Part IThe Story
Fifteen and a Half Miles
Sometime in the small hours of a January night in 2015, a man who worked for the American government in an intelligence capacity borrowed a friend's four-propeller drone and took it out for a spin over Washington. Forbes would later report, with some restraint, that he "had had a little too much to drink." He lost the machine in the dark. He searched for it briefly, then gave up and went home. By dawn the drone, about a foot square, white-shelled, a DJI Phantom, was lying on the south lawn of the White House. The Secret Service had opened an investigation, and the photographs were on every screen in the world.
Eight thousand miles away in Shenzhen, the man who had built it was about to sit for his first interview of the year with a Western publication. The reporter, Ryan Mac, opened his eventual Forbes feature with a short inventory of the man's virtues: "Frank Wang Tao has never been arrested. He pays his taxes on time. And he rarely drinks." The joke depends on the gap between the person and the object. The person was careful, sober, and nearly invisible. The object had just gone where no hobbyist's toy was supposed to go, under the noses of the most heavily guarded people on earth.
What Wang's company did next shows how he thinks. There was no press tour and no apology from the founder. DJI spent the following morning writing a software update, and when it was pushed out, every DJI drone in the world became physically incapable of flying within a fifteen-and-a-half-mile radius centered on downtown Washington, D.C. The fix for a political problem was a circle drawn in code. A boundary that no lobbyist, regulator, or drunk could argue with had been laid over a capital city from an office in Guangdong, and almost nobody outside the hobbyist forums noticed that a private Chinese company had just shown it could do this.
The machines kept turning up in strange places that spring. A DJI Phantom 2 carrying radioactive material landed on the official residence of the Prime Minister of Japan, and the National Diet passed a law restricting drone flights near government buildings and nuclear sites. A smuggler used another to fly drugs, a mobile phone, and weapons into a prison courtyard outside London. Forbes noted that "the inconspicuous mastermind behind the world's drone revolution just shakes it off." That was not exactly indifference. It was the posture of someone who had spent his life trying to make a machine stay in the air and do what it was told. He had succeeded so completely that the remaining question was no longer whether the machine would obey. The question was whom it would obey.
As a boy he had wanted a flying thing that would follow him, watch him, and send him the pictures. He ended up building one for everyone. Most of what follows comes back in some way to that wish, and to what happens when a private dream becomes a global infrastructure, the dreamer withdraws from view, and the camera points everywhere except at him.
By the Numbers
The Drone Empire
CN¥80BDJI revenue in 2025 (about US$11.5 billion)
90%+Share of the world's consumer drone market, June 2024
¥50,000Price of the first flight-control system Wang sold, against ¥15,000 in costs
$90,000Early capital from family friend Lu Di, who also kept the books
99%Voting rights held by Wang and three early backers, per DJI's 2024 lawsuit
500M+Hectares treated by DJI agricultural drones by end of 2024
36Wang's age when he became Asia's youngest tech billionaire, 2017
The Fairy That Follows
He was born Wang Tao in Hangzhou in 1980, which in the cosmology of Chinese business is a city with a reputation. It later became the home of Alibaba, and in August 2017, when Forbes named Wang the youngest tech billionaire in Asia, the oldest and richest on the same regional list was Jack Ma. Two sons of the same city: one built a marketplace, the other built an object. The difference matters more than it seems.
His mother was, in Forbes's phrasing, a teacher turned small-business owner. His father was an engineer. The boy was not a good student. He "spent most of his time reading about model airplanes, a pastime that offered more comfort than his middling grades." The origin myth, which Wang has told several times with small variations, begins with a comic book, Dong Naojin Yeye, translated as Grandpa Think-Hard, a popular series in which an old man teaches science to his grandchildren. In one volume there was a red helicopter.
"I remember it so clearly," he told Forbes in 2015. "I imagined that I'd make a plane like that, that could follow me when I'm hiking or on the train, and use a camera to send the images to me. Actually at the time I didn't even know the concept of a camera. I just wanted to fly so badly. Even though I couldn't turn myself into a plane, I was happy to just make a small plane and watch it fly."
The machine he imagined, a small flying companion that tails you and shows you yourself from above, was something he privately called a "fairy." Keep the word in mind.
The second chapter of the myth is a crash. In elementary school he saw his first model helicopter in a shop window. "It cost the equivalent of several months' salary for average people," he recalled. "My family could not afford it." Years later, after a good result on his high-school exams, his parents bought him one as a reward. He assembled it, and it came down at once. "I wasn't able to fly it properly because to do that needed months of practice," he said. In remarks at Shenzhen University in 2015, one of his very few public appearances, he gave the physics instead of the emotion: he "was very disappointed when it crashed due to limited stability during flight. Later, I gradually got a better idea of what a perfect aircraft was and made the decision to build one."
The story has drifted in the retelling, which origin stories tend to do. In a 2018 Harvard Business Review essay, co-written by Jie Gan, who discloses that she sits on DJI's board, the helicopter is "expensive" and comes from the father, and the essay frames the whole episode as "the legend." Wikipedia puts Wang at sixteen. A French documentary turns it into a bargain: bring home better grades and I'll buy you the helicopter, after which the boy works his way to the top of his class. Whatever the details, every version keeps the same plot. A gift is earned, the gift fails, and the failure is a failure of stability. The child does not conclude that he is a bad pilot. He concludes that the machine is badly designed.
That inference is the whole company in miniature. Hobbyists of the era treated a crash as a skill problem: practice more, spend a summer learning to keep the thing level. Wang treated it as an engineering problem to be solved once, inside the machine, so that nobody would ever need the summer.
The grades did not recover enough to matter. He applied to MIT and Stanford and was rejected by both. In 2001 he enrolled at East China Normal University in Shanghai to study electronics. According to Baidu's account he left in his third year, gave up the credits, and started over. In 2003, at twenty-three, older than most of his classmates, he was admitted to the electronic and computer engineering program at the Hong Kong University of Science and Technology.
A C on Presentation Day
"At HKUST I took an electronics course that gave me the fundamental knowledge to build autonomously controlled flying machines," Wang said years later, in the kind of tidy sentence university alumni offices like to print. What the sentence leaves out is how close he came to failing.
