Not Smart, Just Hard-Working: Mark Cuban on AI

On 26 October 2026, Steven Bartlett, the British entrepreneur and host of The Diary of a CEO, released a conversation lasting a good hour and a half with Mark Cuban. Cuban is one of the best-known entrepreneurs in the US: he became rich by selling his internet company Broadcast.com, owned the Dallas Mavericks basketball team for many years and sat as an investor on the TV show Shark Tank, the American counterpart of Germany’s “Die Höhle der Löwen”. The conversation covers selling and hard work, but above all artificial intelligence, a possible stock market bubble and the debt with which American corporations are building their data centres. We listened to the whole episode and checked his figures. Watch from 0:02:06

The GuestWho is speaking

Cuban is 68 years old and grew up in Pittsburgh. His father upholstered car seats and, in Cuban’s words, never earned more than 40,000 dollars a year. At twelve, Cuban sold rubbish bags door to door; a friend of his father’s let him have them for three dollars a pack, and he charged six. Watch from 0:05:45 His first company, MicroSolutions, a computer reseller and service provider, he sold, by his own account, for five million dollars, of which he kept two. Broadcast.com, an early service for radio and video on the internet, was bought by Yahoo in 1999 for 5.7 billion dollars in stock. Today he spends most of his time on Cost Plus Drugs, a mail-order pharmacy that sells medicines at cost plus a fixed mark-up.

Cuban is not an AI researcher but an entrepreneur and investor. He uses AI every day and makes money from companies that depend on it. That is worth bearing in mind when reading, as is his fondness for exaggeration.

SellingIf you can sell, you will always find work

For Cuban, all success starts with selling. Anyone who has learned to deal with a no at strangers’ doors, to answer objections quickly and to keep going anyway “has a job for life”. Watch from 0:06:32 No business, he says, has ever succeeded without sales. But good selling does not mean persuading; it means understanding what the other person needs and helping them. Watch from 0:06:40

Anyone who wants to be an entrepreneur should not pursue three ideas at once but become really good at one thing. Watch from 0:12:07 You get good where you spend your time, not at whatever you supposedly have a passion for. Watch from 0:12:36 His advice to graduates: don’t look for the perfect job, take any job and become better at it than everyone else. Then doors open, even after two or three years at a fast-food chain. Watch from 0:19:06

The EdgeEveryone else shares second place

When the first office computers arrived in the 1980s, Cuban stayed late at night in the computer shop or took the machine home and read the manuals until he knew them better than any customer. His reasoning: only the developer knows a new program better; everyone else shares second place. Whoever makes the most effort stands there alone. Watch from 0:24:06 He approaches AI the same way today. He learns every new model faster than others. Watch from 0:24:45

How big is the upheaval? Cuban has been in the technology industry for decades and says AI overshadows everything he has experienced before put together. Watch from 0:25:05 Used properly, it makes knowledge accessible to everyone, every library, every book, every speech. Watch from 0:25:26 Bartlett describes how many people push the subject aside because it calls their profession or their degree into question. Cuban counters that where things change there are opportunities, and most people won’t do the work anyway. Watch from 0:26:37 He quotes the basketball coach Bobby Knight: everyone has the will to win, but only those who also have the will to prepare will win. Watch from 0:27:11

The AlgorithmsQuick money in the feed

But preparation is invisible, and for Cuban that is a problem with social media. His 17-year-old son, he says, kept showing him videos about getting rich quickly: reselling goods you never hold in stock yourself, limited-edition trainers, stock market tricks. Cuban’s answer: if these were good businesses, nobody would talk about them, they would simply do them. Watch from 0:28:21 He also blames Section 230, a provision of American communications law that largely exempts platforms from liability for their users’ content. The fact that, as he reads it, it also protects the platforms’ recommendation algorithms is holding young people back. Watch from 0:28:53 He cites the late investor Charlie Munger, Warren Buffett’s long-time partner: hardly anyone today can wait for a reward. Watch from 0:30:01

The BubbleShares like trading cards

Bartlett asks whether the stock market is in a bubble. Cuban says yes, but not like in 2000, when the dot-com bubble burst. Watch from 0:31:08 Today, he says, almost every stock has become a speculative stock, from Bitcoin to Google and Apple. He bases this on how violently prices react to small things: if a company misses earnings expectations by a cent or beats them only narrowly, the share falls 18 per cent, even though its value has not changed by 18 per cent. Watch from 0:31:17 “We trade them like baseball cards.” What will burst the bubble, he does not know. Watch from 0:31:55

His advice for investors: be patient, by all means invest a little for fun, but keep cash, because the market will turn at some point. That is when the real money is made, as he says he did in the Covid crash. Watch from 0:30:55 There will still be winners. After the dot-com bubble burst, Amazon made losses for years and is now worth trillions of dollars. Whether the winners this time are called Anthropic or OpenAI, he does not know.

