As of the reports, the suspect had not been caught.
The coverage moved fast and mostly went for the gossip. Several outlets led with Groom's past relationship with Altman, which the Post attributed to people familiar with Altman's history. They pointed to a photo the two had posted on Facebook in 2014 and a Getty image of them together at the Allen & Company conference in Sun Valley, Idaho, in 2018. Groom and Altman had also invested together. A 2024 Forbes headline described them as "doubling down" on the same internet startup. Some crypto-news sites misreported the basics: one said Groom lost the money himself, another said Physical Intelligence had raised $17 billion at a $56 billion valuation. As far as the record shows, Groom said nothing publicly. The work went on.
The episode has a simple shape. A fortune existed only as entries on a blockchain, accessible from a phone, as close to pure abstraction as money gets. It was taken by a man with a cardboard box, a roll of tape, and a gun. The digital world has always depended on the physical one, and that evening the physical one showed up at the door. The internet lets you hide your age, your location, and the size of your accounts. It does not hide your address.
What the Robot Makes
At Physical Intelligence, the black pants will probably get folded eventually. The shirt will probably get turned inside out. Loizos judged the arm determined enough. Not that day, though.
Groom has spent much of his life getting to places before he was supposed to be there: in business at thirteen, in America at seventeen, at Stripe at eighteen, investing at twenty-four, co-authoring research papers at thirty without a degree. Each time, the barrier in front of him was something abstract that could be argued with, like a credential, a visa category, an age limit, or a market gap. Each time he got around it with talent, letters of recommendation, and the fact that on the internet nobody had to see him.
Laundry does not respond to letters of recommendation. A sheet of cardboard does not care who is extraordinary. The problem Groom has finally chosen is one where the work cannot be hidden behind a screen. The only thing that counts is whether, at the end, the thing on the table has been done.
In the back room of Dandelion Chocolate, a San Francisco chocolate maker, the company has installed one of its systems. According to Bloomberg, the point is to take over repetitive, low-value work so the staff can concentrate on making chocolate. The arm cost about as much as a used car. No one is watching it. It picks up a flat piece of cardboard, the kind that will hold a gift, and folds it into a box.
Part IIThe Playbook
The principles below come from the record in Part I: the bedroom businesses in South Perth, the visa problem, six years inside Stripe, a one-person venture firm, and a robotics company founded after five years of deliberate waiting. They are not motivational slogans. Most are about timing, location, and the uncomfortable question of which problems actually deserve your time. Several contradict each other. Groom's career has involved holding opposing instincts, impulse and patience, software and atoms, and choosing between them based on the situation.
The principles
- 1.Let the work speak before the résumé can.
- 2.Anticipate the wave; don't chase its tail.
- 3.Turn your worst operational pain into your next employer.
- 4.Move to where the valuations live.
- 5.Join early, then stay long enough to own something hard.
- 6.Back love before revenue.
- 7.Make speed your structure.
- 8.Give advice away in public.
- 9.Wait for the rare conjunction.
- 10.Build an early-warning system for talent.
- 11.Buy intelligence, not hardware.
- 12.Measure progress in the mundane.
π
The Arc in Dates
Each stage started before the previous one was finished.
2008
At 13, founds PSDtoWP; sells it nine months later to an American buyer.
2010
Launches iPadCaseFinder in May after a friend breaks his iPad; 400,000 unique visitors in the first weeks.
2011
Finishes Year 12 at Wesley College in December, under a one-free-period-a-day arrangement.
2012
Moves to San Francisco on a tourist visa; Dev Bootcamp; joins Stripe as employee #30.
2018
Heads Stripe Issuing; leaves at year's end to invest.
2020
LGF I filed at ~$48.8M; WSJ reports ~$100M raise for LGF II.
2021
Raises $250M LGF III; buys the Dorland Street house.
2024
Co-founds Physical Intelligence; π0 released October 31; $400M round in November.
2025
$600M round led by CapitalG at $5.6B; π*0.6 and "RL is back."
2026
π0.7 in April; two new LGF funds reported in July.