In 2005 he was on the HKUST team that took third prize at ABU Robocon, the Asia-Pacific robotics contest. That same year, for his graduation project, he declined the faculty's list of topics and proposed his own: an easy-to-use flight-control system for model helicopters. He and two classmates got HK$18,000 in funding, about $2,300. Then he vanished into the work. "I spent three months intensively working on the project," he told Forbes. "At that time I was still enrolled at the university, but I skipped all the courses and just went to my home in Shenzhen. I would wake up at 2 p.m. and then work until like 5 or 6 a.m. for days at a time."
One detail from that stretch is worth more than anything an alumni office has published. "One time, when I did go back to the university lab, I tried to use my ID card but it didn't work. My heart sank a little bit, because I thought I was kicked out of college by my professor. But actually I had forgotten to pay my tuition."
Then the demonstration. The accounts disagree about when the helicopter failed. Wikipedia says it failed to stabilize on presentation day and crashed in front of the examiners. A Chinese-language book, DJI Drone, as cited by the South China Morning Post, says it failed the night before. Neither version matters as much as the grade, which was a C. So twice now, a helicopter Wang cared about had fallen out of the air because it could not hold itself steady.
One observer, a French documentary maker, has pointed out a coincidence in that winter, and it is hard to resist even though it proves nothing. In December 2005, Japanese customs stopped a shipping container at the port of Nagoya. Inside were autonomous helicopters made by Yamaha, bound for a Chinese company. Police carried out searches across the country, and Japan then barred its companies from exporting unmanned flying vehicles to China. In the same weeks that a student in Hong Kong was failing to make a helicopter hover, the most advanced unmanned helicopters in Asia were being placed out of China's reach by law. Nobody connected the two events at the time. The documentary maker doubts Wang was even aware of the seizure.
What rescued him was a person. Li Zexiang, a robotics professor at HKUST, had watched the project and noticed something the grade did not show: the student's command of the group and his technical understanding of the problem. Li later said that Wang's poor academic record did not reflect his excellence in practical work, and he admitted him to the graduate program. That is a teacher's judgment, the kind that rarely makes a headline, and it changed everything that followed.
A few weeks later, in January 2006, the machine worked. "The most memorable moment for me," Wang said, "was when our helicopter automatically hovered in the air in January 2006. That was a good time."
He then added a second sentence, and it is the one that stays with you: "Nowadays so much of what I do is management… like catching spies."
Fifty Thousand Yuan
What happened next had nothing to do with a business plan. Wang made a video of the hovering helicopter and posted it on a model-aircraft forum, offering the system for sale. "The people who watched it thought it wasn't bad," he said. "Someone then approached me to buy it. I thought it was a good deal because I sold it for RMB 50,000 (about $6,000) while it cost RMB 15,000 (about $2,000) to make."
Seventy percent gross margin, on the first unit, sold to a stranger on a forum. Even this early, the economics that would later make DJI enormously profitable were visible. The flight-control brain was the scarce part. The rest was hardware anyone could buy. DJI's corporate records give its founding date as January 18, 2006.
He wanted to build the company in Hong Kong. George Chen, writing in the South China Morning Post in March 2015, described what came of that: Wang's efforts "went nowhere, due partly to a lack of funding, lack of government policy support and other operational issues in the city." Chen called DJI "the technology start-up that got away." The column reads as a lament, and the irony is that Hong Kong had already tried to claim the founder. On May 27, 2014, the city's Chief Executive used his official blog to describe a meeting at which Wang, Li Zexiang, and another former HKUST teacher, Ko Ping-keung, briefed about eighty government officials on how DJI had grown from a few employees to nearly 2,000. The post was titled "Young technology entrepreneur nurtured in Hong Kong." Hong Kong had trained him, failed to keep him, and then announced his success as its own.
The company actually began a few kilometers north, across the border, in a three-bedroom apartment in Shenzhen's Lianhua Village. Wang and two classmates moved in and paid the bills with what was left of his scholarship and with money from his family. DJI sold flight-control components at roughly $6,000 each to Chinese universities and state-owned power companies, which built them into homemade unmanned aircraft. It was a niche market, specialist and unglamorous. "I didn't know how big the market could be," Wang told Forbes. "Our idea was to just make the product, feed 10 to 20 people and have a team."
That ambition is worth stating plainly, because almost no one believes it of a man later worth billions. He did not set out to build a $10 billion company. He set out to feed twenty people.
The money that kept them going came from Lu Di, a family friend, who put in $90,000 and, unusually, stayed to run the company's finances. He appears in later documents under his English name, Henry Lu, as one of four people who together hold nearly all of DJI's votes. The arrangement suggests something about the founder. Wang trusted capital only when it came with a face he already knew.
Marketing amounted to whatever they could afford. "We were mainly just going to exhibitions and putting ads on hobbyist websites," Wang said. "The first time we went to an exhibition at Nuremberg, Germany we were placed in the China section, with one of those cheap and small booths, right next to stuffed animals…We were not even in the model section, we were in the toys section."
He tells this as a joke, and it is one. It is also the clearest picture we have of where a Chinese hardware company stood in the European imagination of the late 2000s: next to the plush animals, out of the serious section, filed under cheap. Much of what DJI later did, including the obsession with polish, the borrowed prestige of Hasselblad, the Emmy, and the headquarters designed by the firm that designed Apple's, reads as a long reply to that booth assignment.
The products were real, though. The XP flight controller arrived ready-built at a time when flight controllers had to be assembled by experienced hands. RC purists sneered at it. In-house accounts say the team sold about twenty a month, through word of mouth and online forums. Peter Tao, who joined in 2008 before DJI had shipped anything, described the thinking in terms that sound like a mission statement but were really a customer profile: "We saw a real need of providing a reliable flight experience to an underserved type of user: the kind who doesn't normally know how to fly."
The kind who doesn't normally know how to fly. That was the boy with the crashed birthday helicopter, which is to say almost everyone.
It Shouldn't Be Egalitarian
In 2015 Wang sat for three interviews with a Forbes staffer, Heng Shao. In one of them he told a story about himself that founders do not usually tell, because it makes him look bad and he seems to know that.
