The new strategic resource in AI is speed

For years, the artificial intelligence race was defined by a simple question: who could build the most powerful model. Increasingly, a different question is emerging: who can move the fastest?

That reality was underscored this week when US President Donald Trump signed National Security Presidential Memorandum 11, a directive designed to accelerate the adoption of artificial intelligence across the American national security apparatus. The document seeks to reduce the gap between the pace of innovation in the private sector and the often slower processes of government agencies and the military.

At first glance, the memorandum appears to be an administrative effort focused on procurement and technology integration. Viewed alongside broader developments in the AI sector, however, it points to a more significant shift. Washington appears to recognize that the next phase of AI competition will not be defined solely by who develops the most advanced technology, but by who can absorb, deploy, and continuously update new capabilities faster than rivals.

The issue emerges clearly from a recent paper published by Anthropic. The company argues that Claude now contributes to the majority of the code integrated into its systems and that AI tools are significantly increasing research and engineering productivity. Whether recursive self-improvement eventually becomes reality remains a matter of debate. What is already visible is something more immediate: artificial intelligence is increasingly being used to accelerate the development of more artificial intelligence.

This changes the nature of technological competition. For years, discussions focused primarily on model capabilities. The central question was who could build the most advanced system. Today, another variable is gaining importance: the speed of the innovation cycle itself. If AI systems can help write code, identify errors, test solutions, and support researchers, the advantages accumulated by leading laboratories can compound far more rapidly than before.

In such an environment, speed becomes a strategic resource. It matters inside research laboratories developing next-generation models. It matters for companies building digital infrastructure. And it matters for governments attempting to translate technological advances into national power.

Trump’s memorandum can be understood through precisely this lens. Many of its provisions address the challenge of innovation absorption. The directive encourages the adoption of solutions from multiple vendors, promotes access to secure computing resources, strengthens testing environments, and seeks to deepen cooperation between private industry and federal agencies.

Underlying these measures is a clear concern: preventing the machinery of government from moving at a pace that is incompatible with the technology ecosystem driving innovation.

The memorandum also addresses questions of operational control. It states that AI capabilities used by American warfighters cannot be disabled or degraded without prior authorization. The provision reflects the growing integration between commercial technologies and national security functions. As privately developed software, models, and infrastructure become increasingly central to military operations, Washington is seeking to clarify who ultimately exercises authority over capabilities considered essential to national defense.

The other side of this competition can be observed thousands of miles away. During the Steppe Partner 2026 exercises, China showcased a growing integration of soldiers, drones, autonomous platforms, and AI-assisted command structures, as highlighted by Decode39. The demonstration fits within Beijing’s broader vision of “intelligentized warfare,” under which sensors, autonomous systems, software, and human decision-makers operate as components of a single operational architecture.

The significance lies less in any individual platform displayed during the exercises than in the apparent effort to rapidly incorporate emerging technologies into military planning and operational concepts. In many ways, this mirrors the logic behind the American memorandum, even if the political and institutional environments are profoundly different.

The US-China rivalry is often portrayed as a race to develop the most powerful model or the most advanced semiconductor. Those factors remain critical. Yet other variables are becoming increasingly important: the ability to coordinate research, industry, digital infrastructure, and state institutions; access to energy and computing power; and the speed with which new technologies can be tested, adapted, and deployed.

Viewed this way, speed is not simply about technological acceleration. It becomes an organizational capability. It depends on reducing bureaucratic friction, mobilizing capital, attracting talent, building data centers, updating procedures, and integrating innovations generated outside traditional institutions.

This challenge affects both democracies and more centralized systems. In the United States, the problem is aligning government processes with the pace of frontier technology companies. In China, despite the central role of the Communist Party, much of the country’s innovation still emerges from a highly competitive ecosystem in which companies, local governments, and research institutions compete for resources and advantages.

This is why Anthropic’s paper deserves attention beyond debates about artificial general intelligence or recursive self-improvement. It suggests that the competition may already be entering a phase in which the decisive factor is not a single technological breakthrough. Instead, success may depend on the ability to sustain a continuous process of innovation and rapidly incorporate its results into operational capabilities.

The race for artificial intelligence is still often described as a competition to build the best model. Increasingly, it looks more like a race against time.

The countries, institutions, and companies capable of compressing the distance between innovation and practical application will enjoy advantages that extend far beyond the technology sector. In the emerging AI era, innovation itself may prove less decisive than the ability to operationalize it. The distribution of speed could ultimately shape the distribution of power.