We are hurtling forward with no map
Two simultaneous races — one among American tech giants, one between the United States and China — are reshaping the foundations of global power through artificial intelligence, yet neither is governed by the kind of strategic doctrine that once kept nuclear rivalry from becoming catastrophe. Historian Niall Ferguson's warning echoes Kissinger's Cold War insight: that a competition without shared red lines, verification mechanisms, or credible deterrence is not a race anyone can win — only one everyone can lose. The technology is accelerating faster than human wisdom is assembling to meet it, and the leaders steering this transition range from the visionary to the reckless.
- Two AI races — domestic and geopolitical — are accelerating simultaneously with no meaningful regulatory framework to slow or shape them.
- The people making the highest-stakes decisions include egomaniacal executives, a reality-television president, and a Marxist-Leninist autocrat, raising the probability of catastrophic miscalculation.
- A Trump executive order requiring thirty-day government review of powerful AI models before release was quietly scaled back and applies only to select companies under select conditions.
- Anthropic's near-trillion-dollar IPO filing signals that capital markets are moving far faster than any policy apparatus attempting to govern the technology.
- Without a strategic doctrine analogous to Cold War nuclear theory — red lines, verification, shared risk language — the competition has no built-in braking mechanism.
- The danger is not malice but rationality: each actor, responding logically to incentives in a system without guardrails, accelerates the very instability no one wants.
The question defining the next decade is whether artificial intelligence represents humanity's greatest gift or its most consequential gamble. The answer, historian Niall Ferguson suggests, hinges on whether anyone is actually managing the competition — and right now, no one is.
Two races are unfolding at once. Domestically, five American companies — Anthropic, Google, Meta, OpenAI, and xAI — are sprinting to build the most powerful AI systems. Geopolitically, the United States and China are competing through those same corporate champions for dominance over a technology that will reshape military power, economics, and global order. Both races are accelerating. Neither has a governing framework.
The Cold War parallel is instructive. When Henry Kissinger published his 1957 analysis of nuclear strategy, he gave policymakers a language for navigating existential competition without triggering the catastrophe they were trying to prevent. That kind of strategic doctrine — imperfect but essential — is precisely what the AI age lacks. What exists instead is a cast of decision-makers of wildly uneven quality: tech executives with records of deception, at least two by any measure egomaniacs, a U.S. president whose statements run roughly half bluff, and a Chinese leader who sees himself as Mao's successor.
The material constraints on AI development are real — vast data, enormous compute, trillions in capital — but the constraint that should matter most is absent. There is no regulatory doctrine, no mechanism for verification, no shared understanding of red lines, no agreement on what constitutes unacceptable risk.
President Trump recently signed an executive order requiring government review of powerful AI models thirty days before public release — a scaled-back measure applying only to some companies in some circumstances. The same week, Anthropic filed for an IPO that could value it near a trillion dollars. The market is moving faster than policy. The technology is outpacing our capacity to think about its consequences.
This is how arms races become dangerous — not through evil, but through rational actors responding to incentives in a system with no brakes. The nuclear age eventually produced thinkers who understood that some competitions could not be won, only managed. The AI age has not yet found its Kissinger. Until it does, civilization is navigating unprecedented territory with mixed-competence leadership and no map.
The question hanging over the next decade is deceptively simple: Is artificial intelligence humanity's greatest gift or its most consequential gamble? The answer may depend entirely on whether we can find a way to manage what is unfolding as the most perilous competition in recorded history—and right now, we are managing it not at all.
Two races are happening simultaneously. The first is domestic: five American companies—Anthropic, Google, Meta, OpenAI, and xAI—are locked in a sprint to build the most powerful AI systems. The second is geopolitical. The United States and China are competing through their own corporate champions to achieve dominance in a technology that will reshape economics, military capability, and global power. Both races are accelerating. Both are largely unregulated. Neither has the kind of strategic framework that once kept the world from nuclear annihilation.
The parallel to the Cold War is not casual. In 1957, Henry Kissinger published a book that became foundational to how the West thought about nuclear weapons. Nuclear Weapons and Foreign Policy identified the central trap of the atomic age: that any strategy built on the threat of nuclear war created such catastrophic risk that it became incredible—no rational actor could actually follow through. The book was flawed, but it was essential. It gave policymakers and strategists a language for thinking about existential competition without destroying civilization in the process. Seventy years later, as AI accelerates and tensions mount, we need that kind of thinking again. We do not have it.
What we have instead is a collection of leaders of wildly uneven quality making decisions that will shape the future. Some of the chief executives running the most important AI companies have track records of deception. At least two are, by any reasonable measure, egomaniacs. The President of the United States is a former real estate developer and television personality whose public statements are roughly half bluff. The leader of China is a Marxist-Leninist autocrat who sees himself as a successor to Mao. These are the people steering us through the most consequential technological transition in human history.
The competition itself has real constraints. Training modern AI systems requires vast amounts of data, enormous computational power, and trillions of dollars in capital investment. These are not infinite resources. But there is one constraint that should matter most and does not: regulation. There is no meaningful regulatory framework governing how these systems are developed, tested, or deployed. There is no doctrine for what happens when one side believes it is falling behind. There is no mechanism for verification, no agreement on red lines, no shared understanding of what constitutes an unacceptable risk.
On Tuesday, President Trump signed an executive order requiring some AI companies to allow the government to review powerful new models thirty days before public release. It was a scaled-back version of an earlier proposal, and it applies only to some companies, in some circumstances. Meanwhile, Anthropic filed paperwork for an initial public offering that could value the company at nearly a trillion dollars. Other major AI firms are preparing to follow. The market is moving faster than policy. The technology is advancing faster than our ability to think about its consequences.
This is how arms races become dangerous. Not because the competitors are evil, but because they are rational actors responding to incentives in a system with no brakes. Each side believes it must move faster, invest more, take greater risks, because the alternative—falling behind—is unacceptable. The nuclear age had Kissinger and others who understood that some competitions could not be won, only managed. The AI age has not yet produced that kind of thinking. Until it does, we are hurtling forward into territory no civilization has ever navigated, with leaders of mixed competence and no map.
Notable Quotes
We desperately need someone like Kissinger to write Artificial Intelligence and Global Security— Niall Ferguson
The Hearth Conversation Another angle on the story
When you say this is the most dangerous arms race in history, are you comparing it to nuclear weapons, or saying it's actually worse?
Both. The nuclear race had something AI competition lacks: a shared understanding that mutual destruction was possible and unacceptable. That created a kind of dark stability. With AI, we don't even have that. We don't know what the endpoint looks like, or whether one side can actually "win" in a way that matters.
But surely there's more regulation coming. Governments are paying attention now.
They are, but they're moving at the speed of government while the technology moves at the speed of capital markets. By the time a meaningful framework exists, the competitive dynamics may have already locked everyone into a path that's hard to reverse.
You mention the quality of leadership. Does that actually matter if the incentives are what they are?
It matters enormously. In the Cold War, you had people like Kissinger and Kennedy who understood the stakes and could think strategically. Now you have egomaniacs and a real estate developer. When the pressure gets real, judgment matters.
What would a Kissinger-style solution look like for AI?
Probably something like verification protocols, agreed-upon safety standards, and a shared doctrine about what kinds of AI development are acceptable. But you'd need both sides to believe they have more to gain from stability than from racing ahead. Right now, neither side believes that.
Is there any sign that's changing?
Not yet. The executive order Trump signed is a start, but it's voluntary and limited. It's not a strategic doctrine. It's a gesture.