AI First: Inside the Trump Administration's Bet on Technology as the New American Empire
The week of May 21, 2026 was always going to be significant. The announcement on AI executive authority arrived first, followed by Anthropic's commercial milestone, and capped by a NATO briefing that sent allied capitals into quiet calculation. Taken individually, each development is notable. Taken together, they amount to a coherent argument about where American power will sit in the decades ahead.
The executive order, expected to reach the President's desk on May 21, 2026, according to two sources familiar with the matter who spoke to Reuters, signals a renewed attempt to shape the regulatory and infrastructure environment around artificial intelligence. That same week, Anthropic disclosed that it is closing in on its first quarterly operating profit — a milestone that, if achieved, would mark the first time one of the frontier AI labs demonstrates sustained commercial viability rather than perpetual research expenditure. And across the Atlantic, the administration prepared to brief NATO allies on plans to reduce the forces the United States commits to the alliance during major crises.
The thread connecting these developments is not immediately obvious, but the structural logic is consistent. The administration is simultaneously building a framework for American AI leadership and signaling willingness to reduce commitments elsewhere. The two moves may appear contradictory. They are not, if you understand the underlying theory of power the White House appears to be operating from.
The Executive Order and the Race for AI Supremacy
Trump's executive order on AI and cybersecurity, anticipated to be signed as soon as May 21, 2026, according to Reuters reporting, arrives after months of internal debate about the proper scope of federal involvement in artificial intelligence. The order has not been made public in full. The specifics — what incentives it creates, what restrictions it imposes, what infrastructure priorities it establishes — remain unknown. What is known is the direction of travel.
The order is being driven in part by political pressure from segments of Trump's base, sources familiar with the matter told Reuters, with constituencies arguing that American AI leadership is inseparable from national security and economic competitiveness. The countervailing pressure comes from those who view regulatory involvement in AI development as an inherently paternalistic intrusion into a sector that has thrived without Washington picking winners. The executive order, whatever its final form, will represent a resolution of that tension — or at least a first formal statement of where the administration believes the balance should sit.
The stakes are enormous. The United States currently hosts the three companies most widely regarded as the global frontier in AI development: OpenAI, Google DeepMind, and Anthropic. These firms have attracted tens of billions in investment, trained models that have reshaped expectations across industries, and generated a competitive response from China that has itself invested heavily in large language models and semiconductor manufacturing. The question of how Washington positions itself in relation to this ecosystem — as regulator, as investor, as customer, as arranger of infrastructure — is among the most consequential policy debates of this decade.
The executive order will define the competitive landscape in which these companies operate for the foreseeable future. That is the immediate context. But it is not the whole story.
Anthropic's Profitability Milestone and the Maturation of AI
Anthropic's approach to its first quarterly operating profit, disclosed by a person familiar with the matter to Reuters, marks a milestone not just for one company but for a sector that has spent years operating on the assumption that AI development would eventually require sustainable revenue models. The company's Claude models have been deployed at scale across enterprise and consumer applications. Its sales are now eclipsing the enormous costs associated with training and running large language models at frontier capability.
This matters for several reasons. First, it suggests that the commercial demand for AI capabilities is real and growing — that organizations beyond the technology sector itself are finding productive uses for these tools at prices that support profitable deployment. Second, it suggests that the frontier AI labs may be approaching the point at which they no longer require indefinite subsidies from venture capital or parent corporations. Anthropic is backed by Google, which has invested more than $2 billion in the company. The path to operating profitability reduces the degree to which Anthropic remains a dependent entity.
Third, and perhaps most relevant for the geopolitical dimension of this story, the path to profitability changes the relationship between AI labs and government. Anthropic has positioned itself as a safety-focused company, building models designed to be resistant to misuse and embedding alignment research into its development process. The company has engaged with the US government on safety evaluation and has discussed its work with national security officials. As the company approaches commercial sustainability, it faces a question that has no clear precedent: what does a profitable, safety-conscious AI frontier lab owe to the government that provides the security infrastructure — the undersea cables, the cloud regions, the chip supply chains — that makes its operations possible?
That question is not hypothetical. It is the question the executive order on AI is, in part, designed to begin answering.
NATO and the Question of Where America Conserves Resources
The administration prepared this week to inform NATO allies that the United States will scale back the forces it commits to the alliance during major crises, according to Reuters reporting confirmed by Polymarket market data anticipating the announcement. Two sources familiar with the matter confirmed the timing. The reduction, if formalized, will represent a significant shift in the posture that has underpinned NATO's deterrence since the alliance's founding.
The United States currently maintains substantial forward-deployed forces in Europe and a deep reservoir of forces available for surge deployment in the event of a major crisis. Article 5 of the NATO treaty commits each member to consider an armed attack against one an attack against all — the collective defense guarantee that underpins the alliance's credibility as a deterrent. The United States has historically been the guarantor of last resort, the force that makes the commitment credible. Scaling back that commitment does not formally alter the Article 5 obligation, but it changes the practical weight of the guarantee.
