Without Clear Patent Rules, the U.S. Loses the AI Race Before It Even Starts
There is a paradox that few in Washington want to voice publicly: the United States is investing tens of billions of dollars in artificial intelligence infrastructure—semiconductors, data centers, research subsidies—while leaving unresolved the question that matters most to the private capital that must fund the application layer. Who owns what AI produces, improves, or discovers? As long as that question remains without a clear institutional answer, language models and optimization algorithms will continue to be a high-risk gamble for any company looking to bring them to market and recoup its investments. Infrastructure without legal certainty does not accelerate innovation—it stifles it at the most costly moment: when transitioning from proof of concept to commercial product.
This is the thesis that a former Trump administration official recently articulated in Fortune: the U.S. can win the AI race against China, but only if it resolves the patent policy issue. It is not a technical thesis. It is a thesis about governance, about how the rules of the game determine what game is played and who chooses to play it.
Capital Doesn’t Wait for Lawyers to Agree
The patent discussion in the context of AI is not new, but it has gained urgency because the private money that should finance applied artificial intelligence—the type that generates products, automates industrial processes, and improves medical diagnostics—is increasingly skeptical of the regulatory framework. Venture capital funds and institutional investors do not finance ideas; they finance return theses. And a return thesis in AI largely depends on the ability to protect what is built.
The current problem in the U.S. is structural. The U.S. Patent and Trademark Office (USPTO) operates with eligibility criteria for software and AI patents that have been generating contradictory rulings for over a decade. What one examiner approves, another rejects. What a district court protects, the Federal Circuit Court can invalidate. This level of unpredictability has a concrete cost: it raises the cost of litigation, increases the legal risk premium for any AI startup, and disincentivizes investment in early stages precisely where the most radical advancements occur.
Meanwhile, China operates under an intellectual property policy for AI that, while imperfect from the perspective of international standards, is deliberate and coherent with its industrial goals. The Chinese government has made it clear that it wants to be the largest filer of AI patents in the world and has consistently executed that agenda for years. Not because the Chinese system is better in technical or ethical terms, but because it has a readable direction. Economic actors—including Western companies operating there—can plan based on that. The American unpredictability, compared to that coherence, becomes a competitive disadvantage even though the talent, models, and infrastructure are superior.
What Chip Money Can’t Buy
Washington's bet on infrastructure is understandable. Chips are tangible, photo-ready, and politically communicable. A semiconductor factory in Arizona creates visible jobs, ribbon cuttings, and headlines. Patent policy, on the other hand, is abstract, technical, and electorally thankless. But that asymmetry of political visibility does not change the underlying economic mechanics.
The layer of infrastructure—chips, networks, energy—is a necessary condition but not sufficient for the U.S. to maintain leadership in applied AI. What converts infrastructure into a sustained competitive advantage is the application layer: the thousands of companies that take foundational models and integrate them into health products, logistics, manufacturing, and financial services. That layer requires massive private investment. And that private investment needs certainty about returns, which in turn requires clarity on what can be protected and how.
Here lies the knot that public debate rarely articulates precisely: if a company develops a substantial algorithmic improvement on a foundational model, designs a novel training process, or creates a more efficient inference architecture for a specific domain, the question of whether that is patentable in the U.S. does not today have a predictable answer. Specialized lawyers charge very high fees precisely for navigating that ambiguity. Startups without access to those lawyers simply assume the risk or, more frequently, seek jurisdictions where the rules are more legible.
Capital is not patriotic. It follows the rules of the game where rules exist.
The Arrogance of Believing Technological Advantage Sustains Itself
There is an organizational pattern I recognize in any company that assumes its competitive advantage is so solid that it does not need institutional infrastructure to support it. It is the same pattern that leads brilliant executives to ignore early signals of deterioration because they are overly confident in what they built yesterday. The U.S. has been operating with that logic concerning its technological leadership for years: talent, universities, venture capital, and innovation culture are so superior that the rules of the game can remain outdated indefinitely.
That confidence made sense when the adversary was diffuse. It no longer holds when China has demonstrated the ability to execute long-term industrial agendas with a discipline that Western democracies rarely sustain through electoral cycles. It is not that China will win because it is better; it is that the U.S. can lose because it assumes it does not need to actively manage its advantages.
Patent policy is precisely the type of variable that leaders ignore because its impact is neither immediate nor spectacular. The cost of not resolving it does not appear in one quarter; it appears in five years, when the capital applied to AI has concentrated in companies that found more predictable regulatory environments, or when American inventors begin to register their patents in other jurisdictions because the domestic system generates more uncertainty than certainty.
The conversation Washington needs to have about intellectual property and AI has been postponed for years because it is technically complex, politically dry, and does not generate the type of visibility that moves legislative agendas. But the culture of a nation competing in technology is not the result of its most publicized investments. It is the natural symptom of all the difficult institutional conversations its leaders had the courage to uphold until the end, and the inevitable reflection of all the ones they chose to ignore because the immediate political cost was too uncomfortable.
