Introduction: The Deep Roots of Regional Technology Power
The story of American high-technology economic power has never been a story of the nation as an undifferentiated whole. It has always been, first and foremost, a story of particular places. Silicon Valley, stretching from San Jose northward to San Francisco, became the world’s preeminent innovation ecosystem not by accident but through the convergence of Stanford University’s research culture, DARPA defense spending, Shockley’s semiconductor laboratory, and the open, collaborative organizational culture that AnnaLee Saxenian would later describe as the defining institutional difference between Silicon Valley and its eastern rival. Route 128 in Massachusetts traced a half-circle of defense electronics and minicomputer firms through Boston’s suburbs, animated by MIT’s research excellence and close relationships with the Pentagon. North Carolina’s Research Triangle Park, spanning more than seven thousand acres between Raleigh, Durham, and Chapel Hill, was built through deliberate regional economic statecraft in the 1950s and grew into one of the nation’s premier research-intensive industrial zones.
These were not accidents. They were the product of deliberate investments in universities, infrastructure, talent pipelines, and institutional culture that reinforced one another across time. They produced regional ecosystems whose competitive advantages compounded over decades into positions of extraordinary economic dominance. And they offer a crucial historical lens through which to understand what is happening today, as artificial intelligence reshapes the economic geography of the United States and the world.
During the 1990s, while attending graduate school at the University of Southern California in Los Angeles, I found myself returning week after week, in seminars and informal conversations with professors and fellow students, to a set of texts that seemed to illuminate not just the past but the future of regional economic development. Chief among them was AnnaLee Saxenian’s landmark 1994 volume, Regional Advantage: Culture and Competition in Silicon Valley and Route 128. Saxenian’s central argument was elegant and counter-intuitive: despite sharing similar technological origins in university research and defense spending, Silicon Valley and Route 128 had diverged radically in their economic trajectories, and the reason for this divergence was not primarily technological but organizational and cultural.
“Silicon Valley developed a decentralized but cooperative industrial system while Route 128 came to be dominated by independent, self-sufficient corporations.”[1]
— AnnaLee Saxenian, Dean, School of Information, University of California at Berkeley, Regional Advantage (Harvard University Press, 1994)
Alongside Saxenian, I engaged deeply with Allen J. Scott’s Technopolis: High-Technology Industry and Regional Development in Southern California, Michael Storper’s 1997 masterwork The Regional World: Territorial Development in a Global Economy, and Michael E. Porter’s theory of competitive clusters. Storper proposed that regions were best understood as “nexuses of untraded interdependencies” — webs of informal knowledge, trust, shared conventions, and institutional legacies that could not be reduced to simple factor costs.
“In conceptualizing the region as a nexus of untraded interdependencies, Storper forces recognition of the decisive influence upon economic competitiveness, innovation, and adaptability, of the social relations of proximity such as sedimented actor rationalities, local knowledge and learning environments, social and cultural conventions, and industrial and institutional legacies.”[2]
— Ash Amin, Professor of Geography, University of Durham, reviewing Michael Storper, The Regional World (Guilford Press, 1997)
“Competitive advantage lies increasingly in local things — knowledge, relationships, and motivation — that distant rivals cannot replicate.”[3]
— Michael E. Porter, Harvard Business School, “Clusters and the New Economics of Competition,” Harvard Business Review (November–December 1998)
These foundational 1990s insights were amplified and extended through important scholarship in the 2000s and 2010s. Edward Glaeser of Harvard, in his landmark 2011 work Triumph of the City, demonstrated empirically that dense urban agglomerations remain the irreplaceable engines of innovation because of the knowledge spillovers that physical proximity enables. Glaeser showed that, all else equal, a city with twice the employment density of another would exhibit a patent intensity roughly twenty percent higher.
“Ideas spread more easily in denser places.”[4]
— Edward Glaeser, Fred and Eleanor Glimp Professor of Economics, Harvard University, Triumph of the City (Penguin Press, 2011)
Richard Florida, then at Carnegie Mellon University, argued in his influential 2002 work The Rise of the Creative Class that regional economic vitality in the twenty-first century would be determined less by access to raw materials or capital and more by the ability of places to attract, retain, and stimulate creative talent.
“Creativity has replaced raw materials or natural harbors as the crucial wellspring of economic growth.”[5]
— Richard Florida, University Professor, University of Toronto, The Rise of the Creative Class (Basic Books, 2002)
Enrico Moretti of UC Berkeley synthesized these insights in his 2012 volume The New Geography of Jobs, demonstrating that the United States was fracturing into “brain hubs” — cities with dense innovation ecosystems generating extraordinary economic dynamism — and a vast periphery being left behind. Moretti showed that for every new high-technology job created in a brain hub, approximately five additional service-sector jobs were generated in the same local economy.
In the 2020s, the economists most directly engaging with AI’s geography have brought an even sharper focus on the divergent outcomes that technological concentration can produce. Erik Brynjolfsson and Andrew McAfee of MIT argued in their 2014 work The Second Machine Age that digital technologies were doing for mental power what steam did for muscle — and that the geography of those gains would not be uniform.
“Now comes the second machine age. Computers and other digital advances are doing for mental power — the ability to use our brains to understand and shape our environments — what the steam engine and its descendants did for muscle power.”[6]
— Erik Brynjolfsson and Andrew McAfee, MIT, The Second Machine Age (W. W. Norton, 2014)
Daron Acemoglu of MIT, the 2024 Nobel Laureate in Economics, argued in Power and Progress (2023) that the direction of technological change is a choice made by powerful actors, and that without deliberate effort to broaden both the development and deployment of AI, its gains will flow disproportionately to already-advantaged places and people.
“New ways of organizing production and communication can either serve the narrow interests of an elite or become the foundation for widespread prosperity.”[7]
— Daron Acemoglu, Institute Professor of Economics, MIT, 2024 Nobel Laureate, Power and Progress (PublicAffairs, 2023)
The Central Thesis: The Five-Layer AI Economy as the New Geographic Framework
On April 16, 2026, at Stanford Graduate School of Business, NVIDIA founder and CEO Jensen Huang described artificial intelligence as a “five-layer cake” — a structural model spanning energy, chips, datacenters, models, and applications. This same framing had appeared earlier at the World Economic Forum Annual Meeting, where AI as infrastructure rather than software was repeatedly emphasized.
“AI is not just software. It is infrastructure.”[8]
— Jensen Huang, Founder and CEO, NVIDIA, Stanford Graduate School of Business (April 16, 2026)
This paper adopts and extends that conceptual statement into a fully articulated geographic framework. The Five-Layer AI Economy — as developed by Dr. Stefanus Hadi at Stefanus.AI in April 2026 — is not a metaphor. It is a structural model identifying five vertically stacked and mutually constraining layers through which artificial intelligence is produced, scaled, and monetized: (1) Energy, (2) Chips, (3) Datacenters, (4) Models, and (5) Applications. In this framework, energy determines whether computation is possible; chips determine how efficiently computation can be performed; datacenters determine how much computation can scale; models determine how intelligence emerges from compute; and applications determine how that intelligence is monetized.
“Power flows bottom to top. Revenue flows top to bottom.”[9]
— Dr. Stefanus Hadi, “Five-Layer AI Economy: What Is It? Who Are the Key Players?” Stefanus.AI (April 2026)
An Intelligence Corridor is defined in this paper as a regional ecosystem in which two or more of these five layers achieve sufficient concentration and integration to generate self-reinforcing cycles of intelligence production and economic value. Just as railroads created industrial corridors in the nineteenth century, and just as highway systems created logistics corridors in the twentieth century, the Five-Layer AI Economy is now creating Intelligence Corridors in the twenty-first. The future winners of the AI era may not be individual companies. They may be entire corridors.
