Mark Cuban on the AI Bubble: Who Actually Gets Wiped Out?

This source is an approximately 45-minute interview with Mark Cuban. The interviewer is not formally identified in the transcript, but contextual evidence strongly suggests Jason Calacanis: the interviewer refers to himself in connection with “Jason Calacanis investments,” discusses his venture firm and portfolio companies, and closes with “there’s your 45 minutes with Mark Cuban.” This identification should be verified before publication.

The conversation covers:

• AI investment and whether the current market resembles the dot-com bubble.
• AI infrastructure and data-center spending.
• IPOs, M&A and public stock as acquisition currency.
• Hedging concentrated private-company or technology-company equity.
• Why enterprise AI implementation remains difficult.
• AI's effects on entrepreneurship and employees.
• Lovable, AI coding tools, agents and internal software development.
• World models, video, robotics and future compute demand.
• AI-assisted health care and personal health data.
• Social media algorithms, political polarization and large language models.
• Immigration and global talent.
• U.S. political leadership and polarization.
• Texas versus California and New York.
• Wealth taxes, mobility, housing and founder geography.
• The NBA, the Knicks, Mavericks, Spurs, Oklahoma City and league roster rules.
• Sports-franchise valuations and streaming subscriptions.

Two advertisements interrupt the interview: one for AppLovin and one for Northwest Registered Agent. Their commercial claims are preserved separately below rather than blended into the speakers' editorial discussion.

The interviewer opened by asking whether the current technology market is a bubble and noted that both participants had lived through previous investment bubbles.

Cuban distinguished the present situation from the late-1990s dot-com bubble.

He recalled that during the dot-com period:

• Companies with little or no revenue, traffic or operating substance were going public at extreme valuations.
• Their shares could rise 50% or 100%.
• Speculation became a mass-market phenomenon; Cuban recalled ordinary people, including taxi drivers, discussing speculative stocks.

Cuban argued that the current AI boom does not look like that type of broad retail-market bubble.

His concern was instead concentrated in private capital:

• Venture-capital firms.
• Private funds.
• Private-equity firms.
• Investors heavily concentrated in fashionable private AI companies.

His view was that a reversal therefore might not financially damage the average American in the way a public-market mania could, but could seriously damage professional investment firms that have gone “all in.”

The interviewer reinforced the point from his roughly decade-long venture experience. He said he had watched investors who deployed capital at the wrong point in the cycle effectively disappear because they invested near a peak.

A central lesson was that entry price matters.

The speakers contrasted earlier angel investing, when they could invest in startups valued around $5 million or $10 million, with later periods when unlaunched products were asking for valuations of $40 million, $50 million or $60 million.

The interviewer questioned how those economics could work when the product had not even launched.

The speakers then connected AI valuations with the enormous infrastructure spending by major technology companies.

The transcript says companies including Google and Meta are borrowing “hundreds of millions of dollars.” The precise figure should be verified because the wording may be mistranscribed.

They discussed:

• Existing problems in private credit.
• Large technology businesses generating substantial cash flow.
• Those companies spending much of that cash flow on capital expenditure.
• Additional borrowing being layered on top of that expenditure.

A reference to “Al Capital, getting all the refunds” is [transcript unclear].

The discussion included 50-year bonds, which were characterized as financial planning based on an unusually optimistic long-term outcome.

Cuban described this as effectively planning or pricing for perfection.

The risk is that the industry is committing enormous sums to infrastructure whose long-term economics depend on assumptions that cannot reliably be forecast 10 or 20 years ahead.

Cuban questioned whether today's enormous AI data-center buildout could eventually become excessive.

His reasoning was not that AI usage will disappear. He acknowledged that utilization is likely to grow significantly.

His concern was technological improvement.

If AI experiences major improvements in price-performance efficiency, especially improvements that reduce power requirements, today's assumptions about how much physical infrastructure is required could prove wrong.

Cuban joked that under such a scenario:

“there's going to be a lot of data centers that are going to be turned into pickleball courts.”

The conversation compared the situation with the telecommunications fiber buildout.

The speakers recalled that:

• There was once enormous urgency around installing fiber.
• Fiber capacity then improved dramatically.
• The perceived bandwidth shortage disappeared.
• The market was eventually left with significant “dark fiber.”
• Infrastructure that had been expensive to build could later be acquired extremely cheaply.

Their question was why AI infrastructure should be assumed immune from similar price-performance improvements.

Cuban therefore cautioned against projecting today's compute requirements linearly decades into the future.

The interviewer additionally suggested that public or political resistance to data centers might paradoxically limit excessive construction and therefore “protect” investors from some overbuilding.

Both agreed that making commitments involving tens of billions of dollars over 10- or 20-year horizons requires forecasting beyond what anyone can reliably know.

The speakers applied the “pricing to perfection” idea to companies such as OpenAI.

The transcript uses an example involving putting $100 billion to work at OpenAI.

Cuban's point was that enormous amounts of capital cannot merely generate impressive revenue.

For the investment thesis to work, the spending ultimately has to generate:

• Earnings.
• Margin dollars.
• Profitability.

The speakers agreed that no one yet knows with certainty whether those economics will materialize at the scale implied by current investment.

Cuban argued that an important consequence of the current private-capital-heavy environment is that more companies should conduct IPOs.

He was not referring only to giant companies such as:

• SpaceX.
• OpenAI.
• Anthropic.

He specifically advocated smaller public offerings, giving examples of approximately $50 million or $100 million IPOs.

His reasoning concerned acquisition currency.

If AI creates the degree of industry disruption the speakers expect, AI companies will encounter opportunities to acquire:

• Struggling legacy competitors.
• Businesses with valuable domain expertise.
• Companies owning useful proprietary data.
• Other strategically important assets.

A public company's stock can serve as currency for those acquisitions.

A private company without publicly tradable equity may repeatedly need to raise cash before making acquisitions.

That can become dangerous because:

• Capital can become expensive.
• Capital markets can close.
• A downturn can make fundraising difficult at exactly the time acquisition opportunities become attractive.

A performing public company, by contrast, can potentially:

• Issue stock.
• Conduct secondary offerings.
• Use equity directly in acquisitions.

