OPENING AMERICA’S LABORATORIES

Beyond test beds, America possesses extraordinary research infrastructure, built up over decades of federal investment. The DOE alone operates 28 user facilities, from the Advanced Photon Source at Argonne to the Spallation Neutron Source at Oak Ridge, providing capabilities available nowhere else on Earth.101 These facilities represent billions of dollars in capital investment by the American taxpayer and decades of accumulated expertise.

Yet access to this extraordinary infrastructure has historically been too narrow and too slow. Echoing a theme from Chapter II, evaluations for facility access are built for academic merit review. This system works for university scientists pursuing publications, but fails entrepreneurs who need to validate technology on a timeline set by competitors.

Opening these facilities more broadly to private industry, with evaluations that weigh innovative potential and commercial urgency alongside scientific merit, would multiply the return on existing federal investments. An older culture at the labs holds that industry engagement detracts from the research mission, but in reality these interactions benefit both sides, allowing external users to leverage the lab’s vast knowledge base while exposing lab researchers to new use-inspired problems. Revenue from user fees can also fund expanded capacity and new instrumentation, turning facilities that today operate below capacity into self-sustaining engines of innovation. Every facility-hour that goes unused is a wasted national asset; every dollar of industry revenue reinvested is a dollar of federal appropriation freed to grow the next generation of tools and equipment. Large federal facilities are only part of the picture. Closer to the entrepreneur, shared platforms at smaller scales have proven equally transformative.

The National Quantum and Nanotechnology Infrastructure program provides shared cleanroom access with more than 2,000 available tools, enabling startups to prototype semiconductor, photonics, and quantum devices without building their own fabrication lines, often at the cost of just a few hundred dollars per hour.102 In the life sciences, shared wet laboratories have reduced the capital barriers for early-stage biotech companies, enabling researchers to move from concept to experiment in weeks rather than the years required to build a dedicated facility.103 Shared Good Manufacturing Practice (GMP) facilities address an even larger bottleneck.104 The production of clinical-grade materials under FDA-compliant conditions requires tens of millions of dollars in capital investment that most startups cannot raise before they have clinical data. This has created a catch-22 that can be broken by shared GMP platforms. Expanding these platforms across sectors and geographies would put the physical tools of innovation within reach of any American entrepreneur with a good idea.

Our national laboratories and universities can play a similar role at larger scales, serving as revitalized hubs of testing and evaluation for private industry. Places like Oak Ridge, Sandia, and Lawrence Livermore possess unique capabilities to validate technologies no startup could test alone; universities that host startups in their research infrastructure catalyze knowledge and hiring pipelines that multiply innovation. Yet licensing and partnership processes at national labs remain slow relative to the pace at which technologies must move. American research universities face a parallel challenge. Intellectual property policies vary wildly across institutions, creating friction for companies that want to license from multiple universities. Faculty incentive structures typically reward publications over commercialization, and far too often, equipment purchased with federal grants sits idle between projects while entrepreneurs who could use it have no access. Reforms that streamline university technology licensing, standardize IP frameworks for federally funded research, and open university facilities to outside innovators on flexible terms would unlock a vast reservoir of capability that today remains bottled up behind administrative walls. Likewise, streamlining the CRADAs that govern lab-industry partnerships, further leveraging the OTA, and reducing the administrative burden on companies seeking to license lab technologies, would help our scientific institutions move closer to industry speed.

TAPPING OUR PRIVATE SECTOR

A major task ahead for the Federal Government is to harmonize the efforts of our publicly-funded institutions with our dynamic private sector. The way government funds science has not yet fully integrated the spectacular rise of the private sector in both basic and applied R&D. In the 1960s, the Federal Government funded over 70% of all basic research performed in the United States.105 Today the federal share of basic research funding has fallen to 40%, while industry’s share has grown to well over 35%. Our biggest technology companies and leading pharmaceutical firms now support or directly publish some of the most cited work in fundamental science. Individual technology companies now spend more on R&D than the NSF’s entire annual budget. In fields like AI, quantum computing, and advanced drug discovery, the most important research increasingly requires capabilities that universities alone cannot provide.

In AI, the companies that train frontier models wield supercomputers worth tens of billions of dollars, hold petabytes of proprietary data, and can afford to spend tens of millions to recruit the best engineering talent in ways no university can match. This has produced an academic brain drain; yet academic researchers remain essential to the long-term health of the field, producing foundational work on next-generation algorithms that companies have less incentive to pursue. Without new partnership structures that give academic scientists access to frontier-scale resources, the basic research that underpins the next generation of AI advances will atrophy, and our technological leadership will rest on an increasingly narrow institutional base. What is needed are mechanisms that adequately leverage the comparative advantage of both public and private funding.

