Thomas Laffont: The $4T AI IPO Wave Is Coming… and We’ve Never Seen Anything Like It

Source context: Authorized transcript of an All-In stage/podcast segment centered on a presentation about the “unicorn economy,” AI investment, private-market concentration, SpaceX, Anthropic/OpenAI, semiconductor economics, and the implications of power-law returns. The presentation is followed by a discussion with the All-In hosts. The principal presenter is identified in the conversation as Thomas, associated with Coatue; his surname is not stated in the supplied transcript. Chamath and Sachs are named during the discussion. Other speaker identities are not consistently recoverable from the transcript and are therefore identified below as Host rather than guessed.

The opening montage describes Coatue as a major hedge fund with “$55 billion under management,” references its flagship hedge fund, and says it is seeking another $1 billion for AI investment.

Transcript-quality issues: The source contains extensive rolling-caption duplication, usually repeating phrases three times. Those duplicates have been removed here without removing substantive repetition. Speaker labels are largely absent. Several names and phrases appear mistranscribed, including “Jake Hal,” “Jay Kyle,” “Mac 7,” “code two framework,” “centercorn,” “tuple,” and the SpaceX IPO figure “1.75.” These are flagged below rather than silently corrected. There is also at least one apparent internal inconsistency: Thomas calls a group the “magnificent eight” but the transcript audibly lists only seven companies.

This document records what the speakers said and argued. It does not independently verify their figures, forecasts, market claims, IPO assertions, company financials, or allegations.

The segment opens with joking about waiting to make a “world podcast premiere” on All-In rather than appearing elsewhere.

A montage or introductory sequence describes:

• Coatue as one of the most successful hedge funds of the past two decades.
• Coatue as having approximately $55 billion under management.
• A referenced fund as its flagship hedge fund.
• Coatue's motivation as finding exceptional entrepreneurs and companies.
• An apparent effort to raise another $1 billion to invest in AI.
• A broader investing philosophy that genuinely revolutionary ideas can become extremely large.

Thomas then tells the hosts to sit back while he walks them through slides updating the state of the “unicorn economy.”

Thomas's first argument is that “the markets are back.”

According to his presentation:

• The unicorn economy, on average, had risen approximately 70% since September 2024.
• Public markets had made a broadly similar upward move.
• The unicorn economy's share relative to the Nasdaq had risen substantially since 2015 but had plateaued in recent years.
• Thomas attributed part of that plateau to the strong performance of public companies, mentioning Palo Alto as an example.
• AI was increasingly dominating fundraising.
• AI had increased its share of fundraising for multiple consecutive years.

However, Thomas said the composition of startup funding had changed significantly.

The “unicorn factory” peaked during the 2021 zero-interest-rate-policy, or ZIRP, period and had since normalized to a considerably lower level that Thomas characterized as pre-COVID-like.

Combining lower unicorn creation with continued capital deployment produced a major concentration effect: Thomas said funding per unicorn had increased roughly fivefold since 2021.

His summary was simple: there are fewer new unicorns, but each is raising much more money.

Thomas used cohort data to evaluate whether the venture ecosystem was functioning normally.

For a pre-ZIRP unicorn cohort:

• He cited approximately 73 companies.
• Twenty quarters after becoming unicorns, roughly 80% had either raised another round or exited.
• He characterized that as a healthy result.

For the 2021 unicorn cohort:

• He cited approximately 479 companies, dramatically more than the earlier cohort's 73.
• Twenty quarters later, less than 20% had either exited or raised another round, according to the presentation.

Thomas therefore framed 2021 as a period in which an exceptionally large number of unicorns were manufactured, many of which subsequently failed to achieve the normal follow-on funding or exit progression.

The emerging question is what happens to the newer 2024 AI-company cohort: will it behave more like the healthy pre-ZIRP cohort or the bloated 2021 cohort?

Thomas did not claim that question was already resolved. He presented it as an important future test of the current AI boom.

Thomas argued that concentration is occurring at two levels.

First, AI itself is absorbing a growing proportion of investment capital.