The interviewer called the idea unusually forward-looking.

He said he had been encouraging portfolio companies to go public, using the emphatic phrase:

“go public, motherfuckers, go public.”

He said many founders nevertheless do not view becoming public as the right course.

The conversation moved into U.S. antitrust policy.

Cuban said entrepreneurs have to “play the game on the field” and argued that large-scale M&A had effectively been unavailable for several years.

He attributed this partly to the Federal Trade Commission approach associated with Lina Khan.

He characterized that approach with a Minority Report analogy: regulators behaving like “precogs” trying to predict which companies might become monopolies in the future and stopping transactions preemptively.

Cuban criticized that approach for, in his view, failing to give sufficient consideration to:

• International competition.
• Competition with China.
• Other global strategic pressures.

He described approximately four years of very limited acquisitions.

He said people he knew in corporate-development departments had effectively been told to stand down because the risk and cost of regulatory challenges and breakup fees made transactions unattractive.

Even if the policy question around enormous mergers is treated separately, Cuban argued that AI disruption will also produce many valuable smaller acquisition targets.

His core argument remained that AI companies should have acquisition currency ready rather than constantly needing to raise cash.

The interviewer connected Cuban's point to Broadcast.com.

Cuban recalled that Broadcast.com acquired approximately five companies using stock rather than cash.

The discussion framed this as evidence of the strategic flexibility equity can provide a growing company.

Eventually, a still larger buyer appeared: Yahoo.

The speakers indicated they were happy to become the acquired company when that opportunity arrived.

The conversation shifted from company financing to employees whose personal wealth is highly concentrated in valuable private or technology-company shares.

The interviewer asked whether employees at companies such as:

• Anthropic.
• OpenAI.
• SpaceX.

should consider protecting some of that wealth.

Cuban said that if an equity position has already become life-changing, he would seek a collar or similar hedge.

His philosophy was that an employee does not necessarily need to maximize every final dollar of upside if protecting against catastrophic downside can already secure a transformed financial life.

He summarized his own motivation as:

“I collared my stock because how rich do I need to be?”

His suggested approach was to preserve some upside while protecting enough downside to secure the wealth already accumulated.

Cuban explained that when he wanted to protect his Yahoo holdings, the financial product he needed was not readily available.

He said he therefore arranged for Goldman Sachs to construct an index of internet stocks that he believed were poor investments.

He then shorted that index.

He said the hedge had to remain in place until he was legally and practically able to establish a conventional market collar on his Yahoo stock.

Cuban stated that he lost tens of millions of dollars on the short position but ultimately made up for those losses through the protection and value of the broader strategy.

The interviewer described it as one of the great trades.

The structure, amount, timing and legal details should be independently verified before publication.

Asked what AI capabilities would prove most consequential over the next five to ten years, Cuban said the analysis should begin by “working backwards.”

His central claim was:

“AI is a lot harder to implement than anybody expected.”

He distinguished between easy individual use and difficult institutional deployment.

Simple AI tasks can include:

• Prompting.
• Basic agents.
• Search-like assistance.
• Projects.
• School or workplace shortcuts.
• Very large personal productivity improvements.

Cuban jokingly referred to people using AI to cheat on tests or work and suggested some tasks could see “100x” productivity improvements.

Those successes encouraged a mistaken assumption that enterprise adoption would be equally straightforward.

His view was that it is not.

Enterprise implementation, especially for mission-critical work, remains difficult.

Cuban referred to predictions associated with Dario [surname not stated in transcript] and others that approximately 50% of white-collar workers could lose their jobs, supposedly on a short timeframe.

He said that roughly two years later:

• Employment was still growing.
• Companies were still hiring.
• Businesses actually needed more people who were AI literate.

He therefore rejected the claim that AI, in its current state, is about to remove half of white-collar employment.

His reasoning was not that AI is unimportant.

He repeatedly described AI as extraordinary and potentially the most consequential technology ever created.

His narrower point was that existing systems still fail at many ordinary tasks people need them to perform reliably.

Cuban used the rise of forward-deployed engineers as evidence against exaggerated claims about AI autonomy.

His logic was:

• If AI were already as capable as some AGI rhetoric implies, a user should be able to tell the AI what outcome is required.
• The AI should then be able to determine how to implement that outcome.
• Yet leading AI companies are sending substantial numbers of engineers into customers to make systems work.

The transcript cites:

• Microsoft hiring 6,000 people.
• Anthropic deploying people into companies.
• OpenAI deploying people into companies.

Cuban argued that this demonstrates the difficulty of real implementation.

The transcript also refers to “Alex Carr from Palantir” criticizing companies for giving away their “alpha.”

Cuban interpreted that criticism as essentially another company adopting Palantir's own forward-engineering approach.

The name “Alex Carr” should be verified against the audio.

Cuban offered a practical test.

He suggested asking Claude or ChatGPT something equivalent to:

• Search for Jason Calacanis's investments.
• Produce a report.
• Email that report every week.

He argued that an ordinary user still cannot simply make that request and have the full workflow reliably created and maintained without technical intervention.

Instead, the system may:

• Suggest an agent.
• Generate JSON.
• Generate code.
• Require programming knowledge.
• Produce poor initial output.
• Require repeated iteration.

The interviewer agreed that even relatively straightforward workflows still require some “systems thinking.”

Cuban contrasted this with easy conversational or image tasks, such as asking an AI to assess a picture or suggest what someone might say to another person.

His distinction was between impressive isolated responses and reliable multi-step operational systems.

Cuban stressed that these shortcomings are not merely negative.

They create, in his view, an enormous entrepreneurial opportunity.

People who understand AI plus basic computing can offer practical implementation assistance to:

• Small businesses.
• Medium-sized businesses.
• Large enterprises.

The opportunity is to fix the repeated failures companies encounter when they try to put AI into operational workflows.

Once those integration problems are solved, Cuban said what AI can accomplish is “phenomenal.”

Cuban disclosed that he is an investor in Lovable.

He said “Anton” had reported that users were creating approximately 770,000 applications per week using Lovable.