Principle 1
Let the work speak before the résumé can
At fourteen Groom was managing contractors in Thailand and Central Europe for clients in America, none of whom knew his age. He knew this was an advantage: "People don't know my age unless they do a little bit of research, so I can hide behind the internet in a lot of ways." The usual advice to young operators is to build credentials first. Groom did the reverse. He shipped work through a channel where credentials weren't visible and let the results stand for him.
The limits showed up quickly. Immigration officers, unlike clients, look at the person. When the U.S. visa system asked for proof, his early output became the proof. Letters from "amazing, successful CEOs" supported an extraordinary-ability petition that a degree could not have provided. The anonymous work produced the reputation, and the reputation then got him through the gates that anonymity couldn't.
Tactic: Find a channel where the buyer sees only the output, not the person, and build a record there before applying anywhere that checks credentials.
Principle 2
Anticipate the wave; don't chase its tail
"Start a business in reaction to a trend and you'll ride the tail end of the wave. Anticipate the wave and you'll ride it to the end." He said this at eighteen, and his own history doesn't fully support it. iPadCaseFinder was a reaction to a cracked screen. PAGGStack was a reaction to a bestseller. Both worked and both were sold quickly, which is how you exit when you've caught the tail of a wave.
The maxim applies more clearly to his later career. He started following Levine's and Finn's work in 2021 and 2022, before the robotics foundation-model boom, and was already positioned when the researchers decided to leave. He backed Figma when designers were quietly switching tools, before the switch was obvious. The skill isn't predicting the future. It's noticing which part of the present is still small enough to get into.
Tactic: Separate your opportunities into reactions and anticipations; sell the reactions quickly and hold the anticipations patiently.
Principle 3
Turn your worst operational pain into your next employer
Cardnap taught Groom payments by making them miserable: taking money in, paying it out, managing inventory, verifying balances. "It was just a nightmare to set up," he said in 2018, as head of a Stripe product. The connection is direct. The teenager who traded gift cards ended up running the platform that lets businesses issue cards.
Most people leave a painful problem behind. Groom went to work for the company built to fix it. As a career strategy this is underrated. Your own scars give you an understanding of the customer that no interview can test for, and they make the infrastructure you'd be building feel real instead of theoretical.
Tactic: List the three most painful operational problems from your last venture and find out which companies are building the fix; one of them should be your next job.
Principle 4
Move to where the valuations live
At seventeen Groom was blunt about Australia: "The only reason to stay in Australia – other than it's a great place to live – is if you're targeting the Australian market. The reason why Australia doesn't work is because the valuations are so much higher in the US." He added: "In San Francisco, there's a whole community – every company is a start-up."
The point isn't patriotic or anti-patriotic. It's that location is a strategic decision, not a background condition. Groom treated where he lived as a bet and paid for it: a tourist visa, stalled interviews at Twitter and Zynga, an uncertain extraordinary-ability petition. He took on the legal risk because staying where he was felt riskier.
◐
Conventional Wisdom vs. the Groom Pattern
Where the record departs from standard career advice.
| Conventional wisdom | What Groom did | Phase |
|---|---|---|
| Get a degree, then a visa, then a job. | Got the job first; argued for the visa on demonstrated ability. | Builder |
| Leave a painful domain behind. | Joined the company fixing it; ran card issuing after a gift-card marketplace. | Operator |
| Build a partnership to scale a fund. | Stayed a fund of one; co-led with Sequoia anyway. | Investor |
| Fund founders after they decide to start. | Asked friends to tip him off before researchers decided; joined as co-founder. | Founder |
| Hardware is the moat in robotics. | Use $3,500 off-the-shelf arms; put the value in the model. | Founder |
Tactic: Work out the real valuation and talent gap between where you are and the leading hub in your field, and if it's large, treat moving as the first item on your roadmap.
Principle 5
Join early, then stay long enough to own something hard
Joining as employee #30 is the obvious part. Staying six years is the less obvious one. Groom, who had sold three companies before eighteen, could have left Stripe at any point to start a fourth. He stayed long enough to lead expansion into Singapore, Hong Kong, and New Zealand and then to run Issuing, a product requiring relationships with Visa, Mastercard, and American Express. Those relationships are measured in years, not funding rounds.