Section One: The Geography of Intelligence
From Industrial Clusters to Intelligence Clusters
Economic historians have long understood that technological revolutions do not merely change what is produced but reshape where it is produced. The geography of production is not incidental to economic development; it is constitutive of it. Each great wave of technological change has generated new spatial logics, new principles of agglomeration, and new hierarchies of place. Before the railroad, economic activity clustered around rivers, harbors, and coastlines. The transcontinental railroads of the 1860s and 1870s created the great industrial corridors of the Midwest. The interstate highway system and the commercial jet enabled a new geography of dispersed manufacturing and just-in-time logistics.
Each of these epochs created regional winners and losers. Places that sat at the intersection of the new infrastructure with favorable combinations of talent, capital, and institutional capacity surged ahead. Places that could not access the new infrastructure, or that clung too long to the institutional patterns of the preceding era, fell behind.
The Brookings Institution’s research on AI geography has confirmed empirically what regional theory predicts: just thirty metro areas account for two-thirds of all AI-related job postings in the United States, and more than half of the 261 metro areas surveyed exhibited no significant AI activity at all.
“This is a powerful technology that will sweep through American offices with potentially very significant geographic implications.”[10]
— Mark Muro, Senior Fellow, Brookings Institution Metropolitan Policy Program, Las Vegas Sun (December 2024)
Why Geography Still Matters More Than Ever
For at least thirty years, a persistent strand of commentary has predicted that digital technology would render geography irrelevant. Nicholas Negroponte distinguished between the world of “atoms” and the world of “bits.” Thomas Friedman’s The World Is Flat argued the internet had leveled the playing field. The AI revolution has decisively refuted this thesis at the foundational layers of the economy.
The International Energy Agency’s April 2026 report, Key Questions on Energy and AI, provided the most rigorous quantitative grounding for this geographic argument. Global electricity consumption from data centers soared by seventeen percent in 2025, outpacing total global electricity demand growth of three percent by more than fivefold.
“The largest technology companies are contributing to a surge in data centre investment, as their capital expenditure exceeded USD 400 billion in 2025 – and is expected to jump by another 75% in 2026. Capital expenditure of just five technology companies is now larger than global investment in oil and natural gas production.”[11]
— International Energy Agency (IEA), Key Questions on Energy and AI (April 2026)
Energy cannot be downloaded. Semiconductor fabs cannot be downloaded. Transmission lines cannot be downloaded. The Five-Layer AI Economy is, at its foundation, a physical economy — and physical economies have geography. The IMF reinforced this in its 2026 workshop note, warning that the macroeconomic path of AI adoption would be shaped not merely by frontier AI capabilities but by “the readiness of institutions and infrastructure to absorb the technology.”
“AI is advancing rapidly and has the potential to restructure the global economy, which calls for treating AI as a macro-critical transition rather than a standard technology shock.”[12]
— International Monetary Fund, Global Economic and Financial Implications of Artificial Intelligence, IMF Notes Vol. 2026, Issue 002 (April 2026)
Section Two: The Five-Layer AI Economy — A Framework for Measuring Intelligence Corridors
The measurement standard employed throughout this paper is the Five-Layer AI Economy framework, developed by Dr. Stefanus Hadi and published at Stefanus.AI in April 2026. The framework identifies five vertically stacked layers that together constitute the full production stack of artificial intelligence — from the physical substrate of electrical power through the economic layer of applications and agentic systems. Each layer depends on the one below it and constrains the one above it. Power flows upward through the stack; revenue flows downward. A region’s position in the Intelligence Corridor hierarchy is determined by how many layers it anchors, and how deeply.
“The AI economy is not driven by ideas alone — it is governed by physical, industrial, and geopolitical constraints.”[9]
— Dr. Stefanus Hadi, “Five-Layer AI Economy: What Is It? Who Are the Key Players?” Stefanus.AI (April 2026)
The framework also reveals a fundamental asymmetry: constraints reside at the bottom of the stack, while monetization occurs at the top. This means that whichever regions control Layer 1 (Energy) and Layer 2 (Chips) hold the most durable and most defensible competitive positions in the Intelligence Corridor economy — because the upper layers of Datacenters, Models, and Applications can only exist where the lower layers have already been secured.
Layer 1: Energy — The Physical Substrate of Intelligence
Energy is, in the words of the Five-Layer AI Economy framework, “the non-negotiable constraint that defines the entire system.” Every computation, every model inference, every training run is fundamentally a conversion of electrical energy into information processing. IBM physicist Rolf Landauer established the theoretical underpinning of this relationship in his landmark 1961 paper, arguing that information is physical and that every irreversible computation has an unavoidable thermodynamic energy cost.
“The ultimate limit to computation is energy.”[13]
— Rolf Landauer, IBM Research, “Information is Physical” (1961)
The scale shift in AI energy demand is unprecedented and accelerating. Traditional enterprise datacenters consumed 5 to 20 megawatts. Hyperscale cloud datacenters consumed 50 to 100 megawatts. AI clusters today consume 100 to 300 megawatts. Next-generation AI campuses — such as the Stargate flagship in Abilene, Texas — are approaching one gigawatt of sustained demand. The IEA projects electricity consumption from data centers will roughly double from 485 TWh in 2025 to 950 TWh in 2030, with AI-focused datacenter consumption tripling in that same period.
“Data centers are becoming the new steel mills of the digital economy.”[14]
— McKinsey Global Institute, Data Centers and the AI Economy (2024)
A critical structural mismatch defines this layer: renewable energy is intermittent while AI compute is continuous and non-stop. This drives the renaissance of nuclear energy, particularly Small Modular Reactors (SMRs). The IEA reported in April 2026 that the pipeline of conditional offtake agreements between datacenter operators and SMR projects had grown from 25 gigawatts at the end of 2024 to 45 gigawatts — an 80 percent increase in eighteen months. Major energy players in this layer include NextEra Energy, Constellation Energy, Vistra, Oklo, and Chevron.
For the Intelligence Corridor Index, Layer 1 (Energy) is scored on: total electrical generation and transmission capacity; cost and reliability of the grid; diversity of the generation mix, particularly the share of always-on generation from nuclear, natural gas, or large-scale hydroelectric sources; the pace of new generation development in response to AI demand; and the presence of behind-the-meter or co-located power arrangements serving AI infrastructure.
Layer 2: Chips — The Intelligence Engine
If energy is the physical substrate of AI, chips are the intelligence engine. This is the layer where electrical energy is converted into computation. The Five-Layer AI Economy framework identifies chips as the most geopolitically sensitive layer in the entire stack, because the manufacturing of leading-edge semiconductors is concentrated in an extraordinarily small number of facilities, controlled by an extraordinarily small number of companies, clustered in an extraordinarily small number of geographic locations.
“Compute is the new currency of artificial intelligence.”[15]
— Stanford Human-Centered AI Institute, AI Index Report (2024)
Approximately 75 percent of AI workloads run on GPUs, and approximately 90 percent of those GPUs are produced by NVIDIA. The manufacturing of NVIDIA’s most advanced chips depends entirely on Taiwan Semiconductor Manufacturing Company (TSMC), which is the sole manufacturer capable of producing the leading-edge nodes — 4nm, 3nm, 2nm — that AI accelerators require. TSMC’s production itself depends on extreme ultraviolet (EUV) lithography machines produced exclusively by ASML in the Netherlands, each costing over $200 million and requiring decades of accumulated R&D that cannot be easily replicated.
