Peter Thiel on AI Acceleration, Global Power, and the Values of Its Builders

Walter Isaacson introduces The Builders as an attempt to understand the moral outlook of the people shaping the future. In his conversations, he has encountered strong libertarianism, Nietzschean ideas, skepticism about empathy, and sometimes a bleak view of humanity. Powerful technologies will affect everyone, he argues, so people outside the builders’ immediate circles need a voice in deciding where they lead.

His guest is Peter Thiel, whom Isaacson describes as a builder of builders. Thiel has backed people such as Elon Musk, Alex Karp, and Sam Altman, as well as political figures including J.D. Vance. The opening montage presents the contrast between his reputation as an unusually perceptive investor and critics’ concern about the influence he has accumulated.

Isaacson wants to know what Thiel believes society is building toward, whether current fears about AI are justified, and what lies behind his lectures about the Antichrist. Beneath those questions is a concern about power: people may possess the drive to create transformative technology without having the empathy or moral steadiness to manage its consequences well.

Isaacson opens the interview by asking what explains the backlash against AI. Thiel agrees that it is happening but warns that explanations are speculative. AI’s winners and losers remain unclear. He expects substantial growth accompanied by disruption, while familiar dystopian scenarios—from autonomous machines to a world resembling The Matrix—give people additional reasons to worry.

Thiel notes that these fears did not originate with outside critics. They have circulated among AI developers for years. He describes OpenAI’s founding after Google acquired DeepMind as partly a reaction to the idea that such important and potentially dangerous technology should sit inside a single company. Later departures from OpenAI reflected further disagreements over how seriously to treat those risks.

Isaacson asks Thiel to explain effective altruism, a movement often invoked without much clarity in public debate. Thiel traces part of its intellectual background to futurist circles around Berkeley and Oakland in the 2000s. These were not necessarily the places producing the biggest technology companies, but they hosted energetic arguments about rationality, the future, and the possibility of artificial intelligence.

The futurists’ important intuition, Thiel says, was that AI might arrive sooner than conventional wisdom suggested. After decades of overpromising, the field had become easy to dismiss. Yet computers kept getting more powerful. Perhaps the obstacle was not entirely the absence of a better algorithm; sufficient hardware could make capabilities that had long seemed remote much more plausible.

Those thinkers also understood that AI would be socially transformative and potentially dangerous, Thiel continues. It could be weaponized or become autonomous in ways that resembled science fiction. He considers parts of the movement excessively insular and almost cultlike, but says it is important to recognize what opponents understood correctly rather than dismiss their entire worldview.

Thiel uses Anthropic’s commercial success to complicate his criticism of the movement. He offers a speculative estimate of the company’s possible value at an IPO and asks how a supposedly irrational group could build something so successful so quickly. His broader point is methodological: before criticizing opponents, try to explain their strongest case as well as they could themselves.

The strongest case for the effective altruists begins with their recognition that AI was approaching, Thiel says. He distinguishes several meanings of AI: relatively limited surveillance systems, a godlike superintelligence, and something in between that behaves like a human. The Turing test concerned this middle category—whether a machine’s outputs could convince someone that it was a person.

Even humanlike AI would raise enormous economic questions, Thiel says. Would it complement workers or substitute for them? Isaacson describes students at Tulane worrying that the new technology will remove the jobs they expect to enter. He asks whether greater productivity will create new opportunities or leave society searching for a different role for human labor.

Thiel prefaces his forecast by saying nobody can be certain. His expectation is that AI will raise productivity substantially and exert a disinflationary force, making goods and services cheaper. Unless a small group of companies captures the gains very narrowly, he expects the median person to benefit. That prediction does not imply that every industry or individual will avoid disruption.

He compares the anxiety to nineteenth-century fears that machinery would replace people. Machines did displace workers in particular industries, sometimes very large ones, but also freed people to do other productive work. That remains his intuition about AI. Isaacson stops him at the qualification about concentrated gains: is a handful of companies capturing the technology precisely what people should now expect?

