The most important challenge facing American science is not whether the United States still leads. It does. The more consequential question is whether the institutions that produced that leadership are evolving quickly enough for a world in which artificial intelligence is changing how discovery happens, strategic technologies are reshaping economic and national security priorities, and scientific cooperation increasingly exists alongside technological competition.

That was the larger message I took from the 2026 State of the Science Address, delivered by outgoing National Academy of Sciences President Marcia McNutt at the National Academy of Sciences in Washington, D.C. The address offered a candid assessment of the U.S. research enterprise, but its significance extends well beyond science policy. It speaks directly to how the United States organizes discovery, develops talent, works with industry, uses government research capacity, and positions itself in a world where science and technology are now central instruments of power.

For readers who want the full address, the recording is available here.

The United States remains one of the world's leading scientific powers. It is home to many of the world's top universities, national laboratories, technology companies, research institutions, and frontier innovation ecosystems. It leads in key areas of artificial intelligence, biotechnology, advanced computing, space, energy, and quantum science. Yet leadership today is no longer secured by excellence alone. It increasingly depends on institutional agility.

The postwar research model that helped make American science dominant was designed for a different world. Federal investment supported universities. Universities produced discoveries and talent. Industry translated research into products and markets. International students and researchers came to the United States because the strongest opportunities were here. That model helped generate decades of prosperity, scientific breakthroughs, and technological leadership.

But the conditions around that model have changed. Other countries have built serious research capacity. Industry now performs and funds much of U.S. research and development. Scientific philanthropy has become a larger force. Talent is more mobile. AI is accelerating discovery and changing the infrastructure needed to conduct research. Strategic technologies are now directly linked to competitiveness, security, industrial policy, and diplomacy.

In that environment, the question is not whether the United States should preserve the old model. The question is what comes next.

AI Is Changing the Structure of Science

Artificial intelligence is not simply another tool in the laboratory. It is beginning to change the structure of scientific work itself.

AI can help researchers analyze massive datasets, search scientific literature, generate hypotheses, design experiments, model complex systems, support drug discovery, identify new materials, automate workflows, and accelerate the movement from data to insight. This changes not only the speed of research, but also the organization of research. The laboratory of the future may depend as much on compute, data infrastructure, robotics, shared platforms, and interdisciplinary teams as on the traditional individual investigator model.

This is why recent U.S. government initiatives matter. The National AI Research Resource is designed to expand access to AI tools, computing power, datasets, software, models, and expertise for researchers and educators. The Genesis Mission points in an even more ambitious direction by seeking to connect federal scientific datasets, national laboratories, supercomputing resources, AI agents, and automated workflows to accelerate scientific discovery. The broader U.S. AI Action Plan also places AI infrastructure, innovation, and international technology leadership at the center of national strategy.

Taken together, these initiatives suggest that the federal government is beginning to treat AI not only as a technology sector, but as scientific infrastructure. That is a major shift. If AI becomes part of the operating system of science, then scientific leadership will increasingly depend on who can connect data, compute, talent, laboratories, industry, and policy most effectively.

This is where universities, government research institutions, and industry will need to work differently. Universities remain essential for talent and foundational research. National laboratories bring mission-driven scientific infrastructure, high-performance computing, and long-term public research capacity. Industry brings scale, platforms, capital, deployment pathways, and real-world technical challenges. AI-enabled science will require all three to operate less like separate systems and more like connected parts of a national innovation architecture.

The New Competition Is About Ecosystems

The address highlighted several examples of university-industry collaboration, including models at the University of Washington, Clemson University, and Purdue University. These examples are important, but not because every institution should copy them exactly. Their broader significance is that they point to a shift from isolated research activity toward ecosystem competition.

The countries and regions that lead in strategic technologies will not simply be those with the strongest individual universities or the largest companies. They will be those that can connect universities, government laboratories, industry, capital, infrastructure, talent pipelines, and trusted international partnerships into a functioning innovation system.

This is already visible in AI. Industry is driving much of the frontier model development, but universities remain essential for training talent, advancing foundational research, and sustaining open inquiry. Government remains essential for national research infrastructure, public-interest science, standards, security, and long-term investment in areas where markets alone may not move fast enough or broadly enough. The same pattern is emerging in quantum technologies, biotechnology, advanced manufacturing, semiconductors, energy systems, and space.

This changes how we should think about competitiveness. It is not only a race for discoveries. It is a race to build systems that can turn discovery into capability.

Industry Is Changing What Scientific Talent Needs to Know

One of the strongest workforce messages from the address was that most STEM graduates who work in STEM fields are employed outside academia. Yet many graduate programs are still structured around academic career pathways that only a small share of students will ultimately follow.

This mismatch matters more in the AI era.

Industry increasingly needs scientists and engineers who can work across disciplines, use AI tools, understand data systems, communicate with non-specialists, manage projects, navigate regulatory environments, and connect research to deployment. Technical depth remains essential, but it is no longer sufficient on its own.

Programs such as Keck Graduate Institute, Northeastern's cooperative education model, and Lehigh's Pasteur Partners Ph.D. illustrate a broader shift toward education models that integrate scientific training with industry experience, professional skills, and applied problem-solving. The point is not to make universities subordinate to industry. It is to recognize that the scientific workforce now moves through a much wider set of institutions than the traditional academic pipeline.

For the United States, this is also a science diplomacy issue. If domestic students are discouraged from STEM careers, if international students choose opportunities elsewhere, or if companies cannot find AI-ready talent, the consequences will affect innovation, investment, and national capability. Talent attraction is no longer a background condition of scientific leadership. It is part of the strategy.