“ASML is the sole supplier of EUV lithography systems.”[16]
— ASML Holdings N.V., Annual Report and Company Filings (2024)
The American dimension of Layer 2 is being transformed by TSMC’s $165 billion investment in its Arizona campus — the largest foreign direct investment in a greenfield project in American history. Fab 21’s first facility entered high-volume production on 4-nanometer process technology in Q4 2024. A second fab targeting 3-nanometer production is accelerated to 2027, and a third fab targeting 2-nanometer and A16 processes broke ground in April 2025.
“AI is reshaping our daily lives and semiconductor technology is the foundation for new capabilities and applications. With the success of our first fab in Arizona, along with needed government support and strong customer partnerships, we intend to expand our U.S. semiconductor manufacturing investment by an additional $100 billion, bringing our total planned investment to $165 billion.”[17]
— Dr. C.C. Wei, Chairman and CEO, TSMC, SEC Form 6-K (March 2025)
NVIDIA’s Q1 FY2027 results (May 20, 2026) revealed record revenues of $81.6 billion for the quarter, with Data Center revenues of $39.1 billion — a 69.2 percent year-over-year increase. Management projected Q2 FY2027 revenues of $91.0 billion.
For the Intelligence Corridor Index, Layer 2 (Chips) is scored on: the presence and scale of semiconductor fabrication facilities; the depth of the broader semiconductor supply chain ecosystem (design, packaging, equipment, materials); CHIPS Act investment anchors; the semiconductor engineering and process talent workforce; and the proximity to and integration with Layer 1 power infrastructure.
Layer 3: Datacenters — The AI Factories
Datacenters are the layer where chips are aggregated into massive compute systems — the industrial factories of the intelligence economy. As Jensen Huang described at Stanford, the buildout of AI datacenters represents “the largest infrastructure expansion in human history.” The Five-Layer AI Economy framework characterizes datacenters as the “AI factories” where the raw materials of Layer 1 (energy) and Layer 2 (chips) are combined into the industrial-scale compute capacity that makes frontier model training and inference possible.
“The buildout of AI factories — the largest infrastructure expansion in human history — is accelerating at extraordinary speed.”[18]
— Jensen Huang, Founder and CEO, NVIDIA, Q1 FY2027 Earnings Call (May 20, 2026)
The capital commitment to Layer 3 buildout is extraordinary. The five largest hyperscalers — Amazon, Alphabet/Google, Microsoft, Meta, and Oracle — collectively committed between $660 and $690 billion in capital expenditure for 2026 alone, nearly double 2025’s $388 billion, with roughly three-quarters of that spend directed at AI infrastructure. Goldman Sachs projects total hyperscaler capex from 2025 to 2027 will reach $1.15 trillion — more than double the $477 billion spent from 2022 to 2024.
“Agentic AI has arrived, doing productive work, generating real value and scaling rapidly across companies and industries. NVIDIA is uniquely positioned at the center of this transformation as the only platform that runs in every cloud, powers every frontier and open source model, and scales everywhere AI is produced.”[19]
— Jensen Huang, NVIDIA Q1 FY2027 Earnings Call (May 20, 2026)
Northern Virginia hosts the world’s largest concentration of datacenter capacity, with Data Center Alley’s more than 100 facilities representing over five gigawatts of commissioned power. Texas has 6.5 gigawatts under construction, more than any other single market in North America. The Stargate Project — announced by OpenAI, SoftBank, and Oracle as a $500 billion, four-year commitment to ten gigawatts of AI datacenter capacity — has its flagship campus operational in Abilene, Texas and additional sites across the country.
For the Intelligence Corridor Index, Layer 3 (Datacenters) is scored on: total commissioned and under-construction datacenter power capacity in megawatts or gigawatts; the presence of hyperscaler and cloud provider facilities; fiber connectivity density and network infrastructure; proximity to Layer 1 (Energy) and Layer 2 (Chips); and the regulatory and zoning environment for datacenter permitting and expansion.
Layer 4: Models — The Brains of the Stack
Models are the layer where compute becomes intelligence. This is where the raw capacity of Layers 1 through 3 is transformed into the large language models, foundation models, vision models, and reasoning systems that drive the AI economy. The Five-Layer AI Economy framework describes models as “the brains” of the stack — the intelligence engines that emerge from sufficient compute, data, and algorithmic ingenuity.
“Scaling laws reveal that larger models trained on more data and compute perform better.”[20]
— Jared Kaplan et al., “Scaling Laws for Neural Language Models,” OpenAI / Johns Hopkins University, arXiv:2001.08361 (2020)
Frontier model development requires resources comparable to large industrial projects: tens of thousands of GPUs, months of continuous training runs, and energy consumption measured in hundreds of megawatt-hours. The structural constraint of this layer is therefore not primarily algorithmic but infrastructural — breakthroughs in model capability are increasingly a function of access to compute, which is itself a function of geography.
“The limiting factor in AI progress is increasingly compute, not ideas.”[21]
— Stanford Human-Centered AI Institute, AI Index Report (2024)
Model development is also marked by deep structural alignment between AI labs and infrastructure providers: Microsoft and OpenAI, Amazon and Anthropic, Google and its internal DeepMind stack. This means that the geography of Layer 4 tends to follow the geography of Layer 3, with the major model development centers clustering near or within the hyperscaler infrastructure ecosystems. Key players in this layer include OpenAI, Anthropic, xAI, Google DeepMind, and Meta AI.
For the Intelligence Corridor Index, Layer 4 (Models) is scored on: the concentration of frontier AI research organizations and labs in or near the region; the presence of major AI company headquarters, research campuses, or significant operations; university and academic AI research output; venture capital investment in AI model companies; and the presence of federal or defense AI research programs.
Layer 5: Applications — The Economic Layer
Applications are the layer where intelligence becomes revenue. This is the economic summit of the Five-Layer AI Economy stack — the layer where everything below is monetized. The framework’s most important insight about this layer is what it calls the “Revenue vs Power Paradox”: applications generate the majority of the AI economy’s visible revenue, but they are entirely dependent on the four layers below. Revenue sits at the top. Power sits below.
“AI only creates value when it is applied.”[22]
— World Bank, World Development Report: The Changing Nature of Work (2019)
The application layer encompasses every sector of the economy where AI is being deployed to generate productivity gains, new products, or cost reductions: healthcare AI (drug discovery, diagnostics, clinical workflow), financial AI (fraud detection, algorithmic trading, credit assessment), defense AI (autonomous systems, surveillance, cyber), industrial AI (predictive maintenance, supply chain optimization), agentic AI systems capable of executing complex multi-step tasks autonomously, and the full range of consumer and enterprise software applications.
“AI agents will transform workflows across industries.”[23]
— Andrew Ng, Co-founder, Coursera and deeplearning.ai, “The Batch” Newsletter (2024)
For the Intelligence Corridor Index, Layer 5 (Applications) is scored on: the concentration of AI application companies and the scale of AI deployment across major industry sectors in the region; the density of venture capital investment in AI application startups; the presence of large enterprise or government AI deployment programs; the quality of the AI talent workforce available for application development; and the regulatory environment for AI application deployment.