Thiel says much of the value created by technological innovation over the past two centuries flowed to consumers, workers, and the wider economy rather than remaining with businesses. Software and the internet were partial exceptions. They produced highly profitable companies with characteristics of natural monopolies, so it is reasonable to ask whether an even larger AI market could repeat that pattern.

As an investor, Thiel likes to ask founders what gives their company a monopoly. The social merits of monopolies can be debated, and intellectual property can create legitimate forms of exclusivity. From inside a business, however, relentless competition is unattractive because rivals can eliminate profits. AI pitches, he says, tend to be much stronger on growth and technical progress than on durable protection from competition.

Thiel describes the leading position in AI as repeatedly changing hands. DeepMind led important work in the early 2010s, then OpenAI advanced with transformer models and ChatGPT. He credits Sam Altman with securing chips and data-center capacity early, but says Anthropic may now occupy a stronger position in some respects. Google, Meta, Grok, and Chinese competitors remain close enough to keep the outcome uncertain.

The founders’ behavior looks to Thiel like a struggle for survival rather than a comfortable monopoly. If competition continues, he expects benefits to spread more broadly. Still, he acknowledges that a person can end up better off on average while experiencing frightening volatility along the way. A predictable wage increase feels very different from an uncertain outcome with a wide range of possible gains and losses.

That distinction helps Thiel understand the backlash even though he considers it misguided overall. The potential benefits are unfamiliar and uncertain, while the threatened disruption is easier to imagine. People do not generally welcome risk for its own sake. Isaacson adds that competition may be economically healthy while simultaneously making safety coordination, regulation, and any proposed pause harder to achieve.

Thiel says he worries about the kind of high-technology totalitarianism he sees developing in China. In his account, a system need not reproduce the mass violence of earlier dictatorships to become oppressive. Surveillance and data can give authorities extraordinary knowledge about individuals. He does not want that political model to become the dominant outcome of the AI race.

Isaacson asks whether Thiel agrees with arguments from David Sacks and Donald Trump that competition with China makes pausing AI unacceptable, even if the technology might pose an existential threat. Thiel broadly agrees. What makes him less comfortable presenting the argument publicly is that major AI companies themselves are raising alarms and asking for slower development.

Those warnings have not persuaded Thiel to join the pause camp. He still favors pressing the accelerator. When Isaacson challenges him about recent frightening reports, Thiel argues that developing AI may be necessary to control AI. If China continues developing the technology, he believes maintaining an American lead becomes more urgent, including for protection against increasingly capable cyberattacks.

Thiel then turns to what he sees as the more central economic issue. He describes a middle-class society as one in which people expect their children to live better than they do. The United States long sustained that expectation, but millennials and Generation Z have increasingly lost it. He is uncertain whether AI alone can restore the promise, yet sees it as an important remaining source of growth.

AI’s gains may be distributed unevenly, Thiel admits. But banning the internet in the past and AI now would, in his view, leave very little growth. He associates stagnation with the slower economic experiences of Europe and Japan. In a society without expansion, each winner needs a loser, making politics and social life more zero-sum and potentially harsher.

Isaacson challenges that argument through the experience of globalization. He supported freer trade, immigration, and capital movement because he expected them to produce growth, and they did. What he underestimated was how severely some people would be left behind. Could AI generate the same combination of rising aggregate wealth and deep losses for particular communities or workers?

Thiel agrees that globalization generated growth in the United States and elsewhere, with uneven results. He nevertheless distinguishes it from technological progress. Globalization can exploit differences in labor costs: consumers gain access to cheaper goods while poorer countries expand production. As the wage differences narrow, he says, the opportunity for mutually beneficial arbitrage shrinks and the arrangement can become more competitive.