Additional claims in the transcript were:

• Only about 30% of Lovable's business was in the United States.
• Only about 20% of its users/business came from engineers [precise denominator unclear].
• Significant use was occurring in places including Brazil and India.

Cuban connected those figures to his belief that there has never been a better time to become an entrepreneur.

AI reduces the importance of geography because people in:

• Brazil.
• India.
• France.
• The United States.
• Other countries.

can now perform startup work that once required far more capital, staff and time.

To test startup creation with AI, Cuban said he invented a fictional company.

His prompt described a business involving:

• A button capable of recording 24-hour video.
• A mechanism to send the recorded material to him every 24 hours.

He then asked AI to produce material including:

• A patent.
• A business plan.
• Licensing requirements or related regulatory needs.

He said this initial work took approximately 12 minutes.

The interviewer contrasted that with historical startup timelines of roughly:

• Six months to build a prototype.
• Twelve months to launch.

Cuban then asked AI for:

• A bill of materials.
• Potential source companies or suppliers.

He emphasized that these are basic but essential pieces of early company formation that historically consumed significant founder effort.

His point was that AI can now produce a usable first pass almost immediately.

Cuban dismissed the objection that AI-generated startup documents might be wrong.

His reasoning was that traditional business plans are also wrong because no startup unfolds exactly as forecast.

He said:

“every single mother fucking business plan ever written in the history of business plans is wrong.”

The interviewer agreed and described a business plan as:

• A thought exercise.
• A guideline.
• A template.
• A device for establishing an initial direction.

Because founders will learn and iterate anyway, Cuban argued that an imperfect AI-generated starting point can still be extremely useful.

The relevant comparison is not perfect AI output versus perfect human foresight.

It is fast, cheap initial direction versus the slower and more expensive process previously required to produce another inherently uncertain plan.

Cuban argued that truly advanced AI should be better at learning from the recurring errors experienced by millions of users.

He offered a hypothetical example.

If a user's AI project has failed three times, an advanced system should recognize the pattern and say, in effect:

• Other users encountered the same problem.
• A very high percentage solved it in a particular way.
• Here is the likely fix.

He used the hypothetical figure 97.6%.

His complaint was that present systems often merely report that something failed instead of proactively supplying the accumulated remedy.

He regarded that as evidence of how far AI still has to go.

Despite his criticism, Cuban repeatedly emphasized that he is strongly positive on AI.

He said:

• He loves the technology.
• He invests in it.
• It is going to be “amazing.”
• It may be the most impactful technology humans have seen.

He described AI as especially powerful for programmers.

He also said it can feel like “magic” in narrow, structured domains such as:

• Code.
• Legal work.
• Tax.
• Mathematics.
• Basic structured data.

His skepticism applies principally to the assumption that broad autonomous capability has already arrived.

For more advanced ordinary use, Cuban compared AI literacy with earlier generations of business software.

Companies once routinely had:

• An Excel expert.
• A PowerPoint expert.
• Someone who knew how to make office software perform tasks ordinary employees could not.

Entire businesses were built around supplying templates and specialized support for those tools.

Cuban mentioned investing in a company associated with downloadable PowerPoint templates [company reference/transcription unclear].

His argument was that AI has not eliminated this expert layer.

Instead, it may create a new version of it.

People who understand how to make AI work reliably inside businesses can build careers and companies around that capability.

Cuban warned that even a successfully built AI system may later break.

He described agents as getting “bored” and “drifting.”

His explanation was that the underlying large language model changes over time.

An agent initially designed around one model behavior can become unreliable when:

• The model is updated.
• Its responses change.
• The assumptions embedded in the original implementation no longer match the model.

Therefore AI deployment can create continuing maintenance work rather than permanently removing the need for people.

The interviewer connected this with previous observations that more personnel can ultimately be required to manage complex AI systems.

The interviewer described trying to make his entire firm AI-first.

He observed a striking division even among younger employees.

One group embraced AI and solved roughly five or six pressing problems.

Another group did not embrace it.

He described the productivity and capability gap as enormous.

His analogy was to the early PC era:

• One employee knew how to use a PC and Microsoft Office.
• Another was still working on legal pads.

He said the current difference between AI-literate and non-AI-literate workers can feel similarly dramatic.

The interviewer said AI-literate employees were repeatedly moving between tools as they encountered limitations.

The transcript references:

• “OpenClaw” / “OpenClaught.”
• “Irmez agent” [transcript unclear].
• Claude Co-Work.
• Perplexity Computer.
• Lovable.

Employees initially built with OpenClaw, but the system became brittle.

He described agents:

• Breaking.
• Hallucinating.
• Becoming frustrating to maintain.

The team then used Claude Co-Work, which he described as useful but somewhat limited.

They subsequently moved toward Lovable and began building increasingly substantial internal software.

The exact product names and sequence should be checked against the audio.

The interviewer said he told employees, in effect, to “token max it.”

He allowed individual employees to spend a couple thousand dollars per month on AI usage.

He said he did not care much about the token expense; he cared about the resulting gains. The transcript renders one word as “games,” which may be [transcript unclear].

His point was that a few thousand dollars in AI-compute costs are trivial if the systems eliminate software projects that would otherwise cost hundreds of thousands or millions.

The interviewer gave his venture firm as an example.

Employees used AI tools to build intranets and internal systems that a venture firm traditionally would not commission.

He estimated that a conventional intranet might previously have required approximately:

• $500,000.
• 12 months.

Because that cost and delay would have been unreasonable, the firm would instead have used:

• Notion.
• Google Sheets.
• Existing SaaS tools.

The team would have attempted to force the firm's workflow into off-the-shelf software and likely abandoned parts of the project when customization became too difficult.

Now, he said, two or three employees are building software that approximately five years earlier might have required $2 million to $3 million per year in outsourced development.

His larger conclusion was that AI is creating software that economically would not have existed before.

The speakers discussed repeated predictions that Lovable would be displaced or absorbed by frontier-model companies.

The interviewer congratulated Cuban on the investment.

Cuban said he had told “Anton” that every time people predicted the company would go out of business, it seemed to add another $100 million in revenue.

The transcript then includes overlapping references to:

• $600 million.
• $200 million.
• $300 million.
• Four separate predictions of failure.