It was the early days, with three or four employees. He wanted each to get equity, but not equal equity. Here is what he said, at length:
— Frank Wang, interview with Forbes, 2015I believed that whoever contributed more should be given more shares... It shouldn't be egalitarian. But at the end I pissed everyone off. I pissed off the person who was given fewer shares. He felt that it was unfair so he left. But I also pissed off the person who was given more, [because] he asked if he could gift his shares to the other person. I said, "No, you have to take more shares."
They argued. "My tone was insensitive and I said something like, 'At this company I am in charge.' When he left the company, he said this incident made him feel hurt."
There are two ways to read this, and both are right. In the first, it is a story about a young engineer who treats people like control loops, with output proportional to input and an error term to be corrected, and who cannot grasp that the man he favored might value the friendship more than the shares. In the second, it is a story about a founder who was correct on principle and wrong only on delivery. He did not want a cap table built on sentiment, and he was willing to lose people rather than give one up. He held the same view about votes. Two decades later, he and three early backers still control 99 percent of them.
The cost was steep. "Almost all of the first group of employees left after two years," Wang said. "They may have felt the future was a question mark… I was a perfectionist, and was abrasive when it came to interacting with other people." Both founding classmates were gone. Then came something worse than churn: an early employee tried to steal DJI's intellectual property and sell it to competitors. Wang wanted to sue and could not pay the legal fees, about 70,000 yuan.
Li Zexiang stepped in again. The professor helped settle the dispute, raised another million yuan for the company, and joined DJI as chairman. It is worth looking at what Li supplied. He brought money, lawyers, recruits, and above all an adult in the room. Wang still wrote the code, and Li stabilized the company around him, the way a flight controller holds an unstable airframe level.
In 2008 or 2009 (Baidu says the former, Wikipedia the latter), DJI's XP3.1 flight-control system went to the Himalayas. Baidu records that Wang and Li ran a flight test themselves in the Mount Qomolangma region of Tibet. The footage of an unmanned machine flying around the peak of Everest did what years of trade-show booths had not. It drew steady demand for the XP3.1, and it put the company in front of the world for the first time. The highest place on earth turned out to be DJI's best advertisement. That will matter at the end of this story.
The other person who stayed had known Wang the longest. Swift Xie Jia was a friend from high school. He joined in 2010 to run marketing and to act, in Forbes's word, as a confidant. At some point he sold his apartment and put the money into DJI, which made him one of its largest individual shareholders and the fourth member of the bloc that would later hold nearly all its votes. The company's inner circle, then, is made up of a family friend, a professor, and a classmate from Hangzhou. Every one of them is someone Wang knew before the company was worth anything.
In 2011 he finally submitted his master's thesis, titled Control systems for autonomous helicopters. The work had been finished, in practice, years earlier in a dormitory. It had taken him five years to put it on paper, because he had been busy building the thing it described.
The company's name, 大疆, is usually translated as "Great Frontier." It holds both a sense of open territory and the old meaning of a border region, the edge of an empire. At the time it was a joke at the company's own expense, or a prophecy. It was not yet clear which.
The Game-Show Contestant
In 2011, at a trade show, Wang met Colin Guinn. Guinn had once been a contestant on a television game show, and he understood something Wang did not yet understand: Americans with cameras would pay a great deal to get those cameras into the air. The two founded DJI North America, a subsidiary meant to sell to the mass market. The pairing was strange from the start. One man had been rewarded for performing well on television. The other had been rewarded, in the end, for staying off it.
The product Guinn was selling arrived in January 2013. The Phantom was the first DJI machine sold fully assembled, a white quadcopter priced at $629 that you could take out of the box, charge, and fly, with a mount for a GoPro. Popular Mechanics later described it as "the first gyro-stabilized quadcopter with GPS that came in a nice, pre-assembled package," contrasting it with earlier quadcopters that were "extremely difficult to control." The flight controller from the dormitory now lived inside a shell that a real-estate agent could operate.
In Chinese, DJI called it 精灵. The word means a sprite, a spirit, a fairy.
There is no record that Wang ever said the name came from the comic book, and it would be overreaching to claim that it did. The rhyme is there whether or not anyone intended it. The flying companion the boy imagined, the one that would follow him on the train and send back pictures, went on sale in 2013 under the name he had privately given it.
DJI expected the Phantom to roughly break even. The thinking was defensive. An entry-level product priced near cost would discourage competitors from starting a price war at the bottom of the market. That expectation turned out to be badly wrong. The Phantom quickly became DJI's best-selling product and multiplied the company's revenue about fivefold. Wang, by his own account, was surprised by the demand, and DJI turned toward consumer drones that were high quality and affordable. The customer he had built the XP for, the person who didn't normally know how to fly, turned out to be the whole market.
That success broke the partnership with Guinn. In the middle of 2013, Wang offered to buy Guinn out, and Guinn refused. By the end of the year, DJI had locked every employee of the North American subsidiary out of their email accounts and was well on its way to shutting the operation down. Guinn sued, and the case was settled out of court. Commentators have since read the episode as a clash between a showman and an engineer, and between self-promotion and the product. That reading may be fair, but it is still interpretation. The facts are colder than the interpretation. Wang made an offer, the offer was refused, and the email accounts went dark.
Compare this to the shares story. There, Wang told a man he had to accept more of the company than he wanted. Here, he took the company back from a man who wanted to keep his share. The principle is the same in both cases: Wang alone decides who holds how much of DJI. In 2015 Forbes estimated his personal stake at about 45 percent. A decade later, DJI's own court filings put the combined founder group at roughly 87 percent of shares and 99 percent of votes. Every outside investor, state-linked or not, sits behind that line.
A Distance From the Masses
The years from 2014 to 2017 are when the company stopped being a story and became a fact.