Second, the top companies inside AI are absorbing a disproportionately large part of that AI capital.

He cited the enormous rounds being raised by companies such as Anthropic and OpenAI as the obvious explanation.

Thomas proposed thinking of the dominant private companies as a kind of new investment index, calling it for the moment the “magnificent eight.”

The transcript lists:

• SpaceX
• Stripe
• Anthropic
• Databricks
• Revolut
• ByteDance
• Anduril

Only seven names appear in the transcript despite the “eight” label; the missing eighth company should be checked against the original slide or video.

Thomas emphasized that the group was not confined to one narrow field: it included internet businesses, AI, fintech and space technology.

He said he would feel comfortable owning such an index for more than a decade.

According to his presentation:

• The group represented almost $4 trillion in value.
• It had strongly outperformed what the transcript renders as the traditional “Mac 7” [likely transcription issue; verify].
• Almost every constituent had individually outperformed that comparison index.

The deeper point was that enormous economic gains were becoming concentrated in a very small collection of companies.

Thomas then shifted from valuations to liquidity.

He said the unicorn economy had always been effective at consuming cash, but the necessary question was how much cash it actually returned.

His framework for a healthy ecosystem required balance between:

1. capital consumed by startups; and
2. capital returned through exits and liquidity events.

Thomas said 2026 was developing into a reasonably strong exit year, although not yet comparable with 2021, and noted that approximately half the year still remained at the time of the presentation.

He then claimed the numbers did not yet incorporate three very large companies expected to go public shortly.

He specifically said:

• SpaceX was expected “in the next few weeks.”
• News that day said Anthropic had confidentially submitted for its S-1.
• The aggregate value of the three expected liquidity events would, according to his slide, be greater than something like the preceding 10 years combined.

The third company is not clearly identified in this part of the transcript and should not be inferred without checking the source.

Thomas contrasted the situation with his presentation at an earlier All-In Summit in 2024, when he believed the ecosystem was fundamentally unbalanced because startups were consuming far more cash than they returned.

His claim was that even before the contemplated major liquidity events, the ecosystem had already become significantly more balanced and should become more balanced still.

Thomas said a major reason for the changed ecosystem was the extraordinary growth of OpenAI and Anthropic.

He described their growth rates as unlike anything previously seen.

A chart beginning in January 2025 apparently compared the combined or relevant scale of the AI companies with established technology businesses. The exact metric represented by the chart is not stated in the transcript and must be verified.

Thomas said the AI companies had successively passed:

• Workday
• ServiceNow
• Adobe
• Salesforce
• Google Cloud
• Azure

He then presented explicit forecasts rather than established outcomes:

• By the end of the year, they could become larger than AWS under the metric in his model.
• By 2028, they could potentially become larger than all of Microsoft.

Thomas stressed that these were based on assumptions and forecasts.

He also made an important strategic observation: the incumbent hyperscalers were not merely being disrupted by AI challengers; they were themselves financing and enabling the disruption, through enormous investments following the ChatGPT breakthrough.

He called the scale of that investment unprecedented.

Thomas then explained how Coatue thinks about SpaceX.

His starting observation was that the variable most strongly correlated with SpaceX's valuation in their analysis was launch cadence.

At the simplest level, this was intuitive: a launch company launching more rockets should have greater economic value.

But Thomas said a second measure produced a more interesting result: valuation divided by launches.

That figure had initially stayed within a relatively fixed range, then began rising. In other words, the market was valuing each incremental launch more highly even as SpaceX increased launch volume.

Thomas's proposed explanation was that the quality of SpaceX's business model itself improves as launch volume grows.

He divided the development into phases.

• The company is primarily trying to make its rockets work.
• Rocket development is difficult.
• Revenue may consist of a limited number of government customers.
• Those contracts can represent unpredictable, one-time business.

Once SpaceX operates a constellation, its economics change.

A constellation provides:

• an end market for the launch capability;
• recurring revenue rather than only one-off launches;
• a relationship in which more satellites can support more subscribers and therefore more revenue.

The next stage is not one constellation but several.