The Cross-Layer Power Dynamics: Why the Stack Shapes the Corridor
The most critical insight of the Five-Layer AI Economy framework for understanding Intelligence Corridors is the directional asymmetry of power and revenue. Power flows upward through the stack: regions that control energy determine who can build chips; regions that control chips determine who can build datacenters; regions that control datacenters determine who can develop models; regions that control models determine who can build the most capable applications. Revenue flows in the opposite direction, from applications down through the stack.
This asymmetry means that the most strategically valuable positions in the Intelligence Corridor economy are not at the top of the stack (where revenue is most visible) but at the bottom (where constraints are most binding). A region that controls Layer 1 Energy and Layer 2 Chips holds a structural advantage over all regions that can only compete at Layers 3, 4, and 5. This is why Arizona’s position in the Corridor Index, anchored by TSMC’s semiconductor manufacturing, is more durable than that of many states with larger and more visible AI application economies.
Section Three: America’s Emerging Intelligence Corridors
America’s Intelligence Corridors are not yet fully formed. They are in the process of becoming — some rapidly, some more slowly, some facing significant structural obstacles. What follows is an assessment of seven of the most significant emerging corridors, evaluated against the Five-Layer AI Economy framework.
The Northern Virginia Corridor: The Datacenter Capital of the World
Northern Virginia — particularly the Loudoun County communities of Ashburn and Sterling — constitutes the world’s most mature Layer 3 Intelligence Corridor. The region hosts over 199 datacenters in Loudoun County alone, with another 117 in development. Approximately 70 percent of the world’s internet traffic flows through this region. The corridor’s fiber density, proximity to the federal government’s massive data-processing needs, established AWS ecosystem, and deep cybersecurity talent pool have made it a virtually irreplaceable node in the global digital infrastructure.
The corridor’s critical vulnerability is Layer 1 (Energy). Northern Virginia’s datacenter capacity is projected to grow from approximately 16 gigawatts in 2025 to over 20 gigawatts in 2026 and potentially 43 gigawatts by 2031. Dominion Energy has proposed a 14 percent rate increase for residential customers in 2026. AI datacenter demand contributed to an 833 percent increase in PJM’s capacity market auction price for 2025–2026. Virginia’s State Corporation Commission approved a new electricity rate class specifically for large-scale AI datacenters in November 2025.
“AI and data are becoming a bigger part of everyone’s lives. About 70% of the world’s internet traffic flows through data centers in the Northern Virginia region.”[24]
— Suhas Subramanyam, former Virginia State Senator and incoming U.S. Representative, 10th Congressional District, InformationWeek (2025)
The Texas Intelligence Corridor: America’s Largest AI Industrial Belt
Texas presents the most dramatic and rapidly evolving Intelligence Corridor story in the United States. As recently as 2020, the narrative was largely centered on Austin’s software ecosystem and Houston’s energy industry. By 2026, however, the emergence of Austin’s Tera Corridor and the expansion of AI-related investments across the state have transformed Texas into a statewide AI infrastructure powerhouse spanning all five layers of the AI stack. Texas’s ERCOT grid, its extraordinary natural gas reserves, and its rapidly expanding renewable and nuclear SMR pipeline give it a Layer 1 advantage that no other state can fully match. Its Stargate Project anchor and Google’s $40 billion investment commitment give it a rapidly developing Layer 3 position. Its energy industry provides natural application infrastructure for Layer 5 industrial AI. And its population growth and business-friendly regulatory environment create powerful gravitational pull for the talent that Layers 4 and 5 require.
“AI can only fulfill its promise if we build the compute to power it.”[25]
— Sam Altman, CEO, OpenAI, Stargate Project Announcement (September 23, 2025)
JLL’s North America Data Center Report for year-end 2025 found that Texas has 6.5 gigawatts of datacenter capacity under construction, more than any other single market in North America, and projected that the Texas market will overtake Northern Virginia as the world’s largest by 2030.
The Arizona Intelligence Corridor: The Semiconductor Frontier
Arizona’s Intelligence Corridor is defined almost entirely by its Layer 2 (Chips) position. TSMC’s $165 billion Arizona campus investment anchors the most significant shift in global semiconductor geography since the industry’s founding. The Greater Phoenix Economic Council estimates that TSMC’s three fabs alone will generate $32.9 billion in total economic output for Arizona over thirteen years, with $9.3 billion in personal income. The Arizona fab posted a $514 million profit in its first full year of operation, with Q1 2026 earnings alone surpassing the full 2025 figure.
Arizona’s key vulnerability is Layer 1 (Energy) in the dimension of water — a special form of infrastructure constraint in the arid Southwest. TSMC has invested in an advanced water reclamation system targeting 90 percent recycling. But as AI infrastructure in Arizona expands — Phoenix is also a growing Layer 3 datacenter hub — the aggregate water demands of the corridor will require sustained policy attention.
The Pittsburgh Intelligence Corridor: America’s Robotics and Physical AI Corridor
Pittsburgh represents perhaps the most intellectually compelling Intelligence Corridor story in the United States. A city synonymous with steel production, that watched its economic base collapse in the deindustrialization of the 1980s, has rebuilt itself around Carnegie Mellon University’s world-leading robotics and AI research programs into an ecosystem with genuine Layer 4 (Models) and Layer 5 (Applications) significance, particularly in physical AI and autonomous systems.
“Carnegie Mellon researchers have been at the forefront of robotics and AI for decades. The Robotics Innovation Center creates a unique environment where researchers from universities, industry, national labs, and startups can collaborate on future generations of intelligent systems that integrate AI and robotics in powerful new ways.”[26]
— Theresa Mayer, Vice President for Research, Carnegie Mellon University (February 2026)
Pittsburgh’s robotics and AI ecosystem now includes more than 120 robotics companies and 120 AI companies, alongside over 55 academic labs and support organizations. Local startups raised over $2.29 billion in 2025. Skild AI, founded by Carnegie Mellon professors, raised a $1.4 billion Series C. The corridor’s primary gap is in Layers 1 through 3: power, chips, and datacenter infrastructure remain significantly less developed than in leading corridors.
The Columbus Corridor: The Midwestern Compute Hub
Columbus, Ohio, is a corridor in its earliest formative stages, but its strategic assets are real. The anchor investment is Intel’s commitment under the CHIPS and Science Act to develop a massive semiconductor manufacturing campus in the New Albany area east of Columbus — a potential Layer 2 anchor of significant scale. Columbus is also emerging as a Stargate site, with Oracle and SoftBank developing Layer 3 datacenter capacity. Its geographic centrality — within a day’s drive of roughly 60 percent of the U.S. population — makes it a natural logistics hub for the physical AI economy.
The Nashville Corridor: The AI Healthcare Hub
Nashville’s emergence as an Intelligence Corridor candidate reflects the deep structural relationship between Layer 5 (Applications) and the healthcare industry. The greater Nashville area is home to over 500 healthcare companies, including HCA Healthcare, Community Health Systems, and Acadia Healthcare. Healthcare AI is among the highest-value application categories in the broader AI economy: the U.S. spends approximately $4.5 trillion annually on healthcare, with a significant fraction consumed by administrative costs and diagnostic inefficiencies that AI systems are increasingly capable of addressing. Nashville’s corridor development strategy rests on the thesis that a city with deep healthcare industry roots can become the dominant hub for healthcare AI development and deployment, even without strong Layers 1 through 3 positions.