In September 2014 the Phantom went on sale in Apple Stores. By early 2015 the U.S. Federal Aviation Administration, which had only just opened American skies to commercial drone operators, had approved ten times as many DJI drones as those of all its competitors. Reuters reported in April 2015 that DJI held "about 70 percent of the commercial market worldwide" and had brought in nearly $500 million in revenue in 2014. One trade profile put 2014 unit sales at roughly 400,000 drones and said revenue had quadrupled every year since 2009. In May 2015, Accel Partners put in $75 million at an $8 billion valuation, along with a joint developer fund called SkyFund. Later that year a $500 million round brought in GIC, the state-owned New China Life Insurance, and New Horizon Capital, a firm co-founded by the son of former Chinese premier Wen Jiabao. That cap-table entry would later be read aloud in Washington. In 2015 it looked like a sign of strength.
Revenue reached about $1 billion that year. In 2017 it was 18 billion yuan, about $2.7 billion, up 80 percent. Forbes noted that roughly 70 percent of sales came from outside Asia, unlike Alibaba or Xiaomi, which had grown mostly by serving the enormous Chinese domestic market. That was the real departure. DJI was arguably the first Chinese company to lead a global technology category from the start, rather than copying one, catching up, or defending a protected home market. The company that had been placed next to the stuffed animals in Nuremberg now set the reference standard that every other drone was measured against.
Wang's own explanation was characteristically short. "In the very beginning, we had different competitors but they were small," he told NetEase in 2015. "We made a lot of the right decisions to stand out in the industry … I think it is DJI's success that made the drone industry attractive to investors and users." To Forbes, the same year, he gave the line that has followed him ever since:
— Frank Wang, interview with Forbes, 2015A distance from the masses. If you can create that distance, you will be successful.
Read without context, it sounds like contempt, and some people heard it that way. Read beside the rest of his career, it sounds more like a technical specification. The distance he meant was a lead in capability large enough that competitors spend their energy explaining why their product isn't a DJI. The Phantom 3, in 2015, added a built-in camera that streamed live video, and that widened the gap again. Rivals left the market. The Mavic Pro in 2016 folded small enough to fit in a jacket pocket. The Mavic Mini in 2019 was light enough to fall below many of the regulatory weight thresholds that governed drones, so the product's design absorbed the law instead of fighting it. By March 2020 DJI held 77 percent of the American consumer market, and no other company had more than 4 percent.
Inside, the company ran on competition. It became known for a brutal hiring process and for setting internal teams against each other to design better products. Its Shenzhen factories ran sophisticated automated assembly lines, and many of the components for those lines were built in-house. The first DJI product was a component nobody else could build well, and the company kept that reflex, including for the machinery that builds its products.
The product names are worth a moment, because in Chinese they read like a book of myths. The heavy-lift hexacopters were called Spreading Wings in English and 筋斗云 in Chinese, the somersault cloud the Monkey King rides in Journey to the West. The early frames were Flame Wheel, 风火轮, the wind-and-fire wheels of the boy-god Nezha. The film-camera stabilizer is Ronin in English and 如影 in Chinese, "like a shadow." The handheld camera is Osmo in English and 灵眸 in Chinese, a "spirit eye." The consumer flagship, Mavic, is 御, the character for driving a chariot, for command, for what belongs to the emperor. In English the names are sleek Latinate inventions that tell an American buyer almost nothing. In Chinese they promise that the gods' vehicles can now be bought at retail. DJI has always spoken to two audiences at once, and it has always told each of them a slightly different story.
In November 2015 DJI took a stake in Hasselblad, the Swedish camera maker whose name is tied to the Apollo program, and later took majority control. A decade on, the Mavic 4 Pro, released May 13, 2025, flies a 100-megapixel Hasselblad sensor on a gimbal that rotates a full 360 degrees. In 2017 DJI won a Technology and Engineering Emmy for camera-drone work used on Game of Thrones, Better Call Saul, and The Amazing Race. The new headquarters, Sky City, was commissioned from Foster + Partners, the British firm that designed Apple Park. When it opened in September 2022, Wang said in a statement that the building would offer a better working environment, and, more importantly, that DJI's pursuits and dreams were being elevated too. He tends to describe even his architecture in terms of altitude.
None of this made him visible. In 2019, when the South China Morning Post asked to interview him, a DJI spokeswoman, Natasha Gray, declined on his behalf: "Frank was passionate about helicopters when he was a child … He is very private and prefers the spotlight [to be on the products]." He had built a machine whose whole purpose was to make things seen, and he had arranged for it to look at everything except him.
Catching Spies
That aside about the spies is easy to read as a joke. It was not a joke.
On January 21, 2019, DJI announced that an internal investigation had found "extensive" fraud. Employees had "inflated the costs of parts and materials for certain products for personal financial gain." The company estimated the losses at "up to CN¥1 billion," about $147 million, and said it had "not incur[red] a full-year loss in 2018." Baidu dates Wang's full anti-corruption campaign to that same month. The supply chain that gave DJI its edge, with its thousands of parts, hundreds of vendors, and suppliers a short subway ride from the office, had become a surface for theft at the scale of the company's success. The perfectionist who once fought over a few percentage points of equity with three employees was now policing procurement in a company of 14,000.
Years later he would tell an interviewer that the company had gone through a crisis of corruption and factions brought on by rapid growth, and that he had spent eight years rebuilding it from an organization dependent on individual ego into one based on rules and systems. Eight years puts the start of that work right around the fraud announcement.
The spies outside were a different matter, and much harder to catch, because it was never certain they existed.
The trouble began in software. In July 2017 the U.S. Interior Department's aviation office found that DJI's software did not meet its information-sharing requirements. In August, The Register reported that the DJI GO app contained a framework called JSPatch, which let DJI change the app without Apple's review or the user's consent. That same month the U.S. Army ordered its personnel to stop using DJI products, citing an Army Research Laboratory report on cyber vulnerabilities. Late in 2017 DJI launched a bug bounty program, which was still unusual for a company its size, and handled it badly. The non-disclosure terms were not ready. A researcher named Kevin Finisterre turned down a $30,000 payment and published what he had found: among other things, DJI had accidentally left its private SSL keys exposed on GitHub. What should have been a story about openness became a public-relations disaster.