The exact revenue figures and chronology are unclear and require verification from the original audio.

Cuban mentioned several investments.

He named:

• Lovable.
• Synthesia.io.
• “AMI,” described as connected to “Jan LeCoon's world model” [transcript spelling unclear].

Cuban said he was an early or first investor in Synthesia, perhaps approximately 10 years earlier.

He characterized the company as performing extremely well.

The speakers then contrasted:

• Large language models.
• Transformers.
• World models.

This led into a discussion of AI systems that understand physical environments rather than relying principally on text and images.

Cuban argued that the current AI paradigm is too heavily based on text and static images.

He said:

“Nothing's going to fucking be text and pictures in 10 years.”

His broader point was that intelligence operating in the real world needs models of:

• Physical causality.
• Motion.
• Objects.
• Human behavior.
• Spatial environments.
• Time.
• Consequences of actions.

He therefore sees world models and video as central to future AI.

Cuban used a two-year-old child as an example of knowledge current AI lacks.

If a toddler is in a high chair with a sippy cup, the toddler understands the likely consequences of pushing the cup over the edge:

• The cup falls.
• A parent may come running.
• The child may find the reaction amusing.

Cuban argued that an AI observing the video does not necessarily possess this intuitive causal understanding.

For him, that gap illustrates how far AI remains from basic embodied understanding that even a very young child can develop naturally.

Cuban offered another thought experiment.

If a blindfolded person stood at a street corner and had to cross the road, would that person rather trust:

• A phone equipped with current AI.
• A trained seeing-eye dog.

Both speakers chose the dog.

Cuban used this to emphasize:

“It's not ready.”

The example supported his argument that impressive language ability should not be confused with robust real-world intelligence.

The interviewer suggested that enormous amounts of YouTube video provide a representation of much of the world.

Cuban responded that this is insufficient.

The conversation referred to collecting richer physical-world training data, including:

• People wearing gloves.
• People in Manila performing recipes or physical tasks.

The precise meaning and companies involved in this portion are unclear and should be checked.

The essential argument is that world models need data containing physical interaction and environmental consequences, not merely passive visual recordings.

Cuban said he invested in a company identified in the transcript as Matter.com.

He described the company as launching satellites and using a technology he associated with “spectrography.”

According to Cuban's description, the satellites collect data or video about the Earth beneath them and use spectral information to help generate representations useful for world models.

The transcript contains the phrase “spectrapeum,” which is [transcript unclear].

Cuban said he did not fully understand all the technical details himself.

His essential investment thesis was that the resulting data or algorithms could feed other world models.

All company names and technical claims should be verified.

The discussion returned to the earlier skepticism about data-center overbuilding.

Cuban identified a major reason he could be wrong.

Video and world models require dramatically more computation and token-like processing than text.

Future demand could therefore expand enormously through:

• Video.
• World models.
• Robotics.
• Rich physical-world inference.

Cuban said:

“If I'm going to be wrong on the data centers, it's going to be because of video.”

Thus his position was deliberately qualified.

He believes current infrastructure spending may be vulnerable to efficiency breakthroughs, but admits that entirely new workloads could consume far more computation than today's text-centric AI.

Robotics was described as closely connected to world models, although Cuban said robotics still has “a long way to go.”

The interviewer complimented Cuban's appearance and said he looked better than he had 20 years earlier.

Cuban credited eating better.

The conversation also contains references rendered as “Reda True Tie,” “T” and “TAM,” which are [transcript unclear] and should not be interpreted without checking the audio.

Both men joked about aging and nighttime urination.

Cuban said he invested in OpenEvidence, which the speakers described as a medical-focused AI product.

He also said he personally uses it.

Cuban described taking:

• A supplement because an unspecified level was low; the transcript renders this as “eye level.”
• A medication.

He had been taking them together and was experiencing problems.

He entered information into OpenEvidence describing:

• Everything he was taking.
• Everything he was eating.

According to Cuban, the system identified a timing issue and advised taking the medication separately, including a suggestion to take it when he woke around 3 a.m. to urinate.

The medical substance, supplement, condition and accuracy of the advice are not clear from the transcript and require verification before publication.

Cuban strongly predicted that AI will improve health care.

However, he explicitly rejected the idea that it will replace physicians.

His position was:

“for health, it's going to make a huge difference, but it's not going to replace doctors.”

The speakers instead described AI as making both patients and doctors better informed.

The speakers discussed:

• Apple Watch.
• Whoop [rendered “Woop” in transcript].
• Regular blood tests.
• AI-assisted self-directed health care.

The interviewer argued that combining wearable data with laboratory testing could produce early health wins before a patient even visits a physician.

Cuban said he has been getting blood tests regularly for approximately 10 years.

The transcript indicates a current cadence of roughly every three to six months.

He values this because it creates a personal longitudinal benchmark rather than a one-time snapshot.

Cuban said he publicly discussed this idea on Twitter/X roughly eight to ten years earlier, including debates with doctors about the value of personal trend data.

The interviewer compared the emerging health-AI period with the earlier “big data” era.

Previously, people and companies accumulated data without sufficient intelligence to interpret it.

Now AI can potentially integrate:

• Sleep.
• Steps.
• Heart rate.
• Respiratory information.
• EKG information.
• Blood panels.
• Diet.

The speakers also referred to Apple conducting health studies through Apple Watch.

Their expectation was that longitudinal multimodal data will help patients understand themselves better and give doctors richer context.

Cuban said that “95% of medicine is guessing.”

This is a broad opinion/factual assertion and should not be presented as an established statistic without verification.

His reasoning was that no individual doctor can memorize every new medical development occurring every day.

AI can therefore improve medical practice by bringing current information and patient data into the consultation.

However, Cuban emphasized several capabilities physicians still provide:

• Empathy.
• Communication.
• Visual observation.
• Real-world situational judgment.

He used an intentionally obvious example: a person asking why they are bleeding when a doctor can visually recognize an apparent gunshot wound.

His point was that current AI lacks forms of direct perception and clinical judgment that remain essential.

During a transition into politics and economics, the speakers joked about Texas heat and air conditioning.

One speaker claimed that a record number of air conditioners were coming from China to Europe that summer.