Thomas predicted that many:

• corporations,
• governments, and
• militaries

would want constellations of their own so that they could “control their own destiny.”

That would further broaden SpaceX's customer base.

At sufficient scale, Thomas argued, SpaceX stops being merely a launch company and becomes a technology platform.

Platform economics create both:

• more customers in the core business; and
• entirely new businesses.

He mentioned possibilities including:

• space-based data centers;
• lunar opportunities;
• Mars;
• other space applications.

Thus the thesis was not merely that more launches produce more revenue. It was that each increase in launch capability unlocks structurally better and more recurring businesses, which could justify rising valuation per launch.

Thomas then presented an analysis of the probability of achieving a 10x return from companies at different valuation stages.

For growth investors, he said, a 10x investment is rare and highly attractive.

His data reportedly showed:

• A unicorn had about an 8% probability of eventually becoming a decacorn.
• A decacorn, defined here as more than $10 billion, had only modestly better odds—approximately 8% to 13%—of becoming a $100 billion company.
• But a “centacorn”, defined as a company worth at least $100 billion, showed approximately a 31% chance of having achieved a further 10x in the public-and-private-company dataset.

Thomas found this counterintuitive.

Instead of the expected pattern in which enormous size makes further exponential returns progressively less likely, the data suggested that the businesses which had survived all the way to extreme scale might actually contain attributes that allowed further compounding.

The result later became a major subject of discussion with the hosts.

Thomas added a fresh chart showing how long companies historically took to move from approximately $500 billion to $1 trillion in market capitalization.

Historically, that transition generally took multiple years.

He said something unusual had recently occurred:

• three companies made the transition during the same year;
• two of them did so within a matter of weeks.

He did not immediately declare a definitive explanation, but presented the phenomenon as evidence that the velocity of value creation among the leading companies had changed materially.

Thomas deliberately contrasted rapid compounders with Cerebras, which he said had “just went IPO.”

He disclosed his own connection:

• he had been a Cerebras board member for a long period;
• he had led its Series B.

Its history illustrated that not all major technology successes compound smoothly.

Thomas described:

• long development periods;
• “dark periods”;
• multiple years without new capital;
• difficult engineering work necessary to complete the technology.

He said this eventually led to a massive OpenAI contract, after which the value of Cerebras could increase dramatically. The transcript says the contract could “tuple” the company's value; the intended word or multiplier is unclear and requires source checking.

Thomas therefore offered Cerebras as evidence that patient technical development can suddenly convert into substantial commercial value after years when financing markets appear unreceptive.

Thomas broadened the Cerebras example into a semiconductor-sector thesis.

He said semiconductors as a whole were experiencing a “generational run.”

Referring to a conversation with Brad Gerstner and performance since the 2024 All-In Summit, Thomas said the semiconductor industry had dramatically outperformed the broader index.

That led into a discussion of memory demand, because Thomas believed increasingly personalized AI would require substantially greater storage or memory resources.

Thomas said one theme he had heard repeatedly from speakers that morning was that an AI system becomes more useful as it knows more about:

• the user's business; or
• the individual user.

He used restaurant booking as a simple example.

An effective AI assistant should already understand matters such as:

• when the user prefers to eat;
• what food the user likes;
• other accumulated personal preferences.

Thomas's conclusion was that AI systems would require much more retained information per user.

Coatue's estimate was that memory per user could quintuple.

Thomas connected that prospective demand directly with strong market performance in memory-related semiconductor companies.

Thomas then addressed what he described as one of the controversial questions of the previous 12–24 months:

Where is the AI revenue and where is the return on investment?

Coatue's estimate, as stated in the presentation, was:

• the AI ecosystem was approximately 140 billion “today”;
• it would reach approximately 300 billion “this year”;
• it would double in 2027.

Currency and exact metric are not specified in the transcript and require verification.

Thomas divided the revenue opportunity into three principal pillars.

Consumer revenue is based on:

• number of subscribers;
• multiplied by ARPU.

Thomas said advertising was an AI revenue source that many observers overlooked.