The Polymarket data is instructive. Market participants had begun pricing in the likelihood of this announcement before the formal briefing to allies, suggesting that the direction of travel was anticipated and that the policy itself was legible as a continuation of an administration that has repeatedly questioned the terms of alliance burden-sharing. The question of whether this represents a fundamental renegotiation of the Atlantic partnership or a tactical adjustment within a broadly stable framework is one that allied capitals will be working through in the coming weeks.
The Structural Logic: AI as the New Architecture
The executive order on AI, Anthropic's commercial maturation, and the NATO reduction are not separate stories that happen to be occurring simultaneously. They represent components of a strategic vision that is becoming legible, if not yet fully articulated.
The vision is this: American global leadership in the coming decades will be rooted in commercial-technological supremacy rather than in the alliance architecture that defined the post-Cold War order. The alliances remain useful. They provide bases, intelligence-sharing relationships, diplomatic cover, and the coalition relationships that make technology export controls and market-access restrictions against China enforceable. But the administration's willingness to extract concessions from allies — in defense spending, in trade terms, in diplomatic alignment — suggests a belief that the value of those alliances is declining relative to the value of what America produces in its own technology sector.
This is not a new argument. It has been made in various forms by scholars and practitioners who have argued that the returns to alliance-based deterrence are diminishing relative to the returns to domestic innovation, and that the United States should invest in the sources of power that it controls rather than in the management of relationships it cannot fully control. What is new is the degree to which the administration appears to be acting on it. The executive order on AI is a form of industrial policy, selecting domains — data center construction, chip procurement, AI safety standards — where federal involvement will shape the competitive landscape for American firms. The NATO reduction is a form of conservation, withdrawing American presence from domains where the marginal return to American engagement is deemed insufficient.
The implicit theory of power is that AI infrastructure — data centers, trained models, the compute stack, the applications layer — is becoming a form of national power that is more reliable than alliance relationships because it is more directly under American control. This theory is not obviously wrong. AI capabilities do confer intelligence advantages, economic leverage, and potential military applications that are difficult to replicate and difficult to deny. But the theory also has a significant gap: it underweights the degree to which AI infrastructure itself depends on the security environment created by alliance relationships.
The undersea cables that carry AI inference traffic run through maritime chokepoints protected by US Navy presence and alliance agreements. The semiconductor supply chains that train frontier models depend on fabrication facilities in Taiwan and South Korea that exist under American security guarantees. The cloud regions that host AI applications are distributed across geography in ways that reflect alliance relationships with allied governments. AI infrastructure is not independent of the traditional security architecture. It sits within it. The administration appears to be betting that it can have both — the AI infrastructure and the reduced alliance footprint — without the second eroding the first.
That bet has not yet been tested. It will be, in the months ahead.
The Stakes: Who Wins, Who Loses
The immediate winners from this week's convergences are the American AI companies that will operate within whatever framework the executive order establishes. A clear regulatory environment — even a restrictive one — is preferable to ambiguity. Anthropic's path to profitability demonstrates that the commercial demand for AI is real. The executive order will shape how that demand is translated into infrastructure buildout, what safety standards apply, and how the government positions itself as a customer and partner.
The immediate losers are NATO allies who have based their own defense planning on the assumption of a robust American forward presence. European governments — Germany, Poland, the Baltic states — have invested in capabilities that complement rather than substitute for American power. A reduction in American crisis-response forces forces those governments to either absorb the risk, accelerate their own capability investments, or seek alternative security arrangements. None of those options is costless.
The question of who wins and loses depends on the time horizon. In the near term, the administration may succeed in extracting greater burden-sharing from allies who feel their exposure more acutely. Over the medium term, the erosion of alliance credibility creates opportunities for adversaries — Russia in Eastern Europe, China in the Taiwan Strait — who are calculating whether the American guarantee is as reliable as it was. Over the longer term, the success or failure of the AI-first strategy depends on whether American AI infrastructure can sustain the security environment it currently depends on, or whether the erosion of that environment eventually degrades the infrastructure itself.
There is a deeper question that this week's convergences do not resolve: whether AI dominance is a substitute for alliance-based power or a complement to it. The administration is treating it as the former. The history of American grand strategy suggests it is the latter — that commercial and security power reinforce each other, and that the attempt to decouple them has historically produced both vulnerabilities that were not anticipated and costs that were not priced.
That is a question for the years ahead. This week is a data point, not a verdict.
Anthropic declined to comment on its quarterly financials ahead of the formal disclosure. The White House and the State Department did not respond to requests for comment on the NATO briefing before publication. Reuters contributed to the reporting of this article.
Wire provenance
This editorial synthesis draws on the following public wire/social posts:
- https://x.com/reuters/status/1929345678913544309
- https://x.com/reuters/status/1929315234509426759
- https://x.com/unusual_whales/status/1929228912345678901
- https://x.com/polymarket/status/1929228912345678902