Technology, by contrast, can mean doing more with fewer resources. At its best, Thiel says, it raises output per person rather than simply expanding the economy by adding more people or relocating production. He thinks the internet helped but was not large enough to restore strong growth. AI could be a bigger transformation and might raise living standards more smoothly, though he presents this as a hope rather than a certainty.

After a break, Isaacson returns to the possibility of catastrophe. He points to accounts of AI agents behaving in ways that appear conspiratorial or misaligned with human welfare, and to warnings from figures including Dario Amodei and Elon Musk. Does Thiel assign even a five- or ten-percent probability to a doomsday outcome, and has he prepared a bunker or other personal escape plan?

Thiel doubts that a bunker would help in a genuinely global AI catastrophe. He is also skeptical of people’s ability to form reliable numerical intuitions about such unprecedented events. Five or ten percent strikes him as extremely high for the destruction of everyone on Earth, and even a very small probability would be troubling. He does not think that outcome is likely, but does not claim to know how to measure it precisely.

His response is that failing to innovate also creates danger. If China surpassed the United States in AI while geopolitical hostility intensified, he fears it could use the advantage as a weapon. Isaacson proposes imagining a world without China to isolate the safety question. Thiel rejects the premise as too far removed from the actual situation in which decisions must be made.

Even under competitive pressure, Thiel argues, AI companies have reasons to prevent disastrous failures and maintain internal controls. They also understand their systems better than governments do. He questions whether public requests for regulation sometimes take advantage of the fact that officials do not know what rules to write and move slowly. Such requests might shift responsibility rather than settle the practical safety problem.

Thiel reads Trump’s insistence that existing criminal penalties already apply as another way of assigning responsibility to companies themselves. He recalls a similar period at Meta, where he served on the board from 2005 to 2022. When Mark Zuckerberg publicly requested clearer regulation, Thiel’s impression was that congressional committees were more interested in criticizing him than in designing a workable regulatory system.

Calls for regulation also reach an internal audience, Thiel continues. Telling employees that they work on a powerful and potentially dangerous technology reinforces the idea that their work is consequential. It can communicate that they are building one of the world’s most important companies. He acknowledges genuine fear and resignations inside labs while arguing that the same language can also motivate staff and strengthen an organization’s identity.

Isaacson quotes Thiel’s argument that an Antichrist figure could gain power by constantly invoking Armageddon and existential risk, persuading people that catastrophe is imminent rather than causing it directly. He asks whether that pattern is appearing in the AI debate. Thiel says his purpose is to prompt questions about the political institutions people might accept in the name of preventing disaster.

Thiel draws a parallel with postwar nuclear debates. Scientists such as Oppenheimer supported stronger international authority over nuclear weapons because they feared a future nuclear war. Isaacson responds that these warnings helped produce arrangements that have, to a degree, worked. Thiel agrees that partial controls were beneficial, while cautioning that the historical test has lasted decades rather than indefinitely.

He explicitly acknowledges serious existential dangers from climate change, other environmental problems, nuclear weapons, and potentially AI. His additional concern is a softly totalitarian world government. If authoritarian power covered the whole planet, there would be nowhere outside it. Unlike the Soviet Union, which existed alongside external alternatives and eventually collapsed, a universal system might become very difficult to reverse.

Asked what he means by the Antichrist, Thiel describes a biblical and medieval image of a tyrant who rules the world at history’s end. He treats it as a literary and political thought experiment: how might such a ruler come to power in different periods? In an earlier era, one might imagine a scientist creating superior weapons and conquering everyone else.

Thiel thinks that picture may fit the seventeenth century better than contemporary politics. Today, fear of science and technology may be a stronger organizing force than enthusiasm for technological conquest. People frightened of environmental or technological catastrophe might support increasingly powerful global institutions. He emphasizes the institutional risk rather than claiming that any particular environmental activist is literally the Antichrist.