Coatue estimated:

• approximately one-quarter of ads served by Meta and Google were already AI-enabled;
• eventually AI penetration could reach 100%;
• that represented approximately 150 billion under their estimate.

Again, the units and methodology should be verified from the slide.

The third pillar was enterprise adoption.

Thomas cited advances associated with:

• Claude Code;
• Codex;

and their use inside businesses.

His conclusion was that consumer subscriptions, AI-enabled advertising and enterprise AI together produced a much more substantial revenue ecosystem than critics sometimes acknowledged.

Thomas contrasted the current period with previous investment eras.

His distinctive claim was not simply that AI was creating a new software cycle. Instead, almost every sector of the economy was undergoing major technological change at the same time.

He gave several examples.

Software was the obvious AI-affected category and required little explanation in his presentation.

Thomas predicted that within a few years Starlink would support a device capable of making a phone call anywhere in the world.

He argued that people tend to treat global connectivity as a solved problem until they experience a dropped call.

Using an earlier “Nikesh” framework around profit pools, Thomas argued that Starlink should be thought of as attacking the global telecommunications profit pool for broadband and wireless, not merely a small satellite niche.

Compute demand was already transforming semiconductor markets.

Thomas referred to senators who had spoken earlier about data centers changing the energy equation in Pennsylvania.

Thomas referenced Ferrari's attempt “last week” to introduce electric and autonomous technology.

He said the reaction raised questions about the future of established automotive franchises in an electric and autonomous world.

Thomas said GLP-1 drugs were producing profound changes in:

• food consumption;
• alcohol consumption;
• dietary composition;
• consumer focus on wellness.

His larger thesis was therefore one of concurrent disruption across technology, communications, energy, transportation, food and consumer behavior.

Thomas summarized his presentation around several conclusions.

First, he believed the new unicorn economy was healthier, with AI playing a substantial role in restoring that health.

Second, the strongest businesses were compounding faster than ever.

That produced an uncomfortable investment consequence: if gains are increasingly concentrated among a few extraordinary businesses, the cost of failing to own a winner becomes higher.

Third, disruption was affecting almost every part of the global economy.

Fourth, Thomas emphasized that all of these changes were occurring even though, in his framing, “we don't even have super intelligence yet.”

Looking two years forward, he expected the environment to change substantially again.

The hosts summarized the theme with the phrase:

“The power law rules our lives.”

They framed the central issue as the consolidation of extraordinary gains into a small number of companies.

A host asked what the concentration of enormous outcomes meant for the structure of private-market investing.

The question contrasted several positions in the ecosystem:

• very early-stage investing;
• Series A investors such as Craft Ventures;
• Coatue moving between public and private markets;
• Brad Gerstner considering where to allocate his attention;
• Andreessen Horowitz potentially moving toward increasingly broad or index-like venture exposure.

The host asked what the resulting market would look like for:

• limited partners;
• angel investors;
• venture firms;

over the next decade or two.

The underlying observation was that private markets now operate very differently from the venture playbook of approximately 20 years ago.

Thomas identified an important positive difference from prior cycles: the outcomes are much bigger.

Private companies were reaching sizes previously thought impossible.

For him, SpaceX illustrated this point particularly well.

He contrasted the current environment with the ZIRP period:

• ZIRP generated numerous companies but smaller ultimate outcomes;
• many companies did not go public;
• the current era was producing extraordinarily large companies;
• those companies also appeared more willing to enter public markets.

Thomas said he believed both Anthropic and OpenAI had publicly expressed a desire to go public.

He saw that willingness as healthy because it would create genuine liquidity and public-market accountability.

However, he also identified a major risk: the same K-shaped, power-law concentration appearing throughout economic life seemed to be appearing among startups.

He noted that the number of “centacorns” had been relatively stagnant.

If the ecosystem produced no new $100-billion-scale companies over the next decade, especially after already seeing few new entrants for the preceding couple of years, Thomas said that would be a warning sign.

A host drew the provocative implication from Thomas's probability slide.

