Introduction
If you’ve been following tech news lately, you might have noticed something unusual happening in the American economy. The companies that once prided themselves on being “asset-light” — scaling with nothing more than clever code and a few servers — are now spending billions on concrete, steel, and silicon . Microsoft, Google, Amazon, and Meta have collectively transformed into infrastructure companies, pouring money into data centers, custom chips, and the power grids needed to run them.
This isn’t just a corporate trend. It represents a fundamental shift in how the United States approaches technology innovation. After decades of relying on the private sector to lead, the federal government has quietly built one of the most ambitious public-private innovation networks in the country’s history. Through programs like the Small Business Innovation Research (SBIR) initiative, strategic semiconductor investments, and the newly launched NSF Tech Accelerators, the U.S. is rewriting the playbook on how technology moves from research labs to real-world applications .
What’s driving this shift? The answer lies in a simple but urgent realization: America’s technological leadership depends on rebuilding the physical and institutional infrastructure that turns breakthroughs into industries.
The SBIR Model: How Federal Procurement Became an Engine for Innovation
Most people don’t think of government procurement as a tool for technological advancement. But the SBIR and its companion program, the Small Business Technology Transfer (STTR), have quietly become one of the most effective innovation engines in the federal government .
How It Works
The premise is straightforward: federal agencies with research budgets are required to set aside a portion of their funding for small businesses. These companies compete for Phase I awards, which typically average about $170,000, to test the feasibility of their ideas. Those that succeed can move to Phase II, with average awards of around $1.1 million, to develop prototypes .
What makes the program powerful is the relationship it creates between federal agencies and small businesses. The government acts as a “legitimized customer,” providing not just funding but validation. For a small tech company, winning an SBIR award signals to private investors that their technology has been vetted by serious experts.
The AI Evolution
As artificial intelligence has emerged as a priority, the SBIR/STTR network has adapted. The program has helped small AI companies navigate what researchers call the “pathway from emerging technology to general-purpose technology” . In other words, it has helped AI move from a niche academic interest to a foundational technology that underpins everything from healthcare to manufacturing to national security.
The Federal Government’s AI market development has followed a pattern researchers call “isomorphism across federal organizations” . One agency’s adoption of AI technology paves the way for others to follow. This creates a cascading effect that accelerates the diffusion of new technologies throughout the economy.
A Wicked Problem Remains
For all its success, the program has struggled with diversity. Only about 15% of SBIR/STTR awardees have been women-owned businesses, just 4% have been HUB Zone-owned, and only 1% have been socially and economically disadvantaged businesses . Researchers suggest that policymakers need to experiment with demand-side innovation strategies to reverse this pattern, noting that “demand-based innovation policy through strategic procurement may be able to reverse this wicked problem” .
The Priority Technologies Roadmap: Semiconductors, Biotech, and Beyond
Earlier this year, a group of MIT faculty published a book that has quietly become required reading for policymakers interested in American innovation . “Priority Technologies: Ensuring U.S. Security and Shared Prosperity” identifies six areas where the U.S. must act to maintain its technological edge: semiconductors, biotechnology, critical minerals, drones, quantum computing, and advanced manufacturing.
Semiconductors: The Oxygen of Modern Society
The semiconductor chapter, written by MIT’s Jesús A. del Alamo, describes chips as “the oxygen of modern society” . It’s not hyperbole. When the semiconductor shortage hit in 2021, it accounted for about a third of the inflation the economy experienced that year.
The U.S. essentially pioneered the semiconductor industry, then watched as manufacturing shifted overseas. As the book explains, “people thought the U.S. could lose the manufacturing, stay on top of the innovation and design side, and would be fine. But it’s turned out to make the country quite vulnerable” .
This vulnerability is why the CHIPS Act and subsequent investments matter. The U.S. is in the process of rebuilding its capacity to manufacture leading-edge logic chips, and early signs suggest the strategy is working. “That’s been a successful strategy in recent years,” Reynolds notes .
Biotechnology: Bringing Manufacturing Home
Biotechnology presents a similar pattern: American leadership in research, but a lack of manufacturing infrastructure to bring discoveries to market at the same pace as competitors . The book argues that rebuilding domestic biomanufacturing capacity is essential, and that the technologies developed in the process could themselves create new industries.
Why These Technologies, and Why Now?