He said he had researched the companies installing them, apparently with an investor or trading mindset.

The other speaker joked:

“you always got to trade.”

The import-volume claim and timing should be verified.

The interviewer described the U.S. as experiencing an unusually intense ideological divide between socialism and capitalism.

He suggested the political atmosphere was among the most polarized of their lifetimes, possibly comparable to the 1960s or 1970s.

There was a joke about their ages, including Cuban saying he looked good for being born in “1923.”

This was clearly humorous rather than a serious biographical claim.

Asked about socialist politics, Cuban divided the issue into two parts.

First, he said much of the DSA-related political movement is local.

He used New York and “Mandami” [name as transcribed; verify] as an example.

His argument was that local politicians ultimately have to govern within concrete constraints:

• They control a budget.
• They must implement policies.
• They have to balance the budget.

He expected similar political movements to appear in places including Michigan.

He did not expect the trend by itself to be nationally “game-changing.”

Cuban's second and more important point concerned social-media algorithms.

He argued that the politicians gaining traction are particularly effective at social media.

His claim was:

“the algorithms drive how people vote in the United States.”

He qualified that he was speaking about the U.S. and could not make the same claim with confidence about the rest of the world.

He went further:

“Whoever controls the algorithm controls, you know, the election in many cases.”

Cuban cited Donald Trump as someone who understands how to use attention and algorithms effectively.

He also described “Mandami” as someone whose entire adult life occurred in the era of Trump and social media and who therefore had an opportunity to study both.

He said an unnamed Michigan political figure represented something similar.

The speakers described successful online political strategy as partly a matter of commanding attention.

Examples included:

• “Flooding the zone.”
• Trump's “they're eating cats and dogs” statement.

Their argument was not necessarily that such statements are true.

It was that attention itself changes algorithmic exposure.

If a political figure can make people:

• Search for a claim.
• Click on it.
• Demonstrate interest.

the platforms will tend to show those users more related content.

This can create separate information environments.

A supporter may receive positive content about a politician continuously.

An opponent may receive continuous negative or counter-content.

Either way, the politician dominates attention.

Cuban then offered a much more optimistic view of large language models.

He said LLMs may become one of the most important technologies for reducing information asymmetry in politics.

His theory depends on a difference in business or trust incentives.

He argued that an LLM company needs users to believe it is:

• Accurate.
• Literal.
• Honest.
• “Truth seeking.”

If users conclude that an AI system simply lies about political facts, the product loses trust.

By contrast, Cuban characterized social-media platforms as principally rewarded for keeping people engaged.

He summarized the distinction as:

“their currency is getting you the correct answer and the correct knowledge, social media's currency is keeping you engaged.”

For Cuban, these are fundamentally different missions.

Cuban predicted that as people become more uncertain about politics, some will increasingly ask large language models questions such as:

• Who should I vote for?
• What policies fit my priorities?
• What constitutes a reasonable immigration policy?

He imagined the model responding by asking users about:

• Their interests.
• Their values.
• Their policy priorities.

He expected the system then to provide what he described as honest or reasonable answers.

As an immigration example, the speakers discussed possibilities including:

• Tighter border control.
• A points-based legal immigration system resembling systems associated with Canada or New Zealand.

These were examples in the conversation, not verified policy prescriptions.

The interviewer said one of his current online habits is replying to political figures and asking “GROC” [likely product name requires verification] to vet their claims.

He mentioned figures including:

• “JD.”
• AOC.
• Stephen Miller.

His preferred instruction to the AI was essentially:

• Vet the claims.
• Be objective.
• Be factual.
• Check the work.
• Show the sources.

He described this partly as “trolling” politicians by having an AI publicly examine their assertions.

He said he had particularly been doing this to Stephen Miller.

The interviewer described himself as strongly supportive of:

• Legal immigration.
• Recruiting talented people internationally.

He argued that the United States needs exceptional people.

The speakers used professional basketball as an analogy.

Obtaining an elite player means another team cannot have that player.

The transcript garbles the examples as “Dirk and a whiskey” and “Joe and Brunson,” which should be checked against the audio.

The intended reasoning is clear: nations compete for talent much as sports teams do.

The speakers connected this to why they attend events in Europe: talented and intelligent people are distributed globally.

They referred to differences across:

• European countries.
• India.
• The wider world.

One speaker said he believed some entrepreneurs were beginning to leave the United States for other places.

Both regarded that possibility as concerning.

Cuban hoped truth-seeking LLMs might counter misinformation or algorithmic political narratives that discourage beneficial immigration.

Asked whether he remained optimistic about America, Cuban said every country has problems and joked that America's current problems happen to be unusually “insane.”

He nevertheless cited presidential term limits as one stabilizing feature.

The conversation then turned to speculation about Donald Trump potentially seeking a third presidential term.

The interviewer asked whether Cuban would run for president.

Cuban replied:

“if he decides to run for a third term, then I'll run. But he's not.”

The interviewer jokingly called the statement “breaking news.”

This exchange should be presented in its conversational context rather than as a formal campaign announcement.

The interviewer began suggesting that political leadership might benefit from executives who are already financially secure and therefore motivated by public service or civic pride.

Cuban declined to fully pursue that argument.

He then criticized both major U.S. parties.

His characterization was:

Democrats:

• “don't know how to actually do anything.”
• Take events and extrapolate them toward the end of democracy or catastrophic outcomes.

Republicans:

• “do crazy shit.”
• Lack connection with ordinary people.
• Lack empathy.

These are Cuban's political opinions, not neutral factual conclusions.

Despite the criticism, he predicted that after:

• The next midterm election.
• The next presidential election.

the country would begin returning to “normalcy.”

His reasoning was that most people ultimately want normal political conditions.

Cuban discussed lessons from living in:

• New York.
• California.
• Texas.

He claimed Texas spends roughly half as much per person as New York, sometimes less.

The transcript gives approximate numbers of:

• $6,000 per person in Texas.
• Roughly $12,000–$14,000 per person in New York.
• A ratio of approximately 2.2 times.

Cuban said he considers quality of life better in Texas despite the lower government spending.

However, he explicitly added a counterpoint:

• New York contributes more to the federal treasury than Texas.