If companies that have already reached $100 billion are statistically more likely to generate another huge return, a rational LP might simply:

• wait until a company reaches $100 billion;
• put as much capital as possible into it;
• avoid the fragility and work associated with much earlier investments.

But the host identified the obvious counterforce: supply and demand determine valuation.

As more capital chases the same tiny group of apparent winners, valuation multiples can become extreme.

The host said the market was reaching valuations disconnected from traditional metrics and cited an earlier explanation from Bill Ackman: investors were effectively making venture-style bets in trillion-dollar companies at valuations of 50 times revenue or 100 times revenue.

The host therefore asked where LPs, private investors and high- or ultra-high-net-worth individuals should rationally put capital, and whether it was dangerous for everyone to race into the same three names.

Thomas agreed that concentrating in the biggest winners had been an effective strategy over the previous five years, but emphasized that the relevant investment problem is what will work during the next five years.

His main pushback against the bubble analogy was:

“These are not fake companies.”

He contrasted the present situation with both:

• the 2000 technology bubble; and
• 2021.

Thomas argued that the current leaders:

• generate substantial revenue;
• operate at scale;
• are growing at historically exceptional rates.

He also referred to a report that Anthropic had experienced a profitable month, though he qualified this with “I believe is what was reported.”

He therefore considered the businesses economically real even if their valuations could still prove wrong.

Thomas and the hosts agreed that public markets provide the ultimate test.

One host supplied “equalizer”; Thomas then developed the metaphor:

“It will be the great antiseptic.”

Public markets would not care about his slide presentation.

Thomas welcomed the prospect of SpaceX, OpenAI and Anthropic being subjected to:

• market pricing;
• short sellers;
• critics and commentators;
• public debate;
• political scrutiny.

He described himself as a strong believer in markets and said he was excited to see those companies prove themselves under that scrutiny.

The important qualification came from a host: the traditional idea that the market discovers a company's true value immediately after listing may no longer hold because of large passive flows.

The host suggested that index and passive buying might postpone genuine price discovery as the market works through abnormal supply-and-demand effects.

A possible revised test was proposed as approximately “six month plus one” rather than day one.

In other words, the group suggested waiting around six months after an IPO before treating the market price as a cleaner indication of underlying value.

A host then asked Thomas whether the accelerating returns of already-gigantic companies reflected something structurally wrong or inefficient in markets, or whether observers were simply seeing survivor bias in a tiny sample.

Thomas warned against drawing too much from such a small number of companies.

He used Anthropic and Claude Code as the example.

In his view, Anthropic before Claude Code was substantially different from Anthropic after Claude Code.

One event or product could suddenly alter the trajectory not merely of a company, but of a large portion of an industry.

Thomas compared such an unpredictable event to the Mule in the Foundation series—an unexpected force that could not have been incorporated into normal forecasts.

His conclusion was cautious: it may represent a profound structural pattern, or it may simply be a rare one-time phenomenon.

However, Thomas was much more confident about another conclusion: he believed the previous narrative that frontier AI models would become commodities and their developers would lose differentiation had been “pretty thoroughly disproven.” This was Thomas's opinion, not independently established fact.

A host asked how Coatue keeps its investment strategy organized as:

• its asset base grows;
• it moves into more strategies;
• it invests in areas such as data centers;
• individual winners become so enormous that hindsight makes concentrated bets appear obvious.

Thomas said the process of building presentations like this was itself part of his answer.

He said:

• preparing the presentation had taken substantial effort from him and his team;
• the preceding two weeks had effectively become a full-time job;
• they wanted the information to be accurate.

But research also served a personal investment function: it re-anchored his conviction.

Thomas said that listening to a thousand different people could leave him distracted and unsure what he actually believed.

His solution was to return to:

• models;
• numbers;
• valuation;
• fundamental “ground truths.”

That quantitative reconstruction helps him restore conviction.

He argued that the recent trillion-dollar compounders were not imaginary speculative vehicles:

• they had often existed for decades;
• they had real underlying businesses;
• some traded at very low earnings multiples relative to other S&P 500 companies.

A host characterized the phenomenon as accumulated “energy” finally being released; Thomas agreed.