He does not reject all international regulation. The nuclear arrangements that emerged were, in his judgment, more good than bad. His concern is that proposed authorities are often insufficiently examined. He compares this with globalization: because its advocates did not think critically enough about its consequences, avoidable harms accompanied genuine gains. The political remedy deserves as much scrutiny as the danger it addresses.

Thiel cites Eliezer Yudkowsky’s book If Anyone Builds It, Everyone Dies as an example. He considers its discussion of AI catastrophe technically detailed, even if the probabilities remain debatable. By comparison, he finds its prescription for global regulation vague. The imbalance worries him: extensive analysis of how technology might destroy the world is paired with much less analysis of how the proposed control system would work.

Isaacson asks why Thiel chooses such unsettling religious language when he could describe these concerns in neutral terms. Thiel acknowledges that it gets attention. More fundamentally, he thinks the framing changes the choices people see. A slogan such as one world or none appears to make global government the obvious alternative to destruction, without requiring further examination of how power would be constrained.

By placing the Antichrist alongside Armageddon, Thiel wants to make both unchecked global power and catastrophic technological disaster unacceptable. His stated aim is to find a moderate third path rather than resign society to either extreme. He says the religious comparison forces the political danger into the discussion, where a simpler choice between unity and annihilation can hide it.

Isaacson turns to the backlash against technology billionaires, including people Thiel helped finance. He worries that a small group now has extraordinary power over the future and may have fewer moral guardrails than earlier generations of leaders. Thiel first asks whether the concern is economic or political. Isaacson answers that it is both: a handful of people increasingly shape the conditions under which everyone else will live.

Thiel says the term builders captures something his own Founders Fund name does not. Starting a company matters, but its transformative effect comes from growth compounding over time. The undertaking is a marathon. Accepting Isaacson’s criticisms for the sake of argument, he asks what alternative form of organization would produce comparable innovation without relying on an unconventional founder.

The alternative Thiel sees is an MBA-trained bureaucrat overseen by a cautious board that avoids new projects. He describes a stark choice between charismatic, sometimes dangerous leaders who inspire creation and dull bureaucracies that protect existing routines. Isaacson tests the claim against Apple after Steve Jobs: did the company lose its creative direction when professional management replaced its founder?

In Thiel’s account, Apple after Jobs remained extremely successful but did little fundamentally new. He contrasts research spending with share repurchases and portrays Tim Cook as an efficient operator rather than a charismatic inventor. The smartphone had created a vast opportunity that could still be expanded through manufacturing, branding, and distribution without another invention on the same scale.

Either model can reward shareholders, Thiel says, but he thinks figures such as Musk and Jobs create more value for society by bringing new possibilities into existence. Isaacson then asks about Thiel’s own scorecard. He doubts that money alone explains the people he writes about and wants to know what motivates an investor who has already achieved exceptional financial success.

Thiel admits that he probably likes winning too much. He also describes an enduring interest in ideas and intellectually engaging conversations. Figuring things out can be worthwhile for its own sake, even when he tells himself it might improve his investing. He is skeptical of that convenient justification and presents intellectual curiosity as a motivation that need not reduce to financial utility.

Isaacson asks what broader purpose the builders should pursue: democratic freedom, human flourishing, or something beyond finding the next profitable opportunity. Thiel returns to his long-standing distinction between progress in bits and progress in atoms. He thinks advances in the physical world slowed in the 1970s even as computing continued to improve, narrowing the practical ambitions available to engineers.

Recalling Stanford in the late 1980s, Thiel says mechanical, aerospace, chemical, and nuclear engineering already looked less promising than they once had. In his telling, physical industries moved slowly and faced heavy regulation, while computer science became the exceptional field in which dramatic new things could still happen. The problem matters because human beings live in bodies and depend on material conditions, not just digital services.

He wants AI to help restart science and engineering across the physical world. Medicine is a central example. Even if radical life extension seems unsettling, curing dementia or Alzheimer’s disease should be an appealing goal. He hopes the new technology can stimulate progress in fields where improvement has become difficult, making its contribution larger than another successful software business.