CLOSER PARTNERSHIPS

As a first step, we need to expand the scope of Federal grantmaking. Funding should be open to new types of institutions, whether they are joint industry-university centers or independent research organizations that can raise equity. Some mechanisms already exist but are underused. As discussed in Chapter II, the OTA can surmount procurement constraints, and institution-agnostic grants can reach nonprofits, industry consortia, and independent researchers. SBIR and STTR programs can be deployed strategically to advance new scientific and technological capabilities, coupling federally-seeded companies with the scientific enterprise. Furthermore, our science agencies could establish or strengthen agency-adjacent independent foundations, modeled on the Foundation for the NIH (FNIH).106 One FNIH public-private partnership involving NIH and industry partners, the Accelerating Medicines Partnership (AMP), invests in reducing the timeline to find live-saving therapies and improvements in outcomes. The AMP on Alzheimer’s Disease, one of twelve disease-focused AMPs, experimentally validated 20 candidate drug targets to lead to clinical trial success.107 Such foundations can blend public and private capital in ways that federal procurement rules prohibit, contract on commercial terms, and accept cost-sharing from industry partners, offering a vehicle for public-private collaboration that moves at the speed of industry while remaining responsive to policy priorities. The most powerful conduit between institutions, however, is the flow of human capital itself. Industry Ph.D. programs that enable American citizens to complete doctoral training at leading private organizations or national laboratories would offer higher stipends, real work experience, and exposure to problems at the frontier while also drawing a larger proportion of American citizens into basic research. Such programs already exist in prototype.108 These include industry postdocs, where some of our best researchers join leading companies to drive groundbreaking work, and industry-funded Ph.D. scholarships, which create revolving doors that bring new ideas into our strongest mathematics and physics departments. Programs like Activate at Lawrence Berkeley National Laboratory embed entrepreneurial scientists in national lab environments with stipends, lab access, and mentorship, and have been effectively expanded to talent emerging from the nation’s universities through investments by NSF.109 We can build on these models by creating more flexible cross-institutional talent pathways. Scaling such programs would widen the aperture for our researchers. Instead of being locked into a single institutional track, or forced into a risky, one-way jump into industry, our next generation should be able to move fluidly among a wide range of sectors, institutions, and research cultures.

MARSHALING GRAND EFFORTS

Reforming the bilateral partnership between our government and private companies is only the first step. Many of the most transformative technological achievements in history required the deliberate marshaling of national effort toward goals that no single company, university, or agency could achieve alone. The Human Genome Project is a case in point. It began as a federally directed NIH-DOE partnership in 1990, with initial funding in President Reagan’s 1988 budget submission.110 Its creators wagered that a complete reference of the human genome, and the sequencing technology advanced through the effort, would become a foundational technology for all of biomedicine. Its advocates pressed forward even as many leading biologists in the late 1980s dismissed the project as immature and argued the money would be better spent on individual grants. Only the Federal Government could have marshaled the coalition that executed it. A distributed network of DOE national laboratories and NIH-funded centers, including Washington University and the Whitehead Institute, coalesced around common milestones and operated under the Bermuda Principles, which required immediate public data release. The project depended on the productive entanglement of public and private capacities, most notably in the development of new automated capillary sequencers, where federal demand pulled forward private innovation in instrumentation.111 When Celera Genomics entered as a private competitor in 1998, the resulting public-private dynamic accelerated the timeline further.

The project finished ahead of schedule, and the $3.8 billion federal investment has generated an estimated$ 796 billion in economic activity.112 The human genome story exemplifies the Federal Government driving American leadership in a platform technology. It featured an engineering challenge bound up with a basic science mission; a gap in basic capabilities that could be closed through large-scale coordination; and a network of national laboratories and academic institutions focused on a common goal. Its completion required public-private collaboration that broke institutional walls.

PRE-COMPETITIVE CONSORTIA

The Federal Government wields enormous power to align fragmented actors around shared objectives. As discussed in Chapter II, well-designed grand challenges exemplify this convening power for problems with clear metrics and deadlines, where opening the field to outsiders is an advantage. But not all shared problems lend themselves to this approach. Some technical challenges sit between basic science and commercial application, too applied for academic grants and too risky for any single firm to tackle alone. These pre-competitive problems, shared across an industry, require another instrument: cooperative R&D anchored by federal investment. SEMATECH is the defining American example. By the 1980s, Japanese manufacturers had captured the majority of the global memory chip market. In 1987, fourteen American semiconductor companies pooled resources, matched by federal funding through DARPA, to attack shared manufacturing challenges in lithography, etching, and materials processing.113 The consortium solved technical problems that every American chipmaker needed but none could afford to solve individually.