He said one might initially expect the opposite given the states' differing policy reputations.

His conclusion was that there are genuine trade-offs rather than a perfectly one-sided comparison.

All spending and federal-contribution figures require verification.

The interviewer strongly criticized wealth-tax proposals.

The conversation referred to a wealth tax allegedly tried in France, with a claim that wealthy people left and that the policy operated for approximately two years. The speaker himself expressed some uncertainty, saying “I don't know.”

Cuban then recalled investigating the modeling behind Elizabeth Warren's proposed wealth tax.

He said he contacted an economist at UC Berkeley associated with the model.

According to Cuban:

• The model covered only one year.
• It did not include behavioral analysis examining how taxpayers might change their conduct in response to the tax.

Cuban considered that a major omission.

His broader argument was that policies proposed by figures including:

• “Madami” [transcript spelling].
• Elizabeth Warren.
• California policymakers.

can become political showmanship if they ignore how mobile people and capital actually respond.

The economists, model, timeline and policy details should all be independently checked.

A central theme in the tax discussion was mobility.

Cuban said policymakers sometimes forget that people can move.

The speakers cited prominent people who had relocated to or become associated with Texas, including:

• Travis [surname unstated].
• Sacks.
• Cuban.
• Elon [surname unstated].

The interviewer joked that he had arrived first.

Their broader point was that jurisdictional policy can alter where:

• Founders live.
• Wealth is located.
• Companies are built.
• Talent congregates.

The interviewer contrasted California's development restrictions with Texas.

He said, in broad terms, that California “won't let you build anything.”

He then described buying a ranch in Texas and asking whether he could install a solar farm.

The response he recalled was essentially:

• It is your ranch.
• You can do that.

The anecdote represented his perception that Texas permits greater property and development flexibility.

Cuban or the interviewer added an important qualification that Texas cities can still have substantial restrictions.

Austin was specifically mentioned.

The distinction was between urban regulatory environments and areas outside major cities.

The interviewer claimed that housing prices and rents had fallen for three consecutive years in the relevant Texas market.

He argued that founders moving to Austin can experience a dramatic reduction in personal financial pressure.

Employees and founders may be able to:

• Afford housing.
• Buy homes.
• Build companies without Silicon Valley's cost structure.

He suggested similar dynamics may exist in:

• Dallas.
• Houston.
• San Antonio.

The housing figures and exact geography should be verified.

The interviewer did preserve a benefit for Silicon Valley.

For a first-time founder, he suggested there may still be no better place to become immersed in startup culture.

However, the speakers argued that the environment also has major distortions.

The interviewer said that surviving in Silicon Valley can shape founders' thinking in unhealthy ways.

For a second-time founder, he argued it may make more sense to build in:

• Texas.
• Nevada.
• Florida.

He believes strong talent will now follow good companies to those places.

They contrasted cultural priorities:

Texas:

• Build the company.
• Execute.
• Keep moving.

Silicon Valley:

• Excessive attention to whether a startup has raised a Series A.
• Fundraising status.
• Social signals surrounding venture capital.

The speakers regarded those signals as distractions from building the actual business.

Cuban said he had known Michael Dell since they were approximately 22 years old.

He said he had done business with Dell many years earlier.

The anecdote came during the discussion of Texas's longstanding technology and entrepreneurial culture.

The conversation moved into professional basketball.

The interviewer said he had waited roughly 25 years for the Knicks to win a championship.

Cuban teased him that the Spurs should have beaten the Knicks.

Asked why they did not, Cuban attributed the loss partly to:

• Coaching.
• Lack of experience.
• Late-game mistakes.

Specific examples in the transcript included:

• Wemby throwing the ball into the back of another player.
• Missed free throws.
• De'Aaron Fox going the wrong way on a play.

Cuban emphasized that San Antonio had been ahead in the fourth quarter of multiple games.

The speakers also discussed:

• Mike Brown.
• Concern surrounding his hiring.
• “Tibbs.”
• “J.B.”
• OG.

Several roles and names in this segment are context-dependent and should be checked carefully against the original recording and actual season chronology.

The speakers discussed Jalen Brunson's departure from the Mavericks.

Cuban said Brunson wanted his own team.

He said he could produce a message from Brunson's agent saying:

“he wants his own team. He wants his own team.”

Cuban therefore portrayed the departure as something Brunson actively wanted rather than merely a transaction Dallas accidentally allowed to happen.

The interviewer recalled previously asking Cuban how good Brunson was.

He said Cuban had described Brunson as:

• Really good.
• “The real deal.”
• A great person.

Cuban emphasized Brunson's character, calling him a genuinely good human being.

The agent-message claim and contractual history should be verified before publication.

The interviewer said he attended the fifth game of the relevant Knicks series sitting approximately six rows behind the bench.

He said he cried because he had waited so many years for the championship.

He compared the moment with Cuban's championship experience with Dirk.

The speakers reflected more broadly on how deeply sports affect people's lives.

Cuban said he had attended four World Cup games in Dallas.

He repeatedly said he hates soccer/football.

Despite that, he praised:

• The atmosphere.
• The crowd.
• The energy.
• The overall event experience.

He compared the emotional appeal to attending:

• A Knicks game.
• A Mavericks game.

His point was that major live sports create a communal experience that can transcend whether someone normally follows the sport.

The tournament/date details should be verified.

Asked whether the NBA had peaked, Cuban said not from a fan-interest perspective.

He cited international growth through social media.

As an example, he recalled walking in Paris and seeing a basketball store carrying jerseys for players including:

• Wemby.
• Gobert.

For Cuban, that was evidence of the NBA's continuing global expansion.

The speakers agreed that Wemby had been “humbled.”

Cuban thought that could be beneficial.

He compared the situation with Dirk in 2006, saying Dirk was humbled and returned better.

Cuban predicted Wemby would similarly return:

• Bigger.
• Stronger.
• Better.

The interviewer observed that Wemby's public role had shifted from almost magical new star to perceived “villain.”

The speakers attributed that partly to:

• Playing New York.
• New York's championship drought.
• The emotional circumstances of the series.

Cuban brought up his connection with Indiana University.