Isaacson revisits a previous exchange with Ross Douthat, in which Thiel hesitated before answering whether he wanted humanity to survive. He asks whether preserving the human race should be the main objective. Thiel calls survival extremely important but resists making it the only goal. Simply continuing present conditions, he argues, does not settle the question of what technology and society ought to achieve.

Thiel thinks Silicon Valley is poor at articulating technology’s meaning, and perhaps it should not be the only place expected to provide an answer. There may be no single purpose. Avoiding existential catastrophe is plainly part of the task, but technologies can also reduce existing threats. He gives nuclear power as an example of a technology that could help sustain human civilization while limiting carbon emissions.

His hesitation in the earlier interview, Thiel explains, came from what he heard beneath the question: why not leave things as they are if improvement creates danger? He rejects that as an adequate response to human suffering. The Enlightenment’s scientific ambition was to relieve the conditions of life, making existence less short and brutal. Serious diseases and avoidable limitations still give people reasons to seek change.

Thiel says clearly that he wants humanity both to survive and to flourish. Longer and healthier lives are desirable, and progress may be necessary to make them possible. His defense of innovation is therefore not that survival is unimportant, but that passive preservation can be an insufficient and even misleading way to think about humanity’s future.

After a break, Isaacson asks about René Girard and mimetic desire. Thiel explains that Girard saw imitation as central to human nature. Children learn language by copying others, and culture spreads through the same powerful capacity. Drawing on Aristotle and Darwin, he describes humans as distinguished by an especially strong aptitude for imitation rather than by freedom from it.

The same mechanism can produce conflict when people reach for the same object, status, or position. Imitation is productive, but it can become destructive and sterile. Thiel describes competitions for positional goods that matter less than participants imagine, and recalls the observation that medals can motivate generals. What looks like a freely chosen ambition may actually be a desire borrowed from other people.

A psychological reading of Girard suggests trying to escape these uncreative rivalries, Thiel says. He distinguishes that interpretation from Girard’s own Christian outlook, which was more skeptical of an individual’s power to stop being mimetic. For Thiel, the practical lesson is to remain aware that crowds can display madness as well as wisdom, and that social approval can pull people away from worthwhile original work.

The builders he most admires often sustain a peculiar vision rooted in first principles or fixed ideas. Musk’s desire to reach Mars was not a socially conventional ambition, but it opened an avenue for building that might otherwise have remained unexplored. Thiel puts Musk and Jobs among the strongest examples of creators who pursued directions that were not simply copies of what everyone else wanted.

Jobs combined existing pieces in a distinctive way to create the internet-enabled phone. Isaacson notices that Thiel’s account of imitation now sounds strikingly similar to language used about AI: Turing’s imitation game, generative systems, and dangerous copying. He asks whether Girard’s thought might contain a warning about the creation of artificial intelligence itself.

Thiel thinks Girard would caution against assigning technology too much autonomy in our explanations. People speak about a nuclear bomb as if it were a godlike force in its own right, when the human beings who build and use it are central. Applied to AI, the question becomes what motivates the people running the companies and what ends they are pursuing.

Isaacson asks Thiel to answer that question directly. Thiel admits that understanding their motives is difficult despite the time he has spent considering it. He nevertheless judges figures such as Musk, Amodei, Altman, and Zuckerberg as more good than bad, and believes they are too easily demonized. Their individual character matters because it influences the companies and the world those companies help create.

Thiel connects this view to Isaacson’s biographies, which portray consequential people as complicated rather than simply good or evil. The deeper opposing position, he suggests, is that individuals do not matter at all—that history would proceed much the same way regardless of who leads. Isaacson agrees that builders matter, then asks whether that makes avoiding evil an essential responsibility.

Thiel says yes, but immediately raises complications about appearances. Publicly performing virtue may be less honest than openly displaying unattractive traits. He contrasts what he calls virtue signaling and vice signaling, using Gates and Musk as examples. His point is not that bad behavior becomes good, but that a reassuring public image cannot by itself establish someone’s moral character.