The SEMATECH model produced an even more consequential successor. In 1997, EUV LLC, another semiconductor consortium, contracted with three DOE national laboratories to develop EUV lithography, a technology that required breakthroughs in plasma physics, precision optics, and materials science beyond the reach of any single firm. By 2001, the consortium had built the first prototype EUV exposure tool and filed over 150 patents.114 Refined over the following two decades, that technology now underpins every leading-edge semiconductor manufactured on Earth. While policies undertaken then ceded the dominant market position to a European company, EUV lithography remains one of the most strategically important industrial technologies in our part of the century, and it exists because American federal laboratories and semiconductor companies solved the problem together.115 The pre-competitive consortium model worked because it targeted the right problems where the science was understood, but the engineering had not yet been done, where shared technical risk was the barrier. Today, many challenges of a similar scale and complexity await, such as returning leading-edge semiconductor research to American soil, programming biological tissues with precision, and creating next-generation nanotechnology techniques that allow for self-replication and atomic-level manipulation. No single firm can tackle these problems, traditional academic grants cannot fund them, and the nation cannot afford to leave them to chance.

The Federal Government’s ability to anchor such ventures, leveraging private capital, aligning fragmented actors, and sustaining effort over timelines that no quarterly earnings cycle would tolerate, remains one of its most potent and underutilized capacities.

OUR NATIONAL CHARACTER

Alexis de Tocqueville observed nearly two centuries ago of America that “boldness of enterprise is the foremost cause of its rapid progress, its strength, and its greatness.” He marveled at a society where social station was not fixed by birth, where citizens formed voluntary associations to solve problems rather than waiting for direction from above, and where the frontier, both physical and intellectual, beckoned those willing to take risks.116 That culture persists. Americans believe that merit deserves an opportunity to show itself, that free inquiry produces truth, and that individuals who build useful things deserve reward.

That spirit is alive in our states and cities. America’s federal system affords us the chance to run many experiments simultaneously across jurisdictions, creating regulatory testbeds, distinct infrastructure, and tailored incentives. By letting companies operate robotaxis on public roads years before most states had written their rules, Arizona built itself into the nation’s leading testbed for autonomous vehicles.117 It has further leveraged that permissive environment to attract over $100 billion in semiconductor investment.118 Utah passed the nation’s first general regulatory sandbox in 2021.119 Wyoming enacted a series of laws tailored to blockchain and digital asset companies.120 States and cities that get regulatory frameworks right attract capital, talent, and industry; those that do not learn from those that do.

When our researchers recognized that scaling laws would transform language models, no government committee approved the decision to pursue it. When our engineers concluded that reusable rockets were possible, a deregulated space industry emerged that mobilized massive capital to land rocket stages. This pattern, of creating a vast reserve of scientific talent and knowledge, of permissionless innovation backed by patient capital and enabled by regulatory flexibility, represents America’s great competitive advantage. But in an era of foundational technologies and active rival states, permissionless innovation must be married to strategic purpose. The capacities described in this chapter, especially the ability to discover and test, to move from laboratory to demonstrated viability, are the mechanisms by which American boldness becomes American dominance. The Federal Government should actively encourage experimentation at every level, creating the conditions for more states, more cities, and more institutions to become laboratories. In a competition where early experimentation locks in trajectories, the nation running the most experiments holds the advantage. For more than 250 years, going back to 13 separate colonies, that has been the American way.

THE MARRIAGE OF SCIENCE AND CRAFT

In policy conversations, we often speak as though technology consists solely of intellectual property and gadgets. We focus on the patents that can be filed, the knowledge that can be written down, or the complex machines that can be built. But scientific and technological capability consists of much more than its most visible inputs and outputs. A better taxonomy holds that technology exists in three forms: tools, explicit instructions, and process knowledge.121 Consider chipmaking. The tools are the lithography machines, etchers, implanters, and more. The explicit instructions are the blueprints and recipes. But the process knowledge, like how to troubleshoot semiconductor yields, how complex variables affect wafer cleaning, how the next process node should be designed to balance performance and manufacturing risk, lives in the heads of experienced engineers and technicians. This tacit knowledge cannot be fully codified. Anyone can be placed in front of a piano bench with a score of Rachmaninoff, but playing it well requires a personal command of musical dynamics and tactile skill. As the chemist-turned-philosopher Michael Polanyi observed, “we can know more than we can tell.”122 A skilled welder knows things about metal behavior that no manual captures, like the way aluminum warns you before it warps, or the sound a good bead makes as it forms. A machinist develops intuitions about cutting tools that come only from years at the lathe. A pharmaceutical manufacturing technician recognizes subtle variations in chemical processes that determine whether a drug batch meets specifications.