The Reno Corridor: The Autonomous Systems Corridor
Reno, Nevada, and the Tahoe-Reno Industrial Center represent a corridor that grew from Tesla’s Gigafactory anchor, which has drawn a growing ecosystem of battery manufacturers, EV component suppliers, and autonomous systems companies. This corridor has a distinctive Layer 5 (Applications) character in physical AI — autonomous vehicles, battery systems, and robotics — supported by Layer 2 (Chips) manufacturing elements in battery and power electronics. Reno’s most significant structural advantage is proximity to California’s talent and capital markets without California’s regulatory costs.
Section Four: The Competition Between Corridors
The New Competitive Unit
The Intelligence Corridor framework requires a fundamental reconceptualization of the unit of economic competition. Throughout the twentieth century, the dominant unit was the individual firm. As the twenty-first century has progressed, the nation-state has re-emerged as a competitive unit, particularly in semiconductors and AI. The Intelligence Corridor framework adds a third competitive unit: the corridor itself.
“Today’s economic map of the world is characterized by what Porter calls clusters: critical masses in one place of linked industries and institutions — from suppliers to universities to government agencies — that enjoy unusual competitive success in a particular field.”[27]
— Michael E. Porter, Harvard Business School, “Clusters and the New Economics of Competition,” Harvard Business Review (November–December 1998)
The Brookings Institution’s research has found that winner-takes-most dynamics are firmly established in AI geography. Just two metro areas — San Francisco and Seattle — dominate large language model research and development activity, and a total of fifteen metro areas account for roughly two-thirds of the nation’s AI assets and capabilities.
“It remains a highly concentrated early-stage industry dominated by the Bay Area. But I think it’s fair to say that there’s quite legitimate diffusion of the elements of success. Places are thinking about talent development specifically attuned to this technology.”[28]
— Mark Muro, Senior Fellow, Brookings Institution, Route Fifty (July 2025)
The Five-Layer Stack as a Competitive Weapon
The most important implication of the Five-Layer AI Economy framework for corridor competition is that corridors compete not as undifferentiated bundles of economic capacity but as vertical stacks. A corridor that controls Layers 1 and 2 — Energy and Chips — has a structural competitive advantage over corridors that can only compete at Layers 3 through 5. This is because the lower layers are both more capital-intensive and more geographically sticky: a semiconductor fab, once built, does not move. An energy generation fleet, once constructed, is a multi-decade asset. By contrast, datacenters can be built in two to three years, model companies can relocate relatively easily, and applications are largely placeless.
This means that the most defensible corridor positions are those that anchor the lowest layers. Arizona’s Layer 2 position, though less visible than Northern Virginia’s Layer 3 dominance, is in many respects more durable. Texas’s Layer 1 position, anchored in the scale and reliability of its ERCOT energy system, is the foundation of a corridor buildout that is just beginning to achieve its full potential.
Capital, Talent, and Regulatory Competition
States are competing aggressively on tax incentives, regulatory fast-tracking, utility rate structures, and workforce development programs to attract the massive capital deployments that anchor Intelligence Corridor formation. The capital attraction competition is not merely about tax rates but about the full ecosystem package that a corridor can offer.
Moretti’s research on brain hubs offers crucial insight into the self-reinforcing nature of talent agglomeration: once a hub achieves critical mass, attracting one high-technology worker creates conditions for attracting additional workers and companies. For Intelligence Corridors, this dynamic is amplified by the extraordinary scarcity of the most specialized AI talent — semiconductor process engineers, large language model researchers, robotics system architects — relative to the scale of demand.
Corridor Vulnerabilities and Failure Modes
The first failure mode is Layer 1 power constraint. Corridors that cannot solve their energy infrastructure challenges will find development stunted regardless of other advantages. The second is Layer 2 talent scarcity, particularly the specialized semiconductor process engineering workforce concentrated in Taiwan, South Korea, and a small number of American universities. The third is regulatory gridlock, as AI infrastructure buildouts generate community resistance driven by concerns about electricity costs, water consumption, and land use impacts.
Section Five: The Next Decade
Corridor Expansion and New Technological Drivers
The Intelligence Corridor landscape of 2035 will look significantly different from the landscape of 2026. Nuclear energy is likely to be one of the most important drivers of new corridor formation over the next decade. The IEA’s April 2026 report noted that the pipeline of conditional offtake agreements between datacenter operators and SMR projects had grown to 45 gigawatts, up from 25 gigawatts at end-2024 — an 80 percent increase in eighteen months. States with strong nuclear regulatory environments and existing nuclear workforces — Pennsylvania, South Carolina, Wyoming, and parts of the Midwest — may find that SMR-anchored Intelligence Corridors emerge as a distinct corridor type in the early 2030s.
Physical AI — robotics, autonomous vehicles, drone logistics, and AI-native manufacturing — is poised to drive a second wave of corridor formation, expanding the locational logic of the AI economy from compute concentration toward manufacturing concentration and physical distribution logistics. Corridors with strong manufacturing roots — Pittsburgh, Columbus, Detroit, and parts of the Southeast — may find that the physical AI wave opens Layer 5 Application opportunities that the first wave of AI infrastructure investment did not.
Intelligence Migration and the Corridor Magnet Effect
One of the most powerful dynamics in the development of mature Intelligence Corridors is what I call the Intelligence Migration effect: the tendency of companies, workers, and capital to migrate progressively toward corridors once those corridors have achieved sufficient critical mass to offer compelling ecosystem advantages. Texas held its title as the number one state for new residents in 2025. Arizona’s growing semiconductor ecosystem is drawing semiconductor equipment companies, advanced materials suppliers, and process engineering talent from around the world.
The self-reinforcing character of these dynamics means the corridor competition of the next decade will be characterized by increasing polarization: a small number of very strong corridors pulling farther ahead, and a larger number of weaker corridors finding it progressively more difficult to achieve critical mass. This is consistent with what Storper identified as the “nexus of untraded interdependencies” — the tendency of small early advantages to compound into large durable advantages as institutional legacies deepen over time.
For policymakers, this creates a genuine urgency. The decisions that states and regions make in the next three to five years about energy infrastructure, semiconductor workforce development, datacenter regulatory frameworks, and AI research investment will shape their corridor trajectories for the next thirty years. The window for entering the corridor competition at a meaningful level is not infinite.
Section Six: The Corridor Index — Ranking All 50 States Against the Five-Layer AI Economy
Framework and Methodology
The Corridor Index presented in this section applies the Five-Layer AI Economy framework as the definitive measurement standard for ranking all fifty states by their AI economic capacity. Each state is scored on five dimensions corresponding exactly to the five layers of the stack: (1) Energy Capacity, (2) Chips Capacity, (3) Datacenters Capacity, (4) Models Capacity, and (5) Applications Capacity. Each dimension is scored on a scale of one to ten. Scores are aggregated into a composite Corridor Index score ranging from five to fifty.
The rankings reflect publicly available data current through Q1 2026, including state utility commission filings, data center market reports, CHIPS Act investment disclosures, NVIDIA and TSMC earnings disclosures, university AI research rankings, venture capital investment data, and AI company location records.
Energy Capacity (Layer 1): Total electrical generation and transmission capacity; cost and reliability of grid; diversity of generation mix, particularly always-on nuclear, gas, or hydroelectric; pace of new generation development; behind-the-meter arrangements for AI infrastructure.