He then distinguishes two opposites of good: morally evil and simply bad at doing something. A powerful car with poor fuel economy may be criticized on ethical grounds, while a car that constantly breaks down is incompetent at its purpose. Thiel argues that some attacks on builders condemn their morality while implicitly acknowledging that they can accomplish things, unlike a bland bureaucracy that achieves little.

Isaacson asks whether an unremarkable bureaucrat would still be preferable to an evil leader. Thiel says he ranks good above evil, but evil above incompetence. He invokes the different Faust stories: in his reading, Marlowe’s figure is a fraudulent magician who goes to hell, while Goethe’s gains troubling knowledge but ultimately accomplishes enough good to be saved. Thiel prefers the latter sentiment without identifying himself with either Faust or the devil.

Pressed again about preferring an effective but evil person, Thiel says he is making a broad observation about a society exhausted by incompetence. A frustrating government office may not seem wicked, but it can still fail badly at its job. Isaacson asks him to clarify evil, and Thiel acknowledges that fully unpacking its moral dimensions is difficult. The distinction is provocative rather than a complete ethical framework.

Thiel illustrates his argument through stem-cell research. In his account, conservative critics called it morally wrong because it involved embryos while conceding that the science might yield cures. He thinks that weakened their position because the promise of substantial benefits mattered to the public. His interpretation is that describing an effective project as evil can inadvertently strengthen the perception that it is capable of accomplishing something important.

He contrasts that with controversies over transgressive art in the 1990s. Critics, he says, successfully attacked its quality rather than accepting the premise that its offensive character made it great. Calling the work bad rather than wicked supported the case against public funding. Through these examples, he argues that competence and morality are distinct dimensions that public debate often confuses.

Isaacson brings the discussion back to an actual investment decision: how highly does Thiel rank a founder’s moral compass? Thiel says he ranks it highly. He adds that he dislikes backing companies that are merely businesses. He looks for an important purpose that exceeds short-term transactions, rather than a mechanism that simply collects a small payment whenever someone buys or sells something.

Musk’s ambition to reach Mars is Thiel’s example of such a transcendent purpose. He does not claim to know exactly where it belongs on a moral scale, but sees it as more than a conventional commercial objective. That sense of pursuing something larger than the immediate business is a quality he wants in builders, even though it does not by itself answer Isaacson’s ethical questions.

Isaacson asks what humans can uniquely do in an era of AI. After a pause, Thiel says the question about artificial intelligence is also a question about human identity. Language once appeared to offer a clear dividing line, as in Noam Chomsky’s account of a distinctively human ability. The arrival of modern language models made that answer less secure.

Thiel instead emphasizes consciousness. He describes himself as vaguely Cartesian and says he does not think AI is conscious or ever will be. Humans experience a sunset in a particular conscious way and are aware of having that experience. He does not offer this as a proven fact about machines, but as his belief about a dimension of human life that would be lost if humanity disappeared.

Consciousness may not be the only reason to preserve humanity, Thiel adds. Girard also directs attention to relationships: people are not merely isolated minds contemplating themselves. Family life, companies, and the larger society give human existence a relational character. AI may change those relationships, but he hopes it can improve them rather than reduce human beings to replaceable processors of information.

Less time spent on repetitive labor could leave people more time for one another, Thiel suggests. In that future, relationships might become more important rather than less. Replacing every person with computers would, in his view, sacrifice both conscious human experience and these relational dimensions. He considers those losses severe, even in a world where machines could perform many of the same outward tasks.

Isaacson closes by thanking Thiel. Their conversation leaves the practical tension exposed: Thiel regards innovation as necessary for growth and human flourishing, while Isaacson keeps asking who can be trusted with its power. Thiel’s final answer to what should endure centers on human experience and relationships, not simply on the continuation of increasingly capable machines.