This process knowledge, embodied in an experienced workforce, is the true keystone of technological capability. The same applies to the practice of science. When the sociologist Harry Collins studied laboratories attempting to replicate a new type of laser in the 1970s, he found that no scientist succeeded using published sources alone. Those who built working devices had all spent time in a laboratory with someone who had already done it. The knowledge to build the laser flowed through personal contact, often so subtle that scientists themselves could not fully articulate what they had learned.123

Likewise, a synthetic biologist improves through countless failed experiments while coaxing cells into expressing a novel protein. An immunologist, after years of experience and guidance from senior mentors, develops intuitions for which protocols will work with finicky cell lines, knowledge no methods section can capture. This is why many forms of scientific expertise require years of onthe-job training in working research organizations, and why academic publications alone remain insufficient to transmit the craft of science. Papers and patents are not the ultimate ends of progress, but way stations in the training of better scientists, engineers, and technicians. Science is not simply about the equipment, which any laboratory with enough capital can purchase, nor the instructions, which can be shared on a sheet of paper.124 Our true competitive advantage lies in the process knowledge embodied by America’s talent. Without skilled practitioners who pass their craft to those who follow, the engine stalls.

OUR MANUFACTURING BASE

America has endured a sustained period of deindustrialization. Manufacturing employment peaked at nearly 20 million workers in 1979;125 today it stands at roughly 13 million, a decline of around 35% even as the population has grown by more than 50%.126 Manufacturing’s share of total employment fell from nearly 22% in 1979 to around 8% today.127 This sustained decline has been compounded by the offshoring of contract research and development, especially in the pharmaceutical industry, where laboratories have moved abroad en masse. The fate of the American scientific enterprise is inseparable from the fate of American industry for two reasons. First, most of the economic returns from scientific discovery arise not at the moment of invention, but during the translation of new ideas into products that can be produced at scale. The returns lie in the work that follows invention: the engineering that makes designs manufacturable, the process refinements that bring costs down, the skilled workforce that operates advanced facilities, the supply chain relationships that enable scale. When technological translation moves abroad, so do the jobs, the expertise, and the capacity to produce the next generation of breakthroughs. Second, science and production continually inform one another. As discussed in Chapter I, the linear model in which basic research flows neatly to applied research, development, and production has always been a simplification.

Knowledge circulates not only between the theorist and the experimentalist, but between the experimentalist and the industrial sector as well. The scientist studying semiconductor physics learns from the manufacturing engineer wrestling with yields. The biologist designing a new therapeutic depends on the process chemist who can scale production. The roboticist developing a high-torque actuator benefits from the presence of a local precision manufacturing base, working where he can drive down the street to stand beside a machinist at the turning center to optimize the design in person.

These feedback loops depend on proximity between scientific research and industrial capability. Without local manufacturing capacity, the marriage of science and craft weakens. Scientists lose access to the practical problems that inspire new lines of inquiry and improve research quality, while industry loses the research ecosystem that sustains technological leadership. We need to reshore American manufacturing, not only for the sake of fruitful employment, but for the long-term health of American science itself.

VAST POTENTIAL REMAINS UNTAPPED

Even after decades of offshoring, the United States still nurtures some of the most dynamic trade communities in the world. Ours is a nation of tinkerers, hobbyists, and people who fix things with their hands. Countless Americans learn to repair cars from family members, to operate power tools, to build and maintain their own homes. From barn raisings on the frontier to hot rod culture in the 20th century to today’s maker movement, this do-it-yourself (DIY) spirit runs deep in our culture. Log onto any video sharing platform and you will find a country of builders. Amateur machinists demonstrate techniques for precision manufacturing. Hobbyist welders share tips on joining titanium. Electronics enthusiasts repair broken oscilloscopes. Amateur radio operators, biohackers, synthesizer builders, and drone constructors all participate in communities of shared technical knowledge.

Even highly specialized pursuits such as nuclear fusion and cyclotron construction have attracted dedicated experimenters. Public libraries increasingly offer 3D printers and laser cutters, while local shops provide courses in welding and machining. This grassroots engagement with technical work points to a vast reservoir of talent waiting to be cultivated. More than half of American adults participate in some form of making or building activity.128 The largest DIY conventions have attracted more than a hundred thousand participants.129 Manufacturing employment may have declined, but the cultural foundations of technical skill remain in the same population from which we once drew machinists, toolmakers, and engineers.