Chips Capacity (Layer 2): Presence and scale of semiconductor fabrication facilities; depth of broader supply chain ecosystem (design, packaging, equipment, materials); CHIPS Act investment anchors; semiconductor engineering talent; Layer 1 proximity.
Datacenters Capacity (Layer 3): Total commissioned and under-construction datacenter power capacity; hyperscaler and cloud provider presence; fiber connectivity density; Layer 1 and Layer 2 proximity; permitting and regulatory environment.
Models Capacity (Layer 4): Concentration of frontier AI research organizations; AI company headquarters or major campuses; university AI research output; venture capital investment in AI model companies; federal and defense AI research programs.
Applications Capacity (Layer 5): AI deployment across major industry sectors; venture capital investment in AI application startups; enterprise and government AI deployment programs; AI application workforce quality; regulatory environment for AI deployment.
The following table presents the full Corridor Index for all fifty states across all five layers of the Five-Layer AI Economy, plus composite score and tier classification.
| State | L1 Energy | L2 Chips | L3 Data | L4 Models | L5 Apps | Total / 50 |
| Virginia | 8 | 5 | 10 | 8 | 10 | 41 |
| Texas | 10 | 6 | 9 | 7 | 9 | 41 |
| California | 7 | 7 | 9 | 10 | 10 | 43 |
| Arizona | 8 | 10 | 8 | 6 | 8 | 40 |
| Washington State | 9 | 5 | 9 | 8 | 9 | 40 |
| Georgia | 8 | 5 | 7 | 6 | 8 | 34 |
| Ohio | 8 | 7 | 7 | 6 | 7 | 35 |
| Pennsylvania | 8 | 4 | 6 | 8 | 7 | 33 |
| North Carolina | 7 | 4 | 7 | 8 | 7 | 33 |
| Nevada | 7 | 4 | 7 | 5 | 7 | 30 |
| Tennessee | 7 | 4 | 6 | 5 | 7 | 29 |
| Michigan | 7 | 4 | 5 | 7 | 6 | 29 |
| New York | 7 | 3 | 6 | 9 | 7 | 32 |
| Maryland | 7 | 3 | 6 | 7 | 8 | 31 |
| Oregon | 8 | 5 | 7 | 5 | 6 | 31 |
| Massachusetts | 6 | 4 | 5 | 9 | 7 | 31 |
| Colorado | 7 | 3 | 6 | 6 | 7 | 29 |
| Utah | 6 | 3 | 6 | 6 | 7 | 28 |
| Indiana | 7 | 4 | 5 | 5 | 6 | 27 |
| South Carolina | 7 | 4 | 5 | 4 | 5 | 25 |
| Wisconsin | 6 | 3 | 5 | 5 | 5 | 24 |
| Minnesota | 6 | 3 | 5 | 5 | 6 | 25 |
| New Mexico | 7 | 2 | 5 | 4 | 5 | 23 |
| Florida | 6 | 3 | 5 | 5 | 6 | 25 |
| Idaho | 7 | 3 | 5 | 4 | 4 | 23 |
| Illinois | 7 | 3 | 5 | 6 | 6 | 27 |
| Kansas | 7 | 2 | 4 | 3 | 4 | 20 |
| Montana | 7 | 2 | 4 | 3 | 4 | 20 |
| Wyoming | 7 | 2 | 4 | 3 | 4 | 20 |
| Oklahoma | 7 | 2 | 4 | 3 | 4 | 20 |
| Alabama | 5 | 2 | 4 | 3 | 4 | 18 |
| Alaska | 5 | 1 | 3 | 2 | 3 | 14 |
| Arkansas | 5 | 2 | 4 | 3 | 4 | 18 |
| Connecticut | 6 | 2 | 4 | 5 | 5 | 22 |
| Delaware | 5 | 2 | 4 | 4 | 5 | 20 |
| Hawaii | 4 | 1 | 4 | 3 | 4 | 16 |
| Iowa | 7 | 2 | 4 | 3 | 4 | 20 |
| Kentucky | 6 | 2 | 4 | 3 | 4 | 19 |
| Louisiana | 7 | 2 | 4 | 3 | 4 | 20 |
| Maine | 5 | 1 | 3 | 3 | 4 | 16 |
| Mississippi | 5 | 1 | 3 | 2 | 3 | 14 |
| Missouri | 6 | 2 | 5 | 4 | 5 | 22 |
| Nebraska | 6 | 2 | 4 | 3 | 4 | 19 |
| New Hampshire | 5 | 2 | 4 | 4 | 4 | 19 |
| New Jersey | 6 | 3 | 5 | 5 | 6 | 25 |
| North Dakota | 7 | 1 | 3 | 2 | 3 | 16 |
| Rhode Island | 5 | 1 | 3 | 4 | 4 | 17 |
| South Dakota | 6 | 1 | 3 | 2 | 3 | 15 |
| Vermont | 5 | 1 | 3 | 3 | 4 | 16 |
| West Virginia | 6 | 1 | 3 | 2 | 3 | 15 |
Tier Descriptions and Key Findings
Tier 1: Established Intelligence Corridors (Composite Score: 40–50)
Five states have achieved sufficient multi-layer integration to qualify as Established Intelligence Corridors: California (43), Virginia (41), Texas (41), Arizona (40), and Washington State (40). These states anchor at least three of the five layers at high scores and exhibit the self-reinforcing dynamics of mature corridor ecosystems.
California (43/50) — L1: 7 | L2: 7 | L3: 9 | L4: 10 | L5: 10. California earns a perfect Layer 4 (Models) and Layer 5 (Applications) score anchored by Stanford, Berkeley, Caltech, Silicon Valley’s AI ecosystem, and the global headquarters of Google, Apple, Meta, and the majority of frontier AI labs. TSMC’s design partners and a growing advanced packaging ecosystem give it a meaningful Layer 2 score. Power constraints temper its Layer 1 position, but California’s breadth across all five layers makes it the most complete Intelligence Corridor in the United States today.
Virginia (41/50) — L1: 8 | L2: 5 | L3: 10 | L4: 8 | L5: 10. Virginia earns a perfect Layer 3 (Datacenters) score — Northern Virginia’s Data Center Alley is the global leader in commissioned datacenter power, hosting 70 percent of world internet traffic. Its Layer 5 (Applications) score reflects the extraordinary AI deployment density driven by federal government, defense, and intelligence agency demand. Layer 1 power constraints (residential rate increases, grid strain) and the near-absence of Layer 2 semiconductor manufacturing are the primary gaps.
Texas (41/50) — L1: 10 | L2: 6 | L3: 9 | L4: 7 | L5: 9. Texas earns a perfect Layer 1 (Energy) score reflecting the scale and independence of the ERCOT grid, extraordinary natural gas resources, and a rapidly expanding renewable and nuclear SMR pipeline that is positioning Texas as the energy backbone of the next generation of AI infrastructure. The Stargate Project, Google’s $40 billion investment, and JLL’s projection that Texas will overtake Northern Virginia as the world’s largest datacenter market by 2030 drive its Layer 3 score. A nascent Layer 2 semiconductor ecosystem and strong Layer 4 and 5 growth from the UT system and Austin’s tech ecosystem round out the picture.
Arizona (40/50) — L1: 8 | L2: 10 | L3: 8 | L4: 6 | L5: 8. Arizona earns a perfect Layer 2 (Chips) score anchored by TSMC’s $165 billion investment — the largest single manufacturing investment in American history. Fab 21 in mass production on 4nm since Q4 2024; second fab construction completed April 2026; third fab breaking ground April 2025. Arizona’s Layer 2 position is the most durable in the country, reflecting the Five-Layer framework’s insight that chips are the most geopolitically sticky and strategically defensible layer. Water constraints mildly temper Layer 1; Layer 4 and 5 scores reflect a still-developing model and application ecosystem anchored by Arizona State University.