Thomas relayed an argument someone had made to him about semiconductor economics.

The person told him that if a company wants to design a chip, it can—despite the difficulty—turn to TSMC as a manufacturing partner.

But, according to the argument, if the company wants to manufacture memory, “there is no TSMC.”

Thomas used that difference to pose an investment question:

Should memory companies therefore receive different valuation multiples from ASIC-related companies because the industrial structure and barriers to entry are different?

The presentation did not supply a definitive answer; the observation was offered as a framework for considering memory-company valuations.

A host remarked:

“The wrath of Lina Khan can be seen clearly in this.”

The host then asked Sachs for his view of how policy and elections affect investment and company outcomes.

However, the conversation visible in the transcript immediately pivots to the power-law valuation chart.

There is no developed Sachs answer to the policy/election question in the supplied material.

That missing response should not be reconstructed or inferred.

The group returned to Thomas's surprising 10x probabilities.

A participant extrapolated Thomas's data one level further:

If a $100-billion company displayed a 31% probability in the analysis, what were the odds that a $1 trillion company could become a $10 trillion company?

The participant speculated that the probability might be:

• 50%;
• perhaps even 100%;
• but, at minimum, seemingly greater than 30%.

This was speculation arising from the chart, not data presented by Thomas.

Another speaker suggested a mechanism: every successive valuation threshold acts as a filter for businesses with durable compounding advantages.

To survive to the next level, a business must repeatedly demonstrate:

• a compounding advantage;
• durable earnings;
• an ability to increase or sustain that durability.

The speaker compared this with fundamental, Ben Graham-style valuation analysis.

Another participant argued that to reach the trillion-dollar club at all, a company normally needs a dominant business. Once there, the central limitation becomes market saturation.

Yet observed addressable markets have repeatedly turned out larger than investors previously expected.

Thus enormous size need not automatically mean a business is near the end of its growth runway.

A participant identified another possible constraint besides saturation: government intervention against monopoly power.

The breakup of the Bell System was used as a hypothetical.

The speaker asked how large Bell might ultimately have become had it remained intact, speculating that it might conceivably have established dominant positions in:

• internet access;
• commerce;
• e-commerce;
• other adjacent industries.

The point was not that Bell definitely would have done those things, but that government intervention can stop a dominant company's compounding path before normal commercial saturation does.

The hosts joked that a trading bot should simply begin buying a company's shares as soon as it reaches $1 trillion in market capitalization.

The discussion recalled Apple as possibly the first company to cross the trillion-dollar threshold, although the speaker expressed uncertainty.

Another participant referred to an unnamed study allegedly showing that, over a 10-year period:

• buying the top 10 Nasdaq companies by market cap;
• rebalancing annually;

would have produced something like a 3x advantage or return [the transcript's exact comparison basis is unclear].

The anecdote reinforced the counterintuitive theme that paying up for demonstrated market dominance has historically sometimes beaten trying to identify winners much earlier.

The final major question concerns what happens when enormous private-company gains are actually realized.

A host asked Thomas to consider the effect if roughly $3 trillion to $4 trillion were distributed:

1. back to general partners;
2. from GPs to limited partners;
3. then recycled into new investments.

The host asked about consequences for:

• competitive dynamics;
• entrepreneurship;
• Silicon Valley more broadly.

Thomas initially joked about a host's earlier bearishness on California real estate.

That led to brief banter about:

• San Francisco homes;
• selling property;
• a “40,000 square foot mausoleum”;
• “protesters not included.”

The jokes do not establish substantive property-market claims.

The conversation then returned to SpaceX.

Thomas explicitly said he did not know whether “1.75” was the right price for the SpaceX IPO.

The transcript does not specify whether “1.75” means:

• $1.75 trillion;
• another valuation unit;
• or something else.

Thomas said, “frankly, I have no clue.”

Rather than defending a specific price, he returned to underlying economics.

He claimed the worldwide profit pool for telecom and service providers was somewhere between “two to 400 billion” [transcript unclear; likely missing digits or words].