OUR EDUCATIONAL SYSTEM TILTED THE SCALES

Yet our formal educational and employment systems often fail to develop this potential. Over the past several decades, the expansion of college education has come at the expense of vocational training, and high schools that once taught machining classes have shifted resources toward college preparation. As factories closed and communities hollowed out, the message to young people that working with your hands is a fallback, not a calling, was clear. Success meant escaping physical work, not mastering it.

This cultural shift was reinforced by an economic transformation that treated physical labor as a commodity to be sourced wherever it was cheapest. This prejudice runs deep. In too many communities, politicians and guidance counselors have come to treat trade schools as consolation prizes for those not cut out for a four-year degree. Over the past two decades, shop class equipment from shuttered programs has flooded the used machinery market, tangible evidence of how thoroughly we abandoned hands-on education in our rush toward a so-called knowledge economy.130

The costs are now visible on both sides of the ledger. Millions of Americans have the aptitude and interest for technical work, but lack clear pathways to translate that interest into careers. Meanwhile, millions of skilled jobs are expected to be unfilled even as graduates enter the workforce.131 For decades, the Federal Government tilted the scales against career education that by extending unlimited loans to students attending colleges, universities, and graduate programs, driving up the cost of college and burying millions in debt.132

At the root of this policy failure was the government’s inability to recognize that the work of building, maintaining, and repairing the physical world is not a relic of the past, but the foundation of any future prosperity.

WE MUST RESTRUCTURE SCIENCE AS A BROADER ENDEAVOR

Too many of our universities and elite science and technology curricula have severed the connection between theory and craft. Engineering students study the theory of combustion, but few can disassemble and rebuild a combustion engine. Graduate programs reward theoretical contributions measured in citation counts, but not practical applications measured in jobs and dollars. The result is a generation of researchers who can model phenomena mathematically but cannot repair the apparatus in their own laboratories.

This narrowing departs from how science actually advances. As discussed in Chapter I, the linear model no longer holds in many fields, where discovery increasingly relies on feedback from the real world. Some of the most important breakthroughs in molecular biology and theoretical physics have come from scientists who understood their instruments intimately, who could not only design the experiments but build and modify the equipment themselves. Consider Rainer Weiss, who won the 2017 Nobel Prize in Physics for detecting gravitational waves. Weiss grew up scavenging war surplus electronics in New York, teaching himself to build ham radio transmitters and fixing broken devices for pocket money. After flunking out of MIT, he took a job as a laboratory technician, working alongside veteran craftsmen, learning to machine, solder, and weld. It was this training in what Weiss called “the art of improvisation in experimental science” that enabled him to design and build the prototype laser interferometer that became LIGO, the instrument that detected ripples in spacetime from colliding black holes a billion light-years away. As Weiss put it, “I’m a big believer in what’s called the apprentice system.”133 The divorce between scientific training and craft has held back America’s scientific progress. The tacit knowledge that powers our scientific enterprise, developed through experimental practice and hands-on technical training, will only become more important. The particle accelerators at our national laboratories need electrical engineers who can develop more powerful klystrons; fusion experiments need vacuum specialists and fabricators who can work with tungsten; telescopes that map the universe need craftsmen who grind mirrors to nanometer precision. These are the people who may well be core contributors to the next technological breakthrough. We must give STEM students at every level of study the opportunity for hands-on technical training, and conversely, ensure America’s skilled technical workforce has clear pathways to participate in formal academic training and scientific research. The glorification of craft, industry, and manufacturing that once characterized American culture must be revived, and we will be richer for it.

EXPANDING PARTICIPATION

Our model of scientific training must adapt in three ways: by incorporating technical training, breaking academic credentialism, and connecting grassroots learning to formal scientific research. First, we must reconnect university science and engineering programs with hands-on, practical knowledge. This requires integrating technical training into university curricula, breaking down barriers between “elite” and “vocational” schooling, and reforming accreditation to reward real-world technical work, which would include counting hands-on externships and registered apprenticeship hours toward accredited degrees. Those who set curricula across the country should think seriously about what a world of more abundant intelligence and more constrained craft knowledge means for the expertise most required by the next generation.

Second, we must create new pathways into our scientific enterprise grounded in demonstrated skill rather than academic pedigree alone. Today, the conventional academic ladder is the only widely legible route into research. We must lay alternative paths that allow more makers, crafters, and technicians to participate in academic training and scientific discovery if they choose to do so. National fellowships could place skilled machinists and lab technicians at national labs and research universities, with skill-based pathways to credentials and co-authorship on research outputs. Practitioners-in-residence programs, analogous to artists-in-residence, could embed experienced craftsmen alongside Ph.D. researchers, granting them access to specialized equipment, professional mentorship, and attention for their research without requiring a doctorate. National laboratories could develop portable, industry-recognized credentials in areas like cryogenics and advanced machining. Programs like SBIR open doors for technician-founded ventures, directing resources toward ideas that do not always start with a dissertation.