The Machine“AI is not smart”

Then comes the sentence that defines the conversation: “AI is not smart.” It uses statistics to predict the next word. Watch from 0:33:19 Four AI programs give four different answers to the same question, and even a tiny change to the question changes the answer. “That doesn’t work in the real world.” AI does not know the consequences of its actions and has no morals of its own. Watch from 0:33:26 At a busy junction, he would rather be led across the road by a guide dog than by the best AI, because the AI always lags a moment behind. Watch from 0:34:07 A toddler in a high chair knows that its mother will come running if it throws the cup on the floor. “AI will never figure that out.” Watch from 0:34:41

Bartlett objects that the technology is developing at breakneck speed and his Tesla already drives itself. Cuban sees countervailing forces. People will push back, just as they protest against new data centres today. There will be communities with no technology at all. Watch from 0:35:51 And researchers are increasingly holding back their results as trade secrets instead of publishing them, so that no model swallows them up. Every model therefore learns from different data. Watch from 0:36:39 He stresses that AI is not useless, it is “incredible”. But it is like having four of the smartest people next to you: you get four answers and have to ask the right question and decide yourself. Watch from 0:37:13

For Cuban, smart means knowing where to find information, bringing it together and using it to solve a problem, and recognising in the first place which problem needs solving. Watch from 0:46:50 His example: thanks to AI, a doctor has every study at hand. But if the patient throws up in front of him, no AI will help; then the senses and experience are what count. Watch from 0:47:37 What would change his mind? An AI that solves problems with little computing power, like a highly gifted child from a poor country without any help. We are a long way from that, he says. Watch from 0:48:07

Nor is he impressed by the AI attacks on computer systems one hears about. The programs did not invent anything, he says. They take the same list of known vulnerabilities that every security expert has and try every combination with an unlimited number of AI agents. Watch from 0:45:44 But the same computing power can also be used for defence, to find your own gaps first. Watch from 0:46:07 He is no doom-monger: “I’m more afraid of what people will do out of fear and ignorance about AI than of what AI will do.” Watch from 0:46:20

He calls Elon Musk the best entrepreneur of his generation, but a salesman nonetheless, who has to raise a lot of money for his ventures. Watch from 0:49:07 Bartlett says that safety researchers and heads of the labs describe all sorts of different futures to him in his studio. Cuban: if no two experts say the same thing, that tells you enough. Watch from 0:49:49

The AgentsMore done, other people hired

What becomes of work if the AI bosses are right? Cuban recalls Terre Haute in Indiana, where records, CDs and DVDs used to be pressed. That work has disappeared, yet people still make money from music. There will be new kinds of work, but the road there will be painful for many, as it was for the people of Terre Haute. Watch from 0:39:59

Bartlett says 25 AI agents now work on his phone, that is, programs that carry out tasks on their own. Cuban says that thanks to his agents he has read a non-fiction book for the first time in 25 years, because they take the tedious work off his hands. Watch from 0:41:07 In the past he might have brought in an intern at 15 dollars an hour for that. Now there is time for new business ideas, and for those he hires people. Watch from 0:42:04 For Cost Plus Drugs, one agent alerts him to every lawsuit against the big middlemen in the pharmaceutical market and every one of their tenders. Another writes requests under the Texas freedom of information law to obtain their contracts. That would never have been done before. Watch from 0:42:28 He tells his children: anyone who does not learn to work with agents will be left behind by those who can. Watch from 0:43:25

The DebtData centres on credit

Cuban devotes the largest part of his warning to money. He once said that many data centres would eventually become pickleball courts, a mix of tennis and table tennis popular in the US. He meant it as an image: for AI to reach its goal, it has to do more with less power, not more with more power. Watch from 0:52:19 The US could slide into a debt bubble. Watch from 0:52:38 The government is borrowing more and more, and at the same time the big cloud companies are borrowing more than they take in. If interest rates rise and a company has 40 billion dollars of debt, it can no longer get new money. Watch from 0:52:55