Staying long turned his early access into real expertise. Being at a fast-growing company early teaches you to move fast. Staying teaches you how things work when the company gets big: compliance, partnerships, systems that can't fail. Those lessons later underpinned his judgment as an investor in fintech companies like Ramp and Deel.
Tactic: If you join a breakout company early, stay until you own a product line with external partners; that's when the learning starts paying off.
Principle 6
Back love before revenue
Groom's investing thesis, as observers have pieced it together, prefers products individuals choose before their companies pay: Figma among designers, Notion among individuals, Stripe among developers. His own description of Nooks puts "insane customer love & engagement" ahead of team and timing.
Bottom-up adoption is the most honest early signal, because nobody was told to use the product. That's also why it's hard to read from standard metrics. Revenue arrives late, and the early evidence is behavioral: who switched, who keeps using it, who brings colleagues. Groom learned to see this inside a company that grew because developers kept choosing it.
Tactic: In diligence, find the users nobody required to adopt the product and ask what they gave up to switch; how intense their answers are matters more than any revenue figure.
Principle 7
Make speed your structure
"A 20-minute pitch wins Indian startup Pronto backing from Lachy Groom." That headline from spring 2026 describes the main advantage of a solo general partner. No investment committee, no Monday meeting, no colleague who wants another reference call. LGF grew from about $48 million to $250 million in three funds without adding partners, and co-led a round with Sequoia, the clearest sign that a one-person firm could sit at the same table as the institutions.
The tradeoff is real. A solo general partner has nobody to check their judgment, and many of the firm's processes exist precisely to catch the mistakes of overconfident people. Groom's answer seems to be a disciplined focus on bottom-up adoption plus a large network of co-investors who serve as an informal committee.
Tactic: If you act alone, make speed of decision the core of what you offer, and compensate for having no committee by sticking to a narrow, explicit thesis.
Principle 8
Give advice away in public
In May 2019, newly independent, Groom announced on X that he was helping friends and friends of friends with career decisions and "very often comp negotiations," and invited more requests by DM. His profile lists a personal Gmail address. Six years later, his line "the opportunity cost of a bad job has never been higher" reached hundreds of thousands of readers.
This isn't charity, and it isn't purely strategy either. It's how a solo investor builds deal flow: the people you help with offer letters this year become founders, executives, and sources of referrals later. It also keeps you close to how people actually decide where to work, which matters a great deal when you later need to recruit researchers away from DeepMind.
Tactic: Pick one narrow, valuable kind of help, like compensation negotiation, and offer it publicly and for free; treat the inbound requests as long-term deal flow.
Principle 9
Wait for the rare conjunction
"I was looking for five years for the company to go start post-Stripe. Good ideas at a good time with a good team — that's extremely rare." The teenager registered domains within hours. The adult held off for half a decade. This is the clearest change in his story: he learned when impulse works and when it doesn't.
Small bets that can be sold quickly reward impulse. A company meant to take ten years, raise a billion dollars, and attempt something a whole field had failed at rewards waiting. Groom's three conditions, idea, timing, and team, are easy to list and very hard to find together. Most founders start with one or two and hope for the rest. He waited until all three were present.
Tactic: Before committing years to a venture, write down separately whether the idea, the timing, and the team each meet your bar; if any one doesn't, keep investing and keep looking.
Principle 10
Build an early-warning system for talent
Groom's account of founding Physical Intelligence turns on one instruction to friends: "If you ever hear of them thinking of starting something, let me know." He had read the Levine and Finn papers. He knew which researchers mattered. He arranged to hear early, and when the rumor came in 2023 he was introduced to Hausman before anyone else had a chance.
Then he stayed involved in the technical work. His name is on the π0, π0.5, and Hi Robot papers. Being an investor got him access to the researchers. Being a co-author earned him standing with them.