The company answered with audits. Kivu Consulting in 2018, Booz Allen Hamilton in 2020, and FTI Consulting in 2024 each concluded, in their own wording, that they had found no unauthorized transfer of data to China. The one exception was a crash-reporting tool that sent logs to Chinese servers. In May 2021, the unclassified portion of a Pentagon analysis found "no malicious code or intent" in two DJI Government Edition models and recommended them for government use. The government acted as though it had not read the report.
The sequence of sanctions is worth reading in a single breath. In January 2020, the Interior Department grounded about 800 DJI drones it had been using for wildlife conservation and infrastructure monitoring. In December 2020, the Commerce Department added DJI to the Entity List. In January 2021, an executive order required Chinese-made drones to be removed from federal fleets. In December 2021, the Treasury barred Americans from investing in DJI, accusing it of helping the People's Liberation Army and of complicity in the persecution of Uyghurs. In February 2022, The Washington Post reported that DJI had obscured its ties to Chinese state funding. In October 2022, the Defense Department listed DJI as a "Chinese military company." In September 2024, the House passed the Countering CCP Drones Act, and in October, Customs began holding some DJI imports under the Uyghur Forced Labor Prevention Act. In December 2025, the FCC banned imports and sales of new DJI drone models.
DJI fought back in American courts. It sued the Pentagon on October 18, 2024, arguing that it "is neither owned nor controlled by the Chinese military," and in that filing it disclosed the details of its ownership for the first time: four founders with 99 percent of the votes, and state-linked funds with 5.4 percent of the shares and 0.6 percent of the votes. In September 2025 a federal court ruled against DJI, finding that the Defense Department had substantial evidence that the company contributes to the "Chinese defense industrial base." In February 2026 DJI sued again, this time over the import ban. It had also hired lobbyists. Two of the firms dropped the account in February 2024 when representing Chinese clients became politically expensive.
There were two ways to tell this story, and the industry press told both. Scott Simmie, a drone-industry journalist, argued that DJI's real weakness was where it came from, and that repeated, unproven claims had pushed the company into a permanent defensive crouch. He went further, on what he acknowledged was circumstantial evidence, and suggested a coordinated campaign against it. Reporting in 2024 found that the American rival Skydio, among others, had paid for lobbying in favor of a ban. In September 2024 the Government Accountability Office reported that the restrictions had hurt the Interior Department's own operations. The department had quietly allowed its grounded DJI drones back into the air for wildfires and later for routine work, after deciding the risk was low. The other version of the story was just as well documented: a company whose investors included state funds, whose drones flew with Chinese police in Xinjiang, and which, as a court eventually concluded, belonged to a defense ecosystem whether it wanted to or not.
Wang said nothing publicly through all of it. His wealth, estimated by Hurun at 50 billion yuan in 2023, fell to 35 billion yuan the next year.
Early in 2026 there was one more security story, smaller and stranger. Sammy Azdoufal, a French engineer living in Spain, was trying to steer his own new DJI robot vacuum, the Romo, with a PlayStation controller. In the process he found he could control about 7,000 of them. DJI thanked him publicly on X, and The Verge reported in March that the company had paid him $30,000 for one of his findings. It was the same amount Finisterre had turned down nine years earlier. This time the company paid it and said thank you. The amount was unchanged. The company had learned something.
Everyone's Eye in the Sky
Look at what the machine was used for, and the boy's wish starts to look like something else.
In 2016 ISIS used DJI drones as explosive devices in Iraq. DJI's response was the same one it had used for Washington, scaled up. It quietly placed a no-fly zone over nearly all of Iraq and Syria. The same year a DJI drone nearly collided with a Chinese fighter jet, and Beijing insisted that DJI build a registry to track its drones inside China. In 2017 DJI signed a strategic cooperation agreement to supply surveillance drones to police in Xinjiang. On March 30, 2018, the Israel Defense Forces used DJI Matrice 600s to drop tear gas on protesters in Gaza. On August 4, 2018, two Matrice 600s carrying explosives detonated near Avenida Bolívar in Caracas in what appeared to be an attempt to assassinate Nicolás Maduro. On June 5, 2018, Axon, the maker of the Taser, announced it would sell DJI drones to American police departments. By 2020 about 90 percent of drones flown by U.S. public-safety agencies came from DJI, and by 2022 more than 500 American law-enforcement agencies were using them. During the pandemic, Chinese police flew them to remind people to wear masks, and authorities in Morocco and Saudi Arabia used them to spray disinfectant on streets and take people's temperatures from the air.
The red helicopter in the comic followed a child on a hike. The real one followed everyone, on behalf of whoever bought it.
Then came Ukraine. In March 2022, Ukraine's vice prime minister, Mykhailo Fedorov, wrote publicly to Wang. He said DJI's AeroScope detection system was helping Russian forces locate Ukrainian drone operators, and he asked DJI to restrict its products. DJI said its drones were civilian tools, unsuited to war because of features like AeroScope and Remote ID that made them visible. It said it could geofence parts of Ukraine if the government formally asked, but that this would ground every side's drones equally and could be defeated by anyone who simply refused software updates. The Verge found no confirmed reports that Russia was in fact using AeroScope. In April 2022 DJI suspended all business in both Russia and Ukraine.
That did not matter. Buyers and donors moved the drones across borders through intermediaries and modified the software to get around the restrictions. Hobbyists who already knew how to fly were recruited by both armies. Ukrainian troops used DJI aircraft for reconnaissance, for dropping improvised explosives, and for filming the destruction of Russian armor, footage that circulated as propaganda. In October 2023 the Ukrainian government announced it had bought 4,000 DJI drones and planned to buy 20,000 by May 2024. A machine designed for the person who didn't normally know how to fly had become, in Foreign Policy's phrase, part of a war that was "cheap, deadly, and made in China."
Al Jazeera reported in 2025 that the Israeli military had retrofitted DJI drones, including Agras crop-sprayers, for surveillance and bombing in Gaza, and that DJI, unlike in Ukraine, had not stopped sales to Israel. In January 2025, around the time that series of reports was taking shape, DJI pushed an update in the United States that removed its old geofences over military bases, airports, power plants, and the White House itself. Foreign Policy's headline called it a security rollback "at a very weird time." The fifteen-and-a-half-mile circle from 2015, the company's first act of private airspace policy, had been erased by the same kind of software update that created it.