He described its football program as an extreme historical underdog.

The transcript states that:

• Indiana had not won 10 games [exact historical meaning unclear].
• In the previous season, it beat more top-10 teams than it had in more than 100 years of the program.

Cuban used the example to illustrate how an unexpected underdog run changes public attitudes and emotional investment.

The statistics should be verified.

The interviewer asked about league rule changes, particularly the NBA's salary-cap “apron” system.

Cuban said the second apron is a major structural change.

His reasoning was:

• Teams cannot sustainably carry too many maximum-salary players.
• At some point, a successful team may have to break up its roster.
• Oklahoma City was cited as a future example.
• Draft picks become especially valuable because teams need lower-cost talent.
• If one expensive player becomes injured or underperforms, roster construction becomes extremely difficult.

Cuban said teams now have to be:

• Better at roster construction.
• More disciplined financially.
• Lucky.

Cuban used young stars on rookie-scale deals as examples of valuable roster economics.

He cited:

• Wemby with San Antonio.
• An Oklahoma City player whose name is rendered as “no hat-jet” [transcript unclear].

The speakers then noted that one referenced player was no longer on a rookie contract, increasing the need for financial care.

The larger point was that superstar-caliber players on inexpensive early-career contracts create a temporary team-building advantage.

Asked whether the league could still produce repeat champions, Cuban distinguished between:

• Back-to-back championships.
• Three consecutive championships.

He said back-to-backs remain possible.

He was far more skeptical about a “three-peat” under the new roster rules.

The salary structure makes keeping a championship core together much harder.

Cuban discussed an Oklahoma City injury involving a player rendered as “Jay Williams, J-Dub.”

He said that if the player had not been hurt, he believed Oklahoma City would have beaten:

• San Antonio.
• A team transcribed as “the Nets.”

He also said he had been worried about Oklahoma City against the Knicks.

Because the teams and playoff sequence as transcribed may not be internally consistent, this entire passage should be checked against the audio before being quoted as a factual prediction.

The interviewer said he knew the Knicks could beat the Spurs.

Cuban challenged that confidence.

The interviewer responded that the Knicks had beaten them two out of three times during the year and had beaten them decisively.

He said he had predicted Knicks in six, while the actual result was “Knicks in five” according to the transcript.

Cuban teased the Knicks for barely beating the Hawks.

The interviewer replied that the team had been adjusting its offense and trying to get “KAT” to buy into the system or scoring approach [end of sentence unclear].

This section is conversational sports banter and contains multiple factual details requiring season-specific verification.

The interviewer asked about a championship ring.

He said he had not yet received it but expected to receive it at the ring ceremony.

The precise speaker and team relationship should be verified from the source context.

The interviewer said he had long wanted to buy a piece of the Knicks.

He said that despite his own net worth having increased, team valuations had risen so far that they seemed “disconnected from reality.”

Cuban explained that modern franchise valuations are increasingly not determined primarily by:

• Game attendance.
• Wins.
• Losses.

Instead, he emphasized media economics.

Cuban said the crucial metric is increasingly the number of subscriptions that sports content generates for streaming platforms.

He mentioned:

• Peacock.
• ESPN.

Traditional ratings still matter because they affect advertising sales.

But Cuban argued that subscription acquisition and retention may be even more important to platform economics and therefore franchise values.

The interviewer said watching the Knicks had caused him to subscribe to one or two additional services.

Cuban immediately asked whether he kept those subscriptions.

The interviewer said he would.

That illustrated Cuban's point: a sports property can be valuable because it creates persistent paying subscribers.

Cuban ended the franchise-valuation discussion with a qualification.

If consumers cancel the streaming services after the event or season, the economics may weaken.

If subscribers remain, valuations may continue rising.

Therefore a crucial question is churn.

His framework was:

• Low churn → subscriptions remain valuable → team/media valuations can keep increasing.
• High churn → the assumptions supporting those valuations become more uncertain.

A sponsor segment interrupted the investment discussion.

The advertisement claimed that:

• AppLovin began with an $8 domain.
• It initially had no venture-capital funding.
• It became one of the world's largest advertising platforms.
• Its advertising engine now powers e-commerce advertising.
• Advertisements run inside mobile games.
• They can reach more than one billion people.
• The placement provides full-screen, “distraction-free” attention.
• The platform identifies buyers and optimizes campaigns for profit.
• Advertisers set the target and the platform performs the optimization.

A cookware-brand case study in the advertisement claimed:

• Growth from $4 million to $16 million.
• Achievement of profitability.
• A trajectory toward $80 million in the current year.

These are advertiser claims, not claims independently made or verified by the interview participants.

A later sponsor insertion promoted Northwest Registered Agent.

The advertisement said starting a business traditionally requires coordinating multiple services.

It claimed Northwest can provide a business identity package including:

• Business address.
• Domain.
• Website.
• Phone number.
• Other business-identity services.

The advertisement emphasized privacy protection “from day one.”

It also claimed entrepreneurs could avoid spending hundreds or thousands of dollars by using resources Northwest provides for free.

These are sponsor claims and should not be treated as independently verified editorial statements.

Speaker: Mark Cuban
Exact quote: “it's not a bubble that's going to impact most people in the room, right? Or most people across the U.S., but it could just destroy a lot of VCs and a lot of funds and a lot of P.E.”
Context: Cuban distinguishes today's private AI investment boom from the retail-driven dot-com bubble.
Supports: AI valuations; venture-capital risk.

Speaker: Interviewer
Exact quote: “They just invested at the peak. And entry price matters.”
Context: Discussing venture investors who deployed capital at inflated valuations.
Supports: Startup valuation discipline.

Speaker: Mark Cuban
Exact quote: “there's going to be a lot of data centers that are going to be turned into pickleball courts.”
Context: Cuban argues efficiency improvements could leave some AI infrastructure economically stranded.
Supports: AI infrastructure risk; price-performance improvements.

Speaker: Mark Cuban
Exact quote: “That's planning for perfection.”
Context: Reaction to extremely long-term financing and infrastructure commitments.
Supports: Capital expenditure risk.