Washington State (40/50) — L1: 9 | L2: 5 | L3: 9 | L4: 8 | L5: 9. Washington earns a near-perfect Layer 1 score anchored by abundant and low-cost Columbia River hydroelectric power — a structural energy advantage that has attracted datacenter development for decades. Microsoft’s global headquarters, Amazon Web Services’ headquarters, and the University of Washington’s strong AI programs drive Layers 3, 4, and 5.
Tier 2: Emerging Intelligence Corridors (Score: 30–39)
Ten states occupy Tier 2, each anchoring significant strength in at least one to two layers while working to develop the others. Ohio (35) benefits from Intel’s New Albany Layer 2 investment and Columbus Stargate Layer 3 development. New York (32) scores highest on Layer 4 (Models) and Layer 5 (Applications) through Columbia, Cornell Tech, NYU, and its global financial AI ecosystem. Maryland (31) benefits from dense defense and intelligence Layer 5 AI deployment. Pennsylvania (33) anchors Layer 4 through Carnegie Mellon’s robotics and AI leadership and Layer 1 through its substantial nuclear generation capacity, while Compute and Chip infrastructure remain gaps. Massachusetts (31) mirrors Pennsylvania in its heavy Layer 4 weighting through MIT and Harvard, offset by limited Layer 2 and 3 development. Oregon (31) benefits from Columbia River hydro for Layer 1 and Intel’s Hillsboro manufacturing presence for Layer 2, combined with a growing Layer 3 datacenter cluster.
Tier 3: Developing States (Score: 20–29)
Fifteen states in Tier 3 possess meaningful assets in one or two layers but have not yet achieved the multi-layer integration necessary to function as coherent Intelligence Corridors. Colorado (29) benefits from its growing Denver-Boulder tech ecosystem in Layers 4 and 5 and strong Layer 1 energy development. Utah (28) anchors an emerging Layer 3 datacenter cluster in Salt Lake City, supported by competitive Layer 1 energy costs. Indiana (27) benefits from strong Layer 1 energy infrastructure and emerging manufacturing capacity. States like New Mexico (23) and Wyoming (20) have notable Layer 1 energy resources — including significant renewable and nuclear development — that could form the basis for Layer 3 datacenter attraction if their talent and connectivity gaps can be addressed.
Tier 4: Early-Stage States (Score: 10–19)
The remaining twenty states score between 10 and 19 on the Corridor Index. These states face the steepest climb toward Intelligence Corridor viability, typically due to limited energy infrastructure development capacity, minimal chip or datacenter presence, weak talent pipelines, and small markets for AI deployment. However, the Five-Layer AI Economy framework suggests that even Tier 4 states have a potential entry point: Layer 1 (Energy). States with significant renewable energy resources — wind in Iowa, Kansas, and the Dakotas; hydro in Maine and Vermont; solar in Louisiana and Mississippi — could position themselves as energy suppliers to the broader Intelligence Corridor economy, even if they do not develop indigenous Layers 2 through 5. This energy-first strategy may be the most realistic path to Corridor participation for the majority of Tier 4 states.
Cross-Cutting Findings from the Corridor Index
Several findings from the Corridor Index deserve particular emphasis. First, the Power dimension (Layer 1: Energy) is the most frequently binding constraint. States with strong Layer 4 (Models) or Layer 5 (Applications) scores are often limited in their overall corridor development potential by inadequate power infrastructure. This finding has clear policy implications: investments in electrical generation, transmission, and grid modernization are not merely energy policy — they are AI economic policy.
Second, Layer 2 (Chips) is the most geographically concentrated layer in the entire Corridor Index. Only Arizona (10/10) earns a perfect Layer 2 score. Only Ohio (7) and California (7) join Arizona above 5. This extreme concentration reflects the global reality that leading-edge semiconductor manufacturing is available in only a handful of locations on earth, and that the CHIPS Act investment cycle — concentrated in Arizona, Ohio, and a few other sites — will determine the domestic Layer 2 geography for decades to come.
Third, Layer 4 (Models) is the most unevenly distributed dimension relative to population. California, Massachusetts, New York, and Washington State account for the overwhelming majority of frontier AI model development activity. This extreme geographic concentration is partly a function of talent (PhD-producing universities), partly of capital (venture ecosystems), and partly of infrastructure proximity (hyperscaler data centers enabling fast iteration). Broadening the geographic distribution of Layer 4 activity is among the most important policy challenges for national AI competitiveness.
Fourth, the Five-Layer AI Economy framework reveals that the most durable corridor positions are anchored in the lower layers (Energy and Chips), not the upper layers (Models and Applications). This is the key structural insight for corridor strategy: states and regions that invest in Layer 1 and Layer 2 capacity today are building corridor positions that will compound over decades, while states that focus exclusively on Layer 5 Application attraction are building positions that are inherently more fragile and more easily disrupted by shifts in capital flows, company relocations, and technology transitions.
Conclusion
The artificial intelligence revolution is most commonly discussed in the language of algorithms, foundation models, data, and chips. These are real and important dimensions of the transformation underway. But they are not the whole story, and they may not be the most consequential story from the perspective of long-run economic geography.
The deeper transformation is physical and spatial. The Five-Layer AI Economy — Energy, Chips, Datacenters, Models, and Applications — is a stacked industrial system in which each layer depends on the one below it and constrains the one above it. This vertical dependency structure means that the geography of the AI economy is determined from the bottom up: the location of energy infrastructure determines where chips can be manufactured; the location of chip manufacturing determines where datacenters can scale; the location of datacenters determines where models can be trained; and the location of models determines where the most powerful applications can be built.
Intelligence Corridors are the regional ecosystems in which two or more of these layers achieve sufficient concentration and integration to generate self-reinforcing cycles of intelligence production and economic value. The Corridor Index presented in this paper — applying the Five-Layer AI Economy as the definitive measurement standard across all fifty states — reveals that the geographic concentration of Intelligence Corridor capacity is already extraordinary. The top five states account for roughly 65 percent of the nation’s current AI infrastructure capacity, and the top ten states account for roughly 80 percent.
This paper has drawn on intellectual traditions spanning three decades to make this argument. From Saxenian’s insight that the social architecture of regions determines technological vitality, through Porter’s demonstration that clusters generate competitive advantages no individual firm can replicate, through Storper’s theorization of regions as nexuses of untraded interdependencies, through Glaeser’s empirical confirmation that density generates knowledge spillovers, through Moretti’s documentation of the brain hub multiplier effect, through Florida’s argument that talent follows quality of place, through Brynjolfsson and McAfee’s insight that the second machine age amplifies geographic divergence, through Acemoglu’s warning that the path of technology is a choice, through Muro’s empirical documentation of AI’s winner-takes-most geography, and through the Five-Layer AI Economy framework that provides the definitive structural map of where that geography is being contested — the cumulative weight of this intellectual tradition points in a single direction: place matters, geography matters, and in the AI era, geography matters more than ever.
The future winners of the AI era may not be individual companies. They may be entire corridors. And the corridors that will win are those that understand the Five-Layer AI Economy — and invest accordingly in Energy, Chips, Datacenters, Models, and Applications, in that order of structural priority.