The book’s editors argue that “there are breakthroughs to be had in each of these areas, where the U.S. can leapfrog competitors and gain an advantage” . These six technologies are “front and center for U.S. national economic and security policy” . The goal isn’t just technological supremacy for its own sake. As economist Simon Johnson puts it, “Out of that flow of innovations and ideas, we can create more good jobs for all Americans” .
The AI Infrastructure Boom: Tech Companies Become Infrastructure Builders
While government programs and policy roadmaps are important, the most visible sign of America’s tech resurgence is the extraordinary capital spending by the major tech companies.
The End of “Asset-Light”
For three decades, the Silicon Valley gospel was about scaling to billions of users with minimal physical assets. That era is over. “The asset-light paradigm has been buried under a mountain of capital expenditures,” as one analysis put it, “marking a return to a style of industrial expansion not seen since the heyday of the American steel and railway magnates” .
What are they spending on? The list is extensive: GPUs and custom AI chips (like Google’s TPUs and Amazon’s Trainium chips), data centers, land, buildings, server halls, cooling systems, and increasingly, power infrastructure . The less glamorous items are becoming just as critical: grid connections, transformers, fiber optics, and networking equipment.
The Economic Impact
This investment has had a counterintuitive effect on the broader economy. The Federal Reserve raised interest rates significantly in recent years, which theoretically should have choked off private investment. Instead, U.S. GDP has remained remarkably resilient. In the first quarter of 2026, hardware and software investment combined contributed 1.34 percentage points to overall GDP growth of 2.0% .
The productivity impact is still unfolding. While productivity growth has improved compared to the previous decade, it’s probably too early to attribute much to AI . Some economists have even suggested that AI might have temporarily held back productivity as workers spend time learning new systems and companies reorganize their workflows. The long-term picture, however, remains promising.
The Stargate Project and AI Supercomputing
One of the most ambitious projects underway is the Stargate Project, an OpenAI, Oracle, and SoftBank initiative that plans to build multiple AI data center sites across the U.S. . This private investment is being matched by significant public investment in AI infrastructure. The Department of Energy recently announced a new public-private partnership to build two AI supercomputers, Discovery and Lux, through collaborations with AMD and HPE . NVIDIA and Oracle are also building what will be the DOE’s largest AI supercomputer for scientific discovery .
Meanwhile, the National Science Foundation is funding the National AI Research Resource Operations Center, providing researchers across the country with access to computing resources that were previously only available to the largest tech companies . The American Science Cloud initiative is similarly designed to provide broad access to computational resources for scientific research .
The Birth of NSF Tech Accelerators
The government’s role in the innovation ecosystem is evolving. In May 2026, the National Science Foundation announced a new initiative called NSF Tech Accelerators .
Bridging the “Valley of Death”
The program is designed to solve a persistent problem in research: the gap between basic science and commercial application. Traditional research often struggles “to traverse the ‘valley of death'” — the gap between discovery and market-ready technology .
The Tech Accelerators model takes a different approach. Instead of simply funding research and hoping it becomes commercially viable, the NSF will invest in teams and provide commercialization expertise. The goal is to “remove commercialization barriers, address ecosystem and technology-specific gaps, and ‘crowd in’ investment from venture capital and others” .
The First Topics
The initial topic areas are agricultural technology, materials technology, ocean technology, and scientific instrumentation . These might not sound as glamorous as AI or quantum computing, but they represent areas where America’s research capacity is unmatched, and where commercial applications could have enormous economic and social impact.
Conclusion: A New Model for Innovation
The picture that emerges from these developments is one of a country actively redesigning its innovation system. The SBIR/STTR programs provide a mechanism for small companies to get started . The Priority Technologies roadmap identifies where to focus . The NSF Tech Accelerators provide a path for research to become commercial products . And private investment — massive private investment — is building the physical infrastructure needed to support the AI-driven future .
What makes this moment different from previous technology waves is the scale of public-private coordination. The government is not just funding research; it’s acting as a customer, a coordinator, and a convener. Private companies are not just developing technology; they are building the physical infrastructure that technology requires.
The semiconductor shortages of 2021 and the ongoing concerns about American competitiveness have made clear that the old model — letting innovation happen where it will, without strategic coordination — is no longer sufficient. The new model is more deliberate, more coordinated, and more capital-intensive.
Whether it works remains to be seen. But the American tech ecosystem, for better or worse, is being rebuilt from the ground up.
No responses yet