The agricultural numbers belong in this section too, because the same machine did both kinds of work. DJI's Agras sprayers, launched in 2015, had treated more than 500 million hectares by the end of 2024 and, by the company's count, saved about 222 million tons of water. An aircraft that drops grenades on tanks in one place sprays rice paddies in another, and it is the same flight controller in both.
And So Am I
In April 2024 DJI donated two drones to a cleanup effort on Everest. In April 2025, in temperatures of minus 25 degrees Celsius and winds of 45 kilometers an hour, a DJI aircraft carried 15 kilograms to Camp I at 6,130 meters, reported as the highest-altitude drone delivery ever attempted. In December 2025 Nepal approved DJI's heavy-lift drones for use on the mountain. Seventeen years after the footage of the summit that first put DJI on the map, the company's machines were working the same slope, hauling the loads that Sherpas had always carried.
On April 30, 2026, Nepal's Home Ministry canceled the flight permit for DJI's FlyCart 100. It also refused an application to fly an American alternative, the Freefly Alta X. An official cited security concerns and heightened sensitivity about where the drones were made. A retired Nepali major general described the decision as an effort to stay strategically neutral between the two superpowers. A few days later a large serac collapsed in the Khumbu Icefall, trapping and injuring climbers. The rescue went ahead without drones. Expedition operators protested, and the ban was lifted after five days.
So the geofence had turned around. For a decade DJI had decided, from Shenzhen, where its machines could and could not fly over other people's countries. Now another country had drawn a circle around DJI, on the very mountain that had made its name, at the moment the aircraft were most needed.
That spring Wang gave his first in-depth interview in ten years, to the Chinese business publication LatePost. The ten-year silence had left behind a single line, endlessly repeated: The world is incredibly stupid. He now gave the line a second half.
— Frank Wang, interview with LatePost, spring 2026The world is incredibly stupid. And so am I.
The interviewer had expected the cold, tense "President Wang" of office rumor. Instead she found someone relaxed and calm, who laughed loudly when he was in the mood. His white shirt was tucked into his trousers with one corner always hanging out, and he did not seem to care. He talked about the corruption and the factions, and about the eight years of rebuilding. He said management mattered as much as product. He described the departures of early core members, the classmates and the hurt shareholder among them, as a kind of healthy organizational metabolism. He said he had once thought of competition as crushing the enemy and now thought of it as improving himself and raising the standard for the whole industry. He said DJI would not build cars. His aim was for the company to overtake Sony in imaging within ten years. Its revenue the previous year had been estimated above 80 billion yuan, with profits above 20 billion, and he expected sales to pass 100 billion this year. He had just spent a week with the AI coding tool Cursor building his own editor for designing organizations and processes, which he called his "Lego." Asked what all of it had been for, he said: seeking truth with companions.
Sky City rises over Liuxiandong, the company's machines fly over every continent, and the man has been out of sight for ten years. He keeps four cats in his office, and he weighs them every day, watching for small changes in their appetite and their weight.
Part IIThe Playbook
What follows are twelve principles drawn from Frank Wang's twenty years at DJI. They are not a recipe. Several of them come at a real cost, and several have already been revised by the man who practiced them. They are presented as patterns: decisions he made repeatedly, under different conditions, that compounded into the most dominant position any company has held in a new hardware category in this century. Where the evidence suggests a principle failed, or has run past its useful life, that is said.
The principles
- 1.Treat failure as a design flaw, not a skill gap.
- 2.Sell the prototype before you believe in the market.
- 3.Build the component nobody else will build.
- 4.Design for the person who doesn't know how to fly.
- 5.Price the entry point to prevent the war.
- 6.Sell the shares, keep the votes.
- 7.Recruit the stabilizer you cannot be.
- 8.Never outsource the relationship with the customer.
- 9.Ship policy as code.
- 10.Audit your own supply chain as if it were an adversary.
- 11.Silence is a strategy until someone else writes your story.
- 12.Rewrite yourself in public before the world does it for you.
Principle 1
Treat failure as a design flaw, not a skill gap
The founding event of DJI is a reward helicopter that crashed. The second is a graduation helicopter that crashed. Wang came away from both with the same lesson, and it was not that he needed to practice more. It was that the machine needed to hold itself steady. That shift moves the burden of competence from the user to the product, and it is the most valuable idea in the company's history.
Most technical cultures, hobbyist RC flying in the 2000s among them, treat difficulty as a badge of membership. The purists who dismissed the XP flight controller did so because it made something easy that had been hard. Wang's conclusion, repeated across fifteen years of products, was that every hour of practice the hobby demanded was a market the hobby had failed to serve. The Phantom, the Mini's weight under regulatory thresholds, and obstacle avoidance are all the same instinct applied again: when a user fails, ask what the product failed to do for them.
Tactic: For every complaint that begins "users just need to learn," write down the product change that would make the learning unnecessary, and put it on the roadmap.
Principle 2
Sell the prototype before you believe in the market
Wang did not commission market research. He posted a video of his helicopter hovering to a forum and sold the system for 50,000 yuan against 15,000 yuan in costs. Even then, he still thought the company's ceiling was feeding "10 to 20 people." The sale was not proof of a vision. It was the information that made a vision possible.
This order of events matters. Founders who wait until they believe in a large market before they sell anything tend to design for the market they imagine. Wang designed for the stranger who actually showed up, and let each sale tell him what came next: universities, then state power companies, then hobbyists, then, to his own surprise, everyone. The 70 percent margin on the first unit was a signal about where the scarce value lay. The flight-control brain was the asset, and everything else could be bought.
Tactic: Put your earliest working prototype in front of the most specialized community that might pay for it, and let the price they accept, not the size of the market you imagine, decide your next move.
Principle 3
Build the component nobody else will build
DJI's own developers describe the company's history as inventing the building blocks first: flight controllers, gimbals, video downlinks, propulsion. The fully assembled drone came after, almost as a consequence. By the time the Phantom shipped, DJI owned every critical subsystem inside it. Later the company built the components for its own automated assembly lines in-house as well.