His argument was that SpaceX had entered a core business only a few years earlier and was already targeting a profit pool worth hundreds of billions of dollars with what he considered a substantially better product.

His supporting product claims included:

• Starlink works consistently;
• it does not depend on traditional radio towers.

The central investment case therefore rested on the size of the market SpaceX could attack rather than Thomas claiming precision about the correct IPO valuation.

Thomas said the scale of the coming capital cycle was difficult to forecast because there was little historical precedent.

He compared it with earlier:

• ride-sharing wars;
• food-delivery wars.

In those markets, companies with enormous pools of capital eventually used that money to subsidize price competition.

Thomas therefore asked whether the same dynamic could emerge between OpenAI and Anthropic.

If both companies possess enormous financial resources, one might eventually cut prices aggressively to win market share.

Another participant replied:

“Rationally, they should.”

Thomas agreed that they should under some competitive logic.

He nevertheless qualified the prediction:

• these companies spend extraordinary amounts on infrastructure;
• therefore large apparent pools of capital are not necessarily free cash available for subsidies;
• the outcome is not obvious.

His broader prediction was that the ecosystem would produce counterintuitive changes that cannot presently be forecast with confidence.

Thomas said he hoped to return in two years and compare:

• what the group predicted correctly;
• what it got wrong.

The hosts suggested he should instead return every year.

They expressed appreciation for the amount of work behind the presentation.

One host said the exercise demonstrated the value of periodically slowing down and meditating on “the actual state of reality,” describing the result as grounding.

Another joked that this was “incredibly rich” coming from that particular host, creating a brief exchange over whether the remark was a compliment or insult.

The segment ended with thanks, applause and music.

Quote: “We're in an idea business and when you have a truly revolutionary idea, it can get really big.”

Context: Opening investing philosophy.

Supports: Venture investing; power-law outcomes.

Quote: “So, we have fewer unicorns that are each raising more.”

Context: Funding per unicorn has risen as new-unicorn creation has fallen.

Supports: Capital concentration.

Quote: “We need to have a balance between the amount of cash consumed to the amount of cash returned.”

Context: Explaining why exits matter to venture-market health.

Supports: Liquidity; venture ecosystem health.

Quote: “The quality of SpaceX's business model increases the more you launch.”

Context: Explaining rising SpaceX valuation per launch.

Supports: SpaceX platform thesis.

Quote: “The more an AI system knows about your business or you as a user, the more useful it is.”

Context: Argument for substantially greater memory demand.

Supports: AI memory; personalization.

Quote: “Where's the revenue?”

Context: Introducing Coatue's answer to doubts about AI monetization and ROI.

Supports: AI economics.

Quote: “The winners are compounding faster than ever, which means the cost of not being in a winner are higher than ever.”

Context: One of Thomas's final presentation conclusions.

Supports: Power-law investing.

Quote: “The power law rules our lives.”

Context: Host's summary of Thomas's presentation.

Supports: Concentration; winner-take-most economics.

Quote: “These are not fake companies.”

Context: Pushback against treating today's giant AI companies as purely speculative entities.

Supports: Valuation debate.

Quote: “It will be the great antiseptic.”

Context: Describing public markets as the ultimate test for private-company narratives and valuations.

Supports: IPOs; market discipline.

Quote: “I understand models, I understand numbers.”

Context: Explaining how he regains conviction amid market noise.

Supports: Coatue investment methodology.

Quote: “If I want to make memory, well, there is no TSMC.”

Context: Argument about barriers in memory manufacturing versus chip design/foundry access.

Supports: Semiconductor valuation.

Quote: “To get to that level, let's call it the trillion-dollar club. You have to have a dominant business.”

Context: Discussion of why very large companies might continue compounding.

Supports: Market dominance; power laws.

Quote: “Frankly, I have no clue.”

Context: Thomas explicitly refusing to claim that the “1.75” SpaceX IPO price/valuation was correct.

Supports: Valuation uncertainty.

Quote: “Rationally, they should.”

Context: Response to Thomas asking whether OpenAI or Anthropic might use abundant capital to initiate a price war.

Supports: AI competition.