Notes, Footnotes, and Bibliography
[1] AnnaLee Saxenian. Regional Advantage: Culture and Competition in Silicon Valley and Route 128. Cambridge, MA: Harvard University Press, 1994.
https://www.hup.harvard.edu/books/9780674753402
[2] Michael Storper. The Regional World: Territorial Development in a Global Economy. New York: Guilford Press, 1997. Attribution by Ash Amin, Professor of Geography, University of Durham.
[3] Michael E. Porter. “Clusters and the New Economics of Competition.” Harvard Business Review 76, no. 6 (November–December 1998): 77–90.
https://www.hbs.edu/faculty/Pages/item.aspx?num=46852
[4] Edward Glaeser. Triumph of the City. New York: Penguin Press, 2011. As summarized by LSE Review of Books (2012).
[5] Richard Florida. The Rise of the Creative Class. New York: Basic Books, 2002. Summary at Hill Strategies Research.
[6] Erik Brynjolfsson and Andrew McAfee. The Second Machine Age. New York: W. W. Norton & Company, 2014.
https://betterworld.mit.edu/spectrum/issues/spring-2015/the-second-machine-age
[7] Daron Acemoglu and Simon Johnson. Power and Progress: Our 1,000-Year Struggle Over Technology and Prosperity. New York: PublicAffairs, 2023.
https://news.mit.edu/2023/power-and-progress-book-ai-inequality-0517
[8] Jensen Huang, Founder and CEO, NVIDIA. Talk at Stanford Graduate School of Business, April 16, 2026.
[9] Dr. Stefanus Hadi. “Five-Layer AI Economy: What Is It? Who Are the Key Players? — Mapping Power, Influence, and Value Across the Full Stack of Intelligence.” Stefanus.AI, April 2026.
[10] Mark Muro, Senior Fellow, Brookings Institution. Quoted in “How AI Could Reshape the Economic Geography of America.” Las Vegas Sun, December 27, 2024.
https://lasvegassun.com/news/2024/dec/27/how-ai-could-reshape-the-economic-geography-of-ame
[11] International Energy Agency (IEA). Key Questions on Energy and AI. Paris: IEA, April 2026.
https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary
[12] International Monetary Fund. “Global Economic and Financial Implications of Artificial Intelligence.” IMF Notes, Vol. 2026, Issue 002 (April 2026).
https://www.elibrary.imf.org/view/journals/068/2026/002/article-A001-en.xml
[13] Rolf Landauer. “Irreversibility and Heat Generation in the Computing Process.” IBM Journal of Research and Development 5, no. 3 (1961): 183–191. Popularized in “Information is Physical.” Physics Today (1991).
[14] McKinsey Global Institute. “Data Centers and the AI Economy.” McKinsey & Company, 2024.
[15] Stanford Human-Centered AI Institute. AI Index Report 2024. Stanford University, 2024.
https://hai.stanford.edu/ai-index/2024
[16] ASML Holdings N.V. Annual Report 2024 and company technology documentation.
[17] Dr. C.C. Wei, Chairman and CEO, TSMC. SEC Form 6-K Filing, March 2025.
https://www.sec.gov/Archives/edgar/data/0001046179/000104617925000024/tsmcexpandinvestmentintheu.htm
[18] Jensen Huang, NVIDIA. Q1 FY2027 Earnings Call, May 20, 2026.
[19] Jensen Huang, NVIDIA. Q1 FY2027 Earnings Call, May 20, 2026. Reported by CNBC.
https://www.cnbc.com/2026/05/20/nvidia-nvda-earnings-report-q1-2027.html
[20] Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B. Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, Dario Amodei. “Scaling Laws for Neural Language Models.” OpenAI and Johns Hopkins University, arXiv:2001.08361 (January 2020).
https://arxiv.org/abs/2001.08361
[21] Stanford Human-Centered AI Institute. AI Index Report 2024. Stanford University, 2024.
https://hai.stanford.edu/ai-index/2024
[22] World Bank. World Development Report 2019: The Changing Nature of Work. Washington, D.C.: World Bank, 2019.
https://www.worldbank.org/en/publication/wdr2019
[23] Andrew Ng. “The Batch.” deeplearning.ai newsletter (2024). See also: Andrew Ng, “AI Agentic Workflows.”
https://www.deeplearning.ai/the-batch
[24] Suhas Subramanyam, former Virginia State Senator. Quoted in InformationWeek, “Power Struggles in Data Center Alley,” 2025.
[25] Sam Altman, CEO, OpenAI. Stargate five new sites announcement, September 23, 2025.
https://openai.com/index/five-new-stargate-sites
[26] Theresa Mayer, Vice President for Research, Carnegie Mellon University. CMU News, February 2026.
[27] Michael E. Porter. “Clusters and the New Economics of Competition.” Harvard Business Review 76, no. 6 (November–December 1998).
https://www.hbs.edu/faculty/Pages/item.aspx?num=46852
[28] Mark Muro, Senior Fellow, Brookings Institution. Quoted in Route Fifty, “Usual Cities Dominate AI Readiness, but More Are Rising,” July 2025.
Additional Sources Consulted
Enrico Moretti. The New Geography of Jobs. Boston: Houghton Mifflin Harcourt, 2012. https://www.gsb.stanford.edu/insights/enrico-moretti-geography-jobs
Allen J. Scott. Technopolis: High-Technology Industry and Regional Development in Southern California. Berkeley: University of California Press, 1993.
Michael E. Porter. The Competitive Advantage of Nations. New York: Free Press, 1990.
Mark Muro and Julian Jacobs. “The Case for Promoting the Geographic and Social Diffusion of AI Development.” Brookings Institution, September 2024. https://www.brookings.edu/articles/the-case-for-promoting-the-geographic-and-social-diffusion-of-ai-development/
Mark Muro. “AI Seems Everywhere, but Regional Readiness Is Uneven.” Brookings Institution, 2025. https://www.brookings.edu/articles/ai-seems-everywhere-but-regional-readiness-is-uneven/
OpenAI. “Announcing The Stargate Project.” January 2025. https://openai.com/index/announcing-the-stargate-project/
TSMC Arizona. https://www.tsmc.com/static/abouttsmcaz/index.htm
NVIDIA SEC Filings Q4 FY2025 and Q1 FY2027. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001045810
IEA. “Data Centre Electricity Use Surged in 2025.” April 2026. https://www.iea.org/news/data-centre-electricity-use-surged-in-2025-even-with-tightening-bottlenecks-driving-a-scramble-for-solutions
American Action Forum. “Virginia’s New Data Center Electricity Rate Class.” January 2026. https://www.americanactionforum.org/insight/virginias-new-data-center-electricity-rate-class/
Pittsburgh Robotics Network. Robotics & AI Discovery Day 2026. https://www.prnewswire.com/news-releases/robotics–ai-discovery-day-2026-spotlights-pittsburgh-at-the-center-of-a-10b-innovation-ecosystem-302736124.html
World Bank. Beyond the AI Divide. 2025. https://documents1.worldbank.org/curated/en/099517502242572646/pdf/IDU-5c1b47f8-89d8-4064-8b72-329105034043.pdf
Nature Communications. “The Global Geography of Artificial Intelligence in Life Science Research.” Vol. 15 (2024). https://www.nature.com/articles/s41467-024-51714-x
Dr. Stefanus Hadi. Five-Layer AI Economy framework and related publications. Stefanus.AI, 2026. https://stefanus.ai