Speaker: Interviewer
Exact quote: “go public, motherfuckers, go public.”
Context: The interviewer says he is encouraging portfolio companies to obtain public stock that can later function as acquisition currency.
Supports: IPOs; AI consolidation; M&A.

Speaker: Mark Cuban
Exact quote: “I collared my stock because how rich do I need to be?”
Context: Cuban explains why he hedged a concentrated equity position.
Supports: Risk management; concentrated stock wealth.

Speaker: Mark Cuban
Exact quote: “AI is a lot harder to implement than anybody expected.”
Context: Cuban contrasts easy individual prompting with difficult enterprise deployment.
Supports: Enterprise AI adoption.

Speaker: Mark Cuban
Exact quote: “If you need to have forward deployed engineers, that tells you all you need to know about AI.”
Context: Cuban argues that extensive human implementation work demonstrates the limits of present AI autonomy.
Supports: AI implementation; AGI skepticism.

Speaker: Mark Cuban
Exact quote: “I love it. I invest in it. It's going to be amazing. It's the most impactful technology that we've ever seen and may ever see.”
Context: Cuban stresses that his criticism of current AI limitations is not anti-AI.
Supports: Long-term AI optimism.

Speaker: Mark Cuban
Exact quote: “AI is not going to, you know, take away 50% of the jobs. There's just so much shit it can't do that regular people need it to do.”
Context: Cuban disputes near-term mass white-collar displacement predictions.
Supports: AI employment impact.

Speaker: Mark Cuban
Exact quote: “there's no better time to be an entrepreneur.”
Context: AI allows founders worldwide to perform tasks that previously required months, money and specialist labor.
Supports: Startup creation; global entrepreneurship.

Speaker: Mark Cuban
Exact quote: “every single mother fucking business plan ever written in the history of business plans is wrong.”
Context: Cuban argues that imperfect AI-generated business plans are still useful because all startup planning is uncertain.
Supports: AI-assisted company formation.

Speaker: Mark Cuban
Exact quote: “AI is amazing. And particularly for programmers, it's game-changing.”
Context: Cuban distinguishes strong narrow-domain capabilities from weaker autonomous workflows.
Supports: Coding; AI productivity.

Speaker: Mark Cuban
Exact quote: “Agents get bored, right, and they drift.”
Context: Cuban says agents can deteriorate as underlying models change.
Supports: AI maintenance; agent reliability.

Speaker: Interviewer
Exact quote: “It's almost like when we got into the industry and somebody knew how to use the office suite in a PC and then somebody else was like on legal pads.”
Context: Comparing AI-first employees with workers who do not use AI.
Supports: Workplace productivity; AI skills.

Speaker: Mark Cuban
Exact quote: “Nothing's going to fucking be text and pictures in 10 years.”
Context: Cuban argues that world models, video and physical understanding will supersede today's predominantly text-and-image paradigm.
Supports: World models; future AI.

Speaker: Mark Cuban
Exact quote: “It's not ready.”
Context: Cuban chooses a seeing-eye dog over present AI for safely guiding a blindfolded person across a road.
Supports: Embodied intelligence; AI limitations.

Speaker: Mark Cuban
Exact quote: “If I'm going to be wrong on the data centers, it's going to be because of video.”
Context: Cuban acknowledges video, world models and robotics could create enough compute demand to justify large infrastructure buildouts.
Supports: AI compute demand.

Speaker: Mark Cuban
Exact quote: “for health, it's going to make a huge difference, but it's not going to replace doctors.”
Context: Cuban predicts AI augmentation rather than physician replacement.
Supports: Medical AI.

Speaker: Mark Cuban
Exact quote: “Whoever controls the algorithm controls, you know, the election in many cases.”
Context: Cuban argues that social-media distribution systems have become major political forces.
Supports: Politics; social media.

Speaker: Mark Cuban
Exact quote: “their currency is getting you the correct answer and the correct knowledge, social media's currency is keeping you engaged.”
Context: Cuban argues LLMs have stronger incentives toward truth-seeking than engagement-driven social networks.
Supports: AI and political information.

Speaker: Mark Cuban
Exact quote: “if he decides to run for a third term, then I'll run. But he's not.”
Context: Asked whether he would run for president if Donald Trump sought a third term.
Supports: Cuban politics; presidential speculation.

Speaker: Mark Cuban
Exact quote: “the Democrats don't know how to actually do anything.”
Context: Cuban criticizes Democrats while also criticizing Republicans immediately afterward.
Supports: U.S. political polarization.

Speaker: Mark Cuban
Exact quote: “the Republicans on the other hand, you know, do crazy shit, right? Like they don't care about the people in the country.”
Context: Cuban's counterbalancing criticism of Republicans.
Supports: U.S. political polarization.

Speaker: Mark Cuban
Exact quote: “I think we'll get to normalcy. I think people want it.”
Context: Cuban predicts reduced political chaos after the coming election cycles.
Supports: U.S. political outlook.

Speaker: Mark Cuban
Exact quote: “people are forgetting that people are mobile.”
Context: Cuban argues tax and regulatory models must account for behavioral responses and relocation.
Supports: Wealth taxes; state competition.

Speaker: Mark Cuban / interviewer attribution to verify
Exact quote: “in Texas, it's just like build your company and go.”
Context: Contrasting Texas startup culture with Silicon Valley's emphasis on fundraising milestones.
Supports: Founder geography; Texas.

Speaker: Mark Cuban
Exact quote: “he wants his own team. He wants his own team.”
Context: Cuban says this was communicated by Brunson's agent before Brunson left Dallas.
Supports: Mavericks; Jalen Brunson.

Speaker: Mark Cuban
Exact quote: “He's just really a good human.”
Context: Cuban praises Brunson personally.
Supports: Jalen Brunson.

Speaker: Interviewer
Exact quote: “I just cried. I just cried. It's been so many years.”
Context: The interviewer describes the emotional impact of seeing the Knicks win.
Supports: Sports fandom.

Speaker: Mark Cuban
Exact quote: “it's not driven by attendance. It's not driven by wins or losses. It's driven by subscriptions to streaming services.”
Context: Cuban explains his view of the economics behind rising NBA franchise valuations.
Supports: Sports media; franchise valuations.