The chipmaker Nvidia, he says, is subsidising half the industry with its money, and Meta is borrowing a multiple of its cash flow. Watch from 0:53:22 He refers to the American investor Ray Dalio, who warns that the US has already crossed the threshold into an over-indebted society. It is a race: the ten-year Treasury yields around 5.2 per cent, and at six, eight or ten per cent something will have to give. The US already pays a trillion dollars in interest a year. When the cloud companies compete for the same money, something will be crowded out. What, he cannot say; he does not understand that much about finance. Watch from 0:55:08 He himself has spread his money across several countries and is holding back on new investments. Watch from 0:55:50

What should an ordinary person do? Bartlett suggests: become really good at something and learn to deal with people. Cuban agrees and says he invests in companies where people meet in person, such as silent discos. Watch from 0:56:47

The ManMoney, fame and a late insight

Wealth gave him freedom but did not make him a better person: “If you’re unhappy when you’re broke, you’ll probably be unhappy when you’re rich.” Watch from 1:02:39 He used to shout at his employees until his business partner Todd Wagner told him people were afraid of him. Being friendly got him further. Watch from 1:03:20 Harder than being rich, he says, was being famous. Watch from 1:05:36 After his parents died, he realised how little time is left. That is why he sold the Mavericks and stopped doing Shark Tank, to be with his children. Watch from 1:11:09

He is bad at hiring, he says, because as a salesman he sees the good in every applicant. His companies therefore hire slowly and part ways quickly. Watch from 1:16:02 At the Mavericks he brought back former players who each mentored a current player with similar skills, and, by his own account, with his former statistics professor Wayne Winston he hired the basketball league’s first full-time data analyst. Watch from 1:18:27

HealthMedicines at cost price

What he has set his sights on for the coming years is the American healthcare system. According to him, three big companies dominate the pharmaceutical market there. So he visits the heads of large employers who pay their staff’s healthcare costs themselves and appeals to their conscience. Watch from 1:28:34 Bartlett asks him about a joint appearance with President Trump. Cuban’s answer: anything that makes people healthier and saves them money is fine by him, no matter who it comes from. Watch from 1:29:25

The Fact CheckWhat can be verified

Cuban’s interest figures are right in order of magnitude. According to the US Treasury, the yield on the ten-year US Treasury was between 5.2 and 5.3 per cent in early October 2026, which fits his 5.2 per cent. By the broadcast date of 26 October it had fallen to 5.0 per cent. Net federal interest payments in fiscal year 2026 amounted to 1,143 billion dollars according to the Congressional Budget Office, eleven per cent more than the previous year. The deficit stood at 2.0 trillion dollars.

That Meta is borrowing a multiple of its cash flow is not stated in any document we were able to examine. According to its annual report to the Securities and Exchange Commission, Meta had 43.6 billion dollars left in 2025 after all investments, and its long-term debt stood at 58.7 billion dollars. For 2026, however, Meta plans investments of 115 to 135 billion dollars, considerably more than came in during 2025. That Nvidia is subsidising half the industry is an exaggeration. What is documented is that, according to an analysis by the broadcaster CNBC, Nvidia put more than 40 billion dollars into AI companies in the first four months of 2026, 30 billion of it into OpenAI.

That AI cannot take in video is not accurate as stated. Google explicitly says that its Gemini model can analyse videos. Cuban’s real point is the delay with live images, for instance at a junction. Whether Section 230 also protects the platforms’ recommendation algorithms is disputed in the US. The Supreme Court expressly left the question open in 2023 in the case Gonzalez v. Google.

Cuban bought the Mavericks in 2000. For the sale of the majority stake to the Adelson and Dumont families at the end of 2023, the team was valued at 3.5 to 4 billion dollars according to reports; Cuban puts it at 4.3 billion. He still holds 27 per cent. The joint appearance with Trump took place on 18 May 2026 at the White House. It was announced there that the government ordering site TrumpRx would add more than 600 generic drugs at Cost Plus Drugs prices, according to Bloomberg Government.

Summary and context by kipode.de. The text is our own; statements from the original are timestamped. The video is from The Diary of a CEO and is embedded via the official YouTube player. Not an official translation, no connection to the podcast.