This is vertical integration that grew out of necessity rather than strategy. In 2006 nobody sold a reliable plug-in flight controller, so DJI made one. Nobody sold a stabilizer good enough for aerial video, so DJI made that too. Each forced investment became a moat. Rivals assembling off-the-shelf parts could not match DJI's integration, its cost structure, or its speed. The "distance from the masses" Wang described was largely the distance between a company that owned its components and companies that rented them.
Tactic: Find the one subsystem your product depends on that no supplier does well, and build it yourself even if it costs more today, because the margin and the moat will both come from there.
Principle 4
Design for the person who doesn't know how to fly
Peter Tao, who joined DJI before it had sold anything, defined the company's customer as "the kind who doesn't normally know how to fly." That is not a demographic. It is a stance. It assumes the largest market is always made up of people currently shut out of the category, not the experts already in it.
The Phantom proved the point when its demand surprised its own maker. A product intended to deter price wars at the low end turned out to be the company. The lesson generalizes: the experts in any category are the loudest voices and the smallest group. Wang's habit was to design for people who would never post on an RC forum, such as the real-estate agent, the wedding videographer, and the farmer. Later the same habit produced the Osmo Pocket and the gimbal for smartphones, products that took a professional capability and handed it to beginners.
Tactic: Write a one-line profile of the person your category currently excludes, then test every product decision against whether that person could succeed on the first try without help.
Principle 5
Price the entry point to prevent the war
DJI priced the first Phantom at $629 and expected roughly to break even. The aim was not profit. It was to deny competitors an opening at the bottom of the market, where a price war could have drained the company. The plan misfired in the best possible way, because the defensive product became the best-seller and multiplied revenue fivefold.
The underlying logic still holds. A dominant technology company can choose to occupy the low end of its own market before someone else does, accepting thin margins there in order to keep the high end safe. By 2017 analysts noted that DJI largely stayed out of the toy segment under $500 and focused on the more serious tiers. The company had learned which entry points it needed to hold and which it could leave to others. Being in the entry tier was always a deliberate strategic decision, never a reflex.
Tactic: Identify the lowest price point at which a credible competitor could gain a foothold in your category, and decide deliberately whether to occupy it yourself, even at break-even.
Wang's capital history is a long exercise in separating money from control. Lu Di's $90,000 came from a family friend who stayed to run the books. Accel's $75 million came at an $8 billion valuation. State-linked funds came in later. Through all of it the founder group held the votes. DJI's October 2024 lawsuit disclosed the result: four people hold 99 percent of the votes and about 87 percent of the shares, and state-linked investors hold 5.4 percent of the shares and 0.6 percent of the votes.
The same instinct appears at both ends of the story. In the early share dispute, Wang insisted on deciding the distribution himself, even when that meant forcing more equity on a man who did not want it. With Colin Guinn, he took back the North American business when Guinn refused a buyout. Control, for Wang, was not a matter of pride. It was what let him make decisions on engineering timescales instead of investor timescales. He paid for it with friendships, a lawsuit, and years of distrust from people who thought a company that would not open its books had something to hide.
Tactic: Before each financing, decide in writing what voting share the founding group must keep, and treat any term that crosses that line as a reason to walk away rather than a point to negotiate.
Principle 7
Recruit the stabilizer you cannot be
Wang is candid that his early leadership drove people away: "I was a perfectionist, and was abrasive when it came to interacting with other people." Almost the whole founding team left within two years. When an employee tried to steal DJI's intellectual property, Wang could not afford the 70,000-yuan legal fees. Li Zexiang, the professor who had seen past the C, settled the dispute, raised a million yuan, and became chairman.
DJI's controlling bloc is a portrait of the people who balanced Wang out: Li, who brought legitimacy and judgment; Lu Di, who handled the money; and Swift Xie, a high-school friend who sold his apartment to invest and served as marketer and confidant. None of them was hired from a recruiter's list. Each made up for something specific the founder lacked. Wang did not pretend to have those qualities himself. He found the people who had them.
◎
The Controlling Bloc
Four people hold 99% of DJI's votes. Each supplied something the founder lacked.
| Person | Origin of the relationship | What they supplied |
|---|---|---|
| Frank Wang | Founder | Product The flight controller, and the obsession behind it |
| Li Zexiang | HKUST robotics professor who saw past the C | Legitimacy Capital, recruits, legal rescue, chairmanship |
| Lu Di (Henry Lu) | Family friend | Capital $90,000 and the company's books |
| Swift Xie Jia | High-school friend from Hangzhou | Voice Marketing, confidence, an apartment's worth of conviction |
Tactic: List your three worst traits as a leader, and name one person, with a real stake and real authority, whose job is to make up for each.
Principle 8
Never outsource the relationship with the customer
DJI North America was a necessary partnership. Wang needed someone who understood American buyers, and Colin Guinn did. But the partnership put a separate company with its own founder between DJI and its biggest market. Once the Phantom made that market valuable, the arrangement could not last. The buyout was refused, the email accounts were cut off, and the lawsuit was settled.
The episode was ugly, and a more diplomatic founder would have handled it differently. Its strategic logic is still clear. A hardware company whose customers belong to an intermediary is a parts supplier, whatever its brand says. Wang's reclaiming of North America, and later DJI's 17 international offices and its own retail stores, meant the company's feedback came straight from its users. That feedback is what let it ship new generations faster than anyone else.
Tactic: If a partner currently owns the relationship with your most important customers, negotiate a clear path to taking it back before that market becomes valuable enough to fight over.
Principle 9
Ship policy as code
When a Phantom landed on the White House lawn, DJI did not issue a statement. It issued firmware, and every DJI drone could no longer fly within fifteen and a half miles of downtown Washington. When ISIS turned its drones into weapons, DJI geofenced nearly all of Iraq and Syria. When Beijing asked for a flight registry, DJI built one. The company's answer to political risk was nearly always technical, quick, and universal.