Government Research Must Also Adapt

The discussion should not focus only on universities. Government research capacity is also central to the next phase of American scientific leadership.

The national laboratories, federal science agencies, and mission-oriented research programs have long played a decisive role in areas such as energy, defense, health, space, climate, and advanced computing. In an AI-enabled research environment, that role may become even more important because government holds unique assets: large-scale scientific datasets, public research infrastructure, supercomputing capacity, long-term mission focus, and convening power across sectors.

Genesis is significant in this context because it suggests a more integrated model of federal science. Rather than treating datasets, laboratories, agencies, and computing resources as separate assets, the initiative points toward a research architecture in which AI can connect experimental facilities, national labs, domain experts, and automated workflows around national priorities.

It could also reshape how the United States approaches international science partnerships. If federal scientific data and AI-enabled research platforms become more integrated, future partnerships may increasingly revolve around shared standards, trusted data access, interoperable infrastructure, and clearly defined rules for collaboration in sensitive technology areas. In other words, science diplomacy may move from supporting researcher-to-researcher exchange toward negotiating the conditions under which AI-enabled discovery systems can collaborate securely and productively across borders.

That shift matters. Partners will want access to U.S. scientific infrastructure, datasets, and AI-enabled research capabilities, while the United States will need to decide where openness advances discovery and where safeguards are necessary. Genesis could therefore become not only a domestic research initiative, but also a signal of how the U.S. intends to organize scientific leadership in an AI-driven era.

But adaptation also requires reducing unnecessary friction. McNutt emphasized the growing administrative and regulatory burden on researchers, including the time spent on compliance and reporting. Oversight is necessary, especially in areas involving public funds, security, human subjects, and research integrity. But when compliance systems become fragmented, duplicative, or poorly aligned with risk, they slow the very research they are meant to support.

AI may help here as well. Used responsibly, it could reduce repetitive reporting, harmonize documentation, and improve compliance processes. But the larger issue is institutional design. A research system that wants to lead in fast-moving strategic technologies cannot afford to bury its researchers in avoidable administrative complexity.

Cooperation and Competition Will Coexist

The science diplomacy implications are clear. The United States will need international collaboration to address global challenges, but it will also need to compete in strategic technologies that shape economic power and national security.

These two realities are no longer separate. The same countries may collaborate on health research while competing in AI. The same partners may share climate data while competing in clean energy manufacturing. The same universities that depend on international talent may also face research security concerns. The same companies that build global research networks may operate in markets shaped by export controls, standards competition, and geopolitical risk.

This is the new operating environment for science diplomacy. It is not cooperation or competition. It is cooperation and competition at the same time.

That requires more sophisticated institutions. Universities need clearer strategies for international engagement and research security. Governments need frameworks that protect critical technologies without undermining openness. Industry needs trusted pathways for collaboration, talent development, and responsible deployment. Diplomats need deeper scientific and technological literacy. Scientists need a better understanding of the policy and geopolitical environments in which their work now operates.

The Real Question Is Institutional Adaptation

The most important insight from the address is not that American science is weak. It is that the system around American science must evolve.

The pilot programs and examples McNutt highlighted point in the same direction. Universities are experimenting with new models of industry engagement. Promotion and tenure systems are beginning to recognize team science, public impact, and interdisciplinary work. Graduate programs are being redesigned around workforce realities. Shared and automated research platforms are emerging. Federal initiatives are beginning to treat AI as core research infrastructure. High-risk, high-reward models such as DARPA, ARPA-E, and ARPA-H remain important examples of how government can support ambitious science.

The challenge now is not identifying reforms. It is scaling them. America's advantage has never rested solely on the size of its economy or the scale of its research investments. Its greatest strength has been the ability to build and rebuild institutions when circumstances change. The postwar research model was one such institutional achievement. The AI era may require another.

For science diplomacy practitioners, this moment is especially important because scientific leadership now depends on more than research excellence. It depends on talent flows, trusted partnerships, research security, technology governance, public trust, industrial capacity, and the ability to manage cooperation and competition simultaneously.

The United States still leads in science. But leadership in the next era will depend on whether its universities, government research institutions, national laboratories, and industry partners can adapt to a world where AI is changing how discovery happens and strategic technologies are reshaping global power.

As National Academy of Sciences President Marcia McNutt concluded, innovation should not be America's response to crisis. It should be the baseline expectation of the U.S. research enterprise.

That is also the central task of science diplomacy in the AI era: helping institutions remain open enough to collaborate, strategic enough to compete, and adaptive enough to lead.

More analysis on science diplomacy and strategic infrastructure is available at glsd.ai, where ongoing work connects policy, innovation, and international partnerships through a global lens.

New Resource: The Science Diplomacy Playbook

For readers working across diplomacy, technology policy, research strategy, investment, or international partnerships, The Science Diplomacy Playbook: 2026 Edition is now available to order.

The guide expands on many of the themes covered in Global Signals, including AI governance, critical minerals diplomacy, bilateral science and technology agreements, minilateral coalitions, research security, and the growing role of science diplomacy in industrial strategy.

It is written as a practical field guide for decision makers navigating the intersection of science, technology, and global affairs.

Available as a paperback and kindle on Amazon, with ebook editions on Apple Books and Google Play. Expanded ebook and library distribution is rolling out across North America, the United Kingdom, and Europe, including through Tolino, Vivlio, Smashwords, Fable, Gardners, BorrowBox, and cloudLibrary.

Join Global Signals

A biweekly strategic brief and live briefing series translating science, technology, policy, and global capital flows into strategic insight for decision makers across diplomacy, investment, research, and innovation. Join here.