Third, we must connect the millions of Americans who tinker, fabricate, and repair to engineering and research. This is especially critical for rural students, who face distinct challenges in accessing traditional academic pathways. They are less likely to have family members working in STEM fields. Their schools receive less outreach from industry. The smaller populations in their towns make it harder to find like-minded peers. And yet many of our best scientists and engineers first learned to weld in a barn, or helped their families fix tractors growing up. The answer is not to pluck talented individuals from their communities, but to bring the frontier to them, creating more connection points between maker culture and our formal science enterprise.

INTEGRATED MODELS OF TRAINING

To sustain America’s scientific leadership, we must rebuild integrated communities from community colleges to extension systems to keep the frontier of science and technology open not only to a select few, but to every American who has the aptitude and interest to advance it. Community colleges are the natural foundation for scaling scientific and technical training. Congress has provided major funding to these institutions in recent years, and justifiably so. They enroll around 40% of all undergraduates and serve as the primary entry point to higher education for first-generation students, veterans, and working adults.134 They also provide apprentices with Related Technical Instruction in classrooms, pairing theory with on-the-job training.

With targeted support, community colleges can become regional hubs for scientific and technological innovation, integrating with critical industries, NSF Regional Innovation Engines, and local ecosystem initiatives. These partnerships can channel shared facility access, equipment donations, and industry-led, co-developed curricula into institutions that already train America’s best technical talent. The Trump Administration has made expanding apprenticeships a priority, directing federal agencies to reach and surpass one million active apprenticeships annually. The Department of Labor (DOL) has shifted toward a pay-for-performance model, replacing traditional upfront grants for employer-led apprenticeships with funding tied to apprentice hiring and retention.135 And with the passage of Workforce Pell in 2025, short-term training programs are, for the first time, eligible for Pell Grant funding.

Going further, we could extend registered apprenticeships into fields that have not traditionally used them, particularly in science and technology. These shifts move us toward a more practical system of education that restores integrated mentorship and embeds learning in real-world application.136 And to ensure scientific opportunity is not limited by geography, training must be distributed nationwide. Programs like the U.S. Department of Agriculture’s Cooperative Extension System could expand its remit to include critical sciences and technologies, creating strongholds in rural areas. Employers who most need skilled workers could partner with nearby community colleges by donating equipment and supporting instruction. These programs should serve as entry points into a broader regional innovation ecosystem, helping young talent network with mentors and peers who can channel their entrepreneurial energy into shared projects.

NEW PATHS FOR TRANSLATION

American technological leadership has repeatedly emerged from regional clusters where research and production were inseparable. In the early 20th century, Detroit became the world’s automotive capital not merely because of Ford’s factories, but because a dense ecosystem of suppliers, machinists, and engineering talent made the region uniquely capable of translating automotive innovations into mass production. Thousands of small suppliers could prototype and manufacture new components faster than anywhere else on earth.137 Such clusters work because they create dense connections among industrial production, scientific research, and a skilled technical workforce.138 The model persists today, where different regions specialize in distinct domains, such as optics in Arizona, biotechnology in the Boston corridor, aerospace in Colorado and Alabama, and advanced manufacturing in the Midwest. In these clusters, knowledge accumulates locally and is reinforced by a culture of practical problem-solving that emerges between the skilled technical workforce and dense supplier networks. As the United States offshored manufacturing, it also severed these connections and the process knowledge they sustain. For too long, America’s leaders assumed that we could retain high-value design work while ceding production. But as competitors began to manufacture at scale, they improved their ability to design and iterate. Vice President Vance captured this dynamic precisely, observing that over time, “the geographies that do the manufacturing get awfully good at the designing of things.”139 Recent federal initiatives have taken early steps toward reversing this trend and rebuilding America’s innovation clusters. The NSF’s Regional Innovation Engines and the Commerce Department’s Tech Hubs channel resources to stimulate place-based production ecosystems outside traditional coastal hubs. Programs such as the Manufacturing USA Institutes, NIST’s Manufacturing Extension Partnership, and DOW’s eight Microelectronic Commons regional hubs further incentivize industry partnerships, open shared R&D infrastructure, and drive industry-led workforce training. These clusters lower the barrier for small manufacturers to create new products, test applications of new technologies, and build a manufacturing workforce with the skills of the future. Whole-of-government industrial policy reinforces this shift, driving increased demand for American-made products and changing the calculus for where companies manufacture.