⚙
The Founding Team
Each brought a different piece of the problem.
| Co-founder | Prior role | Contribution |
|---|---|---|
| Sergey Levine | Associate professor, UC Berkeley | Research Deep RL for robotics |
| Chelsea Finn | Professor, Stanford | Research Robotic learning |
| Karol Hausman | Led robotic manipulation, DeepMind | Research Manipulation at scale |
| Brian Ichter | Research scientist, Google Brain / DeepMind | Research Mobile robots, complex tasks |
| Quan Vuong | Google DeepMind | Research Robot learning |
| Adnan Esmail | SVP Engineering, Anduril; Tesla | Hardware Systems engineering |
| Lachy Groom | Head of Issuing, Stripe; sole GP, LGF | Company Capital, operations, co-author |
Tactic: Identify the five researchers or operators whose work you'd quit your job to build with, and ask three well-connected friends to tell you the moment any of them starts thinking about leaving.
Principle 11
Buy intelligence, not hardware
Physical Intelligence uses arms that retail for about $3,500 and could be built for under $1,000. It does not manufacture robots. It sells, or plans to sell, the layer that controls them. The bet, in Loizos's paraphrase of Levine, is that good intelligence makes up for bad hardware.
The same reasoning ran through Groom's investing for years: back the software layer that goes across many endpoints, not one endpoint. Stripe didn't build the card networks. It built the layer that made them usable. Physical Intelligence aims to do the same for robots, which makes every hardware company a possible customer and none of them a necessary partner. The risk is that companies building robots also build their own control systems. The upside is that if the model works across robot bodies, it becomes the standard.
Tactic: In any crowded hardware category, ask which layer would make the cheapest device good enough, and build or back that layer instead.
Principle 12
Measure progress in the mundane
Moravec's paradox reverses the usual idea of what's impressive. Chess was solved decades ago. Pants are still being worked on. Physical Intelligence measures progress with laundry, zucchini, cardboard boxes, and lattes, and reports results on household chores it calls the "Robot Olympics." Its early deployments, folding laundry in short-term rentals and boxes at a chocolate maker, were chosen because the work is repetitive and low-value. That makes it a hard test for robots and a small loss for any person who no longer has to do it.
The broader lesson goes beyond robotics. Hard problems often look trivial, and impressive demos often avoid them. Groom's own life shows the contrast: his visa depended on proving "extraordinary ability," and his company depends on achieving ordinary ability.
Tactic: Define success for your product by the most boring task it must do reliably, and refuse to count demos that skip it.
Free playbook
Get The Business Model Playbook
58 business models, one visual page each: how the money flows, the metrics that matter, and who runs it. Free when you join the Faster Than Normal email.
Free. No spam. Unsubscribe anytime.
Part IIIQuotes and Maxims
In their words
— Lachy Groom, to StartupSmart, June 2012The great thing about the internet is it's not face to face. People don't know my age unless they do a little bit of research, so I can hide behind the internet in a lot of ways.
— Lachy Groom, to Anthill, 2012Start a business in reaction to a trend and you'll ride the tail end of the wave. Anticipate the wave and you'll ride it to the end.
— Lachy Groom, interviewed by Sachin Shah, 2026It feels like we've made so much progress on the Internet on software, and so little progress on technology in the physical world.
— Lachy Groom, to TechCrunch, January 2026I was looking for five years for the company to go start post-Stripe. Good ideas at a good time with a good team — that's extremely rare.
— Lachy Groom, on X, September 24, 2025the opportunity cost of a bad job has never been higher
Maxims
- Ship before you're seen. Build a record in channels where only the output is visible, then use it where people check credentials.
- Sell the reactions, hold the anticipations. Trend-driven businesses are for quick exits; anticipating the next wave is what you build on.
- Follow your scars. The operational problem that hurt you most points to the infrastructure worth working on next.
- Location is a strategy. Move toward the hub where valuations and talent are concentrated, even if it means paperwork and risk.
- Stay past the excitement. Joining early gets you access; staying long gets you responsibility for something hard.
- Unforced adoption is the strongest signal. Back products people chose before anyone required them to.
- Make speed your structure. A fund of one wins on how fast it can decide, so keep the thesis narrow.
- Give help away publicly. Today's career advice becomes tomorrow's deal flow and recruiting network.
- Wait for all three. Idea, timing, and team rarely line up; when the stakes are large, don't start with fewer than all three.
- The ordinary is the hard part. Judge progress by the dullest task done reliably, not by the most impressive demo.