This is a powerful capability, and the record shows how far it reaches. Geofencing let DJI handle safety problems without waiting for regulators to act. But it also showed that a private company could control airspace over sovereign nations, which is part of why governments came to see it as a strategic actor rather than a gadget maker. In Ukraine, DJI noted that a geofence could not tell one side's operators from the other's and could be defeated by anyone who refused updates. The January 2025 removal of U.S. geofences showed that whatever code can grant, it can also take back. Policy written as code is fast, and it is also visible to everyone.
⌖
Policy by Firmware
DJI's habit of answering political events with software updates.
2015
A Phantom crashes on the White House lawn; DJI pushes a no-fly zone covering a 15.5-mile radius around downtown D.C.
2016
After ISIS uses DJI drones as weapons, DJI geofences nearly all of Iraq and Syria.
2017
Local Data Mode added so drones can operate without transmitting data, after U.S. Interior Department concerns.
2022
Ukraine asks DJI to restrict Russian use; DJI offers a geofence and warns it would ground all sides equally.
2025
DJI removes U.S. geofences over airports, bases, power plants, and the White House.
Tactic: Before the next crisis, decide which problems your product could solve automatically through an update, and which would be safer left to regulators, so you are not improvising sovereignty under pressure.
Principle 10
Audit your own supply chain as if it were an adversary
"Nowadays so much of what I do is management… like catching spies," Wang said, long before DJI disclosed in January 2019 that employees had inflated the cost of parts and materials, at a loss of up to one billion yuan. The same feature that gave DJI its edge, a vast and fast-moving supplier network next door in Shenzhen, created countless openings for theft at a scale that grew with the company.
The deeper lesson is about timing. A founder's perfectionism works well when the founder can see every part of the operation. Once the company has 14,000 people, attention cannot reach everything, and things begin to rot where nobody is looking. Wang has since said that rapid growth brought both corruption and factions, and that it took about eight years to move DJI from a company run on individual ego to one run on rules. The anti-corruption campaign was the start of that process, not the end.
Tactic: Once your procurement passes the point where you personally know every major supplier, fund an internal audit function that reports to you directly and assumes, until shown otherwise, that costs are being inflated somewhere.
Principle 11
Silence is a strategy until someone else writes your story
For a decade Wang let the products speak for him. "He is very private and prefers the spotlight [to be on the products]," his spokeswoman said. The silence kept attention on the hardware, avoided the founder-cult dynamics that have damaged other technology companies, and suited a man who by his own account handled people badly.
In Washington, though, the silence was filled by other people. While Wang said nothing, the U.S. government built a nine-year record of restrictions. Rivals paid for lobbying in support of a ban. And the version of DJI that most policymakers came to believe was written by DJI's critics. The company hired lobbyists late and lost two of them to political pressure. Independent audits clearing DJI's data practices hardly registered. In the end, the most detailed account of who owns DJI came out in a lawsuit, filed years after the questions had started.
◐
The Cost of Silence
How the absence of a founder voice shaped DJI's two reputations.
| Arena | What silence achieved | What silence cost |
|---|---|---|
| Consumers | Strength The brand became the product; DJI became the word for "drone" | Very little |
| Internal culture | Strength Attention stayed on engineering | Risk Problems grew out of sight until the 2019 fraud |
| Washington | Nothing measurable | Risk Critics defined the company first; sanctions came in sequence |
| Ownership questions | Privacy for the controlling bloc | Delay First detailed disclosure came in a 2024 lawsuit |
Tactic: Map every audience whose decisions could shut your business down, and make sure at least one of them hears your account of the company from you before it hears another account from anyone else.
Principle 12
Rewrite yourself in public before the world does it for you
The most surprising part of Wang's 2026 interview was not the revenue figure. It was the revision. The man whose most quoted line was "the world is incredibly stupid" added "and so am I." He took apart his earlier positions one at a time. Competition was no longer about crushing enemies. Departures were no longer betrayals. Management was not secondary to product. He also set firm limits: DJI would not make cars, and its stated target was to overtake Sony in imaging.
Founders who build very large companies often get trapped by the version of themselves that built them, and by the time they are ready to change, the market has already moved on. Wang's revisions came after the fact, and they came while the American market was closing, Nepal had grounded his aircraft on Everest, and his net worth had fallen. They came all the same, and they came in his own words. Admitting that he had been wrong is now part of how he leads. A boundary drawn by the founder himself, such as no cars, is more believable than a dozen strategy decks.
Tactic: Once a year, write down the three beliefs that built your company that you now think are wrong, and share them with your leadership before circumstances force the issue.
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
— Frank Wang, interview with Forbes, 2015There was a comic book in the 1980s called "Dong Naojin Yeye" (Grandpa Think-Hard) which featured a red helicopter. I remember it so clearly. I imagined that I'd make a plane like that, that could follow me when I'm hiking or on the train, and use a camera to send the images to me.
— Frank Wang, interview with Forbes, 2015I didn't know how big the market could be. Our idea was to just make the product, feed 10 to 20 people and have a team.
— Frank Wang, on DJI's first trade show in NurembergWe were placed in the China section, with one of those cheap and small booths, right next to stuffed animals…We were not even in the model section, we were in the toys section.
— Frank Wang, interview with Forbes, 2015The most memorable moment for me was when our helicopter automatically hovered in the air in January 2006. That was a good time. Nowadays so much of what I do is management… like catching spies.
— Frank Wang, HKUST alumni profileI started my company in the first year of my PG study and received lots of help from professors who believed in my dreams.
Maxims
- Blame the machine first. When a user fails, look for the product's failure before the user's.
- Let the first sale teach you. A stranger paying 50,000 yuan tells you more than any market-size estimate.
- Own the scarce part. Make the component nobody else can make well; buy the rest.
- Design for the excluded. The biggest market is the people who don't yet know how to fly.
- Hold the entry tier on purpose. A cheap product that blocks a price war can end up being the whole company.
- Separate money from control. Sell the shares if you must. Keep the votes.
- Hire your counterweight. Find the people who can steady you where you are unsteady.
- Code is policy. If your product can draw boundaries, decide how it will draw them before a crisis decides for you.
- Silence has a cost. Not speaking for yourself leaves others free to speak for you.
- Revise yourself early. Say you were wrong before circumstances prove it.