The Federal Government is using innovative economic tools to attract investment commitments from foreign governments, and so financing American production, training American workers, and building American supply chains.

That said, the Federal Government cannot build these ecosystems alone. State and local governments hold critical levers, including land use and permitting, education and workforce development, anchor institutions, and more, which enable them to cultivate competitive advantages. Done poorly, such competition can become a race to the bottom that transfers public money to mobile firms, but done well, smart policies can cultivate long-lasting regional ecosystems. By leveraging the benefits of federalism, the United States can pursue parallel experimentation where 50 states and thousands of localities experiment with different approaches to cultivating industry clusters, competing to attract investment and talent.

Several state-led efforts illustrate the approach. When new semiconductor fabrication plants were announced in New Albany, Ohio, the state committed roughly $2 billion in incentives spanning direct cash “onshoring incentive grants,” infrastructure spending, and job creation tax credits.140 A community college in Columbus now leads a statewide network of 23 Ohio colleges developing open, shareable curricula for a two-year degree pathway into chip manufacturing technician careers.141 A semiconductor company has invested $ 50 million in Ohio higher education to support comprehensive semiconductor workforce development, spanning curriculum development, faculty training, reskilling and upskilling programs, work-based learning, and laboratory equipment upgrades, multiply federal dollars from NSF.142 In Taylor, Texas, state-led support for semiconductor manufacturing helped attract a nearly $5 billion capital investment supporting thousands of high-quality jobs.143 Across these cases, state leaders have treated industrial development as a core priority, aligning local government, educational institutions, and industry on workforce development. As localities experiment, the most successful will discover models others can adapt.

MAKING PROGRESS AVAILABLE TO ALL

American innovation has never been confined to elite university laboratories. In a local workshop, hobbyists young and old work together to retrofit a Computer Numerical Control (CNC) machine to wind carbon-overwrapped pressure vessels. At a private airstrip on the East Coast, a defense technology startup has pitched a trailer at the end of the field, experimenting with new propellant mixtures. Out in the mesas of the New Mexico desert, a student tinkers in a rusty shed, trying to harness the power of the sun through an inertial confinement fusion device cobbled together from laboratory surplus. Innovation advances in these places, far from major research campuses, through hands-on experimentation, the accumulation of craft knowledge, and a bias toward action.

If this tradition is to endure, we must build a future in which entrepreneurship is not confined by geography or credentials, in which a student in a rural town, a machinist in a small city, and a researcher at a major university each has pathways to contribute to technological progress. A future in which science is not gated by pedigree, but open to all with aptitude and drive. The falling costs of computational power, manufacturing equipment, and scientific tools make this increasingly possible, placing capabilities once reserved for major corporations in the hands of small businesses, community colleges, and individual tinkerers. Alongside new pathways to contribute to the scientific enterprise, this future will create pathways for all to experience the benefits of scientific and technological progress. In the past, the integration of science and craft has produced entirely new markets, industries, and forms of work. Aircraft mechanics, modern welding, CNC machining, and semiconductor manufacturing emerged as skilled crafts in the 20th century; none existed a generation before. Each required its own body of tacit knowledge, its own communities of practice, and its own ties to science and engineering. Each created dignified work for millions of Americans and sustained entire communities. We should expect the same from the technologies taking shape today, but only if we make broad participation possible.

Rebuilding our industrial commons is not the work of a single administration. The clusters that once defined American industry took generations to build, and only years to hollow out when production moved overseas. Meeting this moment requires revitalizing the web of skills, suppliers, and tacit knowledge that form America’s industrial commons, and cultivating vibrant communities of scientists, researchers, and craftspeople in a hundred Silicon Valleys across the nation.

The triumphs of American science and technology have never been the work of any singular locale, and have always drawn on distributed strengths. The Manhattan Project pulled talent from across the country, including physicists from Berkeley, engineers from Tennessee, and craftsmen from New Mexico. The space program employed hundreds of thousands in facilities spread across multiple states, from Houston to Huntsville to Cape Canaveral. The agricultural modernization that multiplied farm productivity in the mid-20th century was driven by land-grant universities and extension agents serving communities in every corner of the nation. Throughout our history, American scientific genius has been broad-based and open to all, and it must remain so. 57

THE AGE OF INTELLIGENCE

In July 1945, the same month Vannevar Bush submitted Science: The Endless Frontier to President Truman, he published a companion essay titled “As We May Think.” Where the former laid the institutional foundations for postwar science, the latter imagined its cognitive structure. Bush foresaw growing drags on the scientific enterprise, writing:

There is a growing mountain of research. But there is increased evidence that we are being bogged down today as specialization extends. The investigator is staggered by the findings and conclusions of thousands of other workers—conclusions which he cannot find time to grasp, much less to remember, as they appear.144

Bush recognized that the tools of his era had extended man’s physical strength and perception. Trip hammers augmented the fist; microscopes sharpened the eye. But the instruments to extend human thought remained rudimentary. Bush proposed a remedy he called the memex, a device that could store, retrieve, and link the entire accumulated record of human knowledge: Consider a future device for individual use, which is a sort of mechanized private file and library… a device in which an individual stores all his books, records, and communications, and which is mechanized so that it may be consulted with exceeding speed and flexibility. It is an enlarged intimate supplement to his memory.145 Over the eighty years since, the instruments Bush imagined have arrived in the form of computers, the internet, and now, AI. Today, the world stands at the threshold of a dramatic transformation. In 2025 alone, American companies committed more than $400 billion to building out AI infrastructure, more than the inflation-adjusted cost of the Apollo Program and Manhattan Project combined.146 Capital exceeding the gross domestic product of most nations is now spent on matrix multiplications, as city-scale symphonies of chips work in concert to train the next generation of AI models.

The result is AI systems that would have been unbelievable even to the most forward-thinking AI researchers five years ago, machines capable of reasoning through complex problems, understanding context and nuance, maintaining large codebases autonomously, and solving mathematics problems at a graduate level.

The early returns from AI for science are striking. Narrow systems have predicted protein structures with atomic accuracy,147 designed novel proteins from scratch,148 modeled molecular dynamics for drug discovery,149 advanced plasma control for fusion research,150 and uncovered phenomena hidden in vast quantities of particle physics data that human analysts would have missed entirely.151 In 2026, AI agents can write entire software applications, analyze experimental data, and operate laboratory equipment without human intervention. In mathematics, perhaps the purest field of reasoning, AI systems have begun to prove novel theorems and enable new forms of mathematical collaboration. These capabilities are improving rapidly as billions of dollars pour into AI research every year.

Our goal, however, should not merely be to accelerate existing methods. It should be reforms of the scientific enterprise that allow us to reach beyond the limits of human cognition and organization.152

Our brains, today, impose hard boundaries. We can hold only so many variables in working memory, read only so many papers in a career, and master only so many techniques in a lifetime. So too does the sociology of science. Disciplines fragment knowledge, incentives reward incremental work, and hierarchies suppress unconventional thinking. Even our means of communication are constrained by the acceptable mediums of words, equations, and charts. AI tools offer the possibility of breaking through all of these limitations. Consider how scientists digest knowledge. A researcher today faces the same problem Bush identified in 1945, only greatly magnified. Millions of papers are published each year,153 more than any human can read in a fraction of their own subfield, let alone adjacent domains where the most important connections hide. AI systems, by contrast, can extract and synthesize findings from vast literatures, identify patterns that span disciplinary boundaries, and search combinatorial spaces of hypotheses and ideas inaccessible to human capacities.154 Consider the collection of scientific data. Until recently, only cleaned and structured data could be used reliably at scale. AI systems can now process and annotate raw data buried in old publications, archives, and video recordings.155

The Python notebooks used to discover the transformer architecture,156 or the boxes of paper Andrew Wiles filled before finally proving Fermat’s Last Theorem, can now be processed by machines as well as humans. The metadata behind the tacit knowledge of science, the kind that never makes it into print, is becoming legible for the first time.

And consider the process of experimentation itself. Traditionally, experiments have been designed by hand, run sequentially, and adjusted only after results were reviewed, limiting both their speed and scope. AI-driven planning, combined with industrial-scale autonomous laboratories, will let us parallelize data collection with custom techniques tailored to each run. Operating in closed loop, AI systems can identify which measurements will yield the most information and adjust course in real time, as experimentation scales up by orders of magnitude.

The cognitive tools Bush imagined in “As We May Think” have finally arrived. If we get this right, AI will accelerate every stage of the scientific process. It will help identify the most impactful questions, generating hypotheses that would never occur to researchers constrained by their training and field. It will design experiments, optimize protocols, and anticipate pitfalls based on the full shared record of prior work. It will analyze data at scales far exceeding what is currently possible and predict behaviors in complex systems that were previously impenetrable to mathematical modeling, from large-scale brain dynamics to longrange weather patterns. It will accelerate how scientific results are communicated and validated, breaking the constraining form of the scientific paper. And increasingly, it will participate in the engineering work that translates scientific discoveries into deployable technologies.