The hidden alignment between "old industrial infrastructure" and "new computing demands" is the true starting point of the Rust Belt's rebirth.
Author and source: Fudan Business Knowledge
Fast Reading
- On July 15, 2025, Trump announced at Carnegie Mellon University in Pittsburgh that over $90 billion in private investment is flowing into Pennsylvania—$25 billion from Google, $25 billion from Blackstone, $20 billion from Amazon, and $16 billion from Microsoft—transforming this traditional industrial state, long labeled the “Rust Belt,” into a “Smart Belt.”
- Pennsylvania’s transformation is no accident: its massive energy output capacity (the largest net electricity exporter in the U.S.), abandoned industrial land, talent pool from Carnegie Mellon University, and the rigid baseload power demands of AI data centers have historically aligned—a real-world example of dynamic capabilities theory.
- On a deeper level, this is a national-level awakening of dynamic capability: U.S. energy policy has shifted from “climate first” to “AI first,” with the federal government amplifying local dynamic capabilities through regulatory relaxation. In any organization, what matters is not “what you own,” but whether you can find new value coordinates for dormant assets.
On July 15, 2025, U.S. President Trump, along with executives from numerous technology and energy companies, financiers, and politicians, gathered on the campus of Carnegie Mellon University in Pittsburgh, Pennsylvania, to launch a global competition for AI infrastructure. Trump announced at the event that over $90 billion in private investment would flow into this traditional industrial state, long labeled as part of the "Rust Belt."
Let’s break down the $90 billion in detail: Google is investing $25 billion in building data centers; Blackstone has committed $25 billion to develop AI infrastructure; Amazon AWS plans to invest over $20 billion to construct two data centers in Pennsylvania, one of which is located next to a nuclear power plant; Microsoft has signed a $16 billion agreement to restart the Three Mile Island nuclear plant; data center specialist CoreWeave is investing $6 billion to build an AI data center in Lancaster; and Westinghouse has even announced plans to construct ten next-generation nuclear reactors specifically to provide clean power for data centers.
The transformation of the "Rust Belt" is the result of resource- and capability-driven factors. The resource-based view in management tells us that a firm’s competitive advantage stems from its possession of scarce, difficult-to-imitate resources—such as patented technologies, brand reputation, or unique human capital.
However, as the way the economy operates begins to change, previously valuable resources may suddenly lose value, and traits once seen as disadvantages may transform into new competitive advantages. Therefore, in a rapidly changing environment, “what resources you have” is less important than “how you reconfigure your resources.” This is the core of dynamic capabilities theory.
The importance of building dynamic capabilities holds equally true at the regional and national levels. Pennsylvania’s geography, workforce, and energy resources once formed the backbone of American industrialization and now serve as core assets for developing AI infrastructure. As Pennsylvania Senator Dave McCormick stated at the summit: “The victory in the battle for AI innovation will go to those states that can provide computing power, electricity, and talent—and Pennsylvania is at the center of this competition.”
Why Pennsylvania? The hidden alignment between "old industrial infrastructure" and "new computing demands"
In their book "Smart Transformation: The Economic Miracle from Rust Belt to Smart Belt," "Father of Emerging Markets" Antoine van Agtmael and former CEO of Financial Daily Fred Bakker describe the Rust Belt as a sleeping beauty waiting for a prince to awaken her potential.
More than a century ago, the sleeping beauty once stood at the center of the stage. In the first half of the 19th century, one of the largest anthracite coal deposits in the United States was discovered in northeastern Pennsylvania, becoming the fuel that powered the entire American Industrial Revolution. With abundant coal and iron ore resources, Pittsburgh in the southwest, driven by industrial titans like Andrew Carnegie, emerged as the world’s steel capital.
After the 1970s, the waves of globalization and deindustrialization dealt a severe blow to Pennsylvania, as steel mills shut down one after another and unemployment soared into double digits. Abandoned factories, rusted railway tracks, and nostalgic industrial sites became the new defining features of the region—giving rise to the term “Rust Belt”—as once-bustling cities fell silent.
But the fact that the sleeping beauties lie motionless does not mean they have lost everything—their inherent strengths remain: energy, expertise, knowledge, and potential. The awakening of the sleeping beauties typically requires the arrival of a new character. In today’s context, that new character is AI.
In the past two years, the rapid advancement of AI has produced an unexpected side effect: an energy crisis. Training a large AI model requires electricity equivalent to the annual consumption of hundreds of households, while the data centers running these models demand 24/7 baseload power. According to 2025 data from the International Energy Agency (IEA), global data center electricity demand is projected to more than double by 2030, reaching approximately 945 terawatt-hours—slightly exceeding Japan’s total national electricity consumption. The United States and China are experiencing the most significant increases in data center power consumption, expected to account for nearly 80% of global growth.
The innovation stories we once knew—where a genius or a couple of geeks built something in a garage—no longer fit this era; the high demand for electricity has pulled tech companies back from the cloud into the physical world. Securing a stable and sufficient energy supply has become a critical component of the AI race.
Pennsylvania is the largest net electricity exporter in the United States, generating a total of 241.5 million megawatt-hours in 2024, of which approximately 80 million megawatt-hours were exported to other states. It is also the second-largest natural gas-producing state in the U.S., accounting for 20% of the nation’s natural gas output. More importantly, according to Jon Gray, President of Blackstone, Pennsylvania can “build data centers directly next to power sources,” a geographic-energy coupling advantage that reduces the need for expensive transmission lines—this is Pennsylvania’s greatest competitive advantage.
This land, once steeped in American industrial history, has now found a new role on the stage of the AI era.
In addition to its natural energy advantages, assets accumulated during past development have also prepared Pennsylvania for its transition.
Abandoned industrial sites are ideal locations for data centers—already connected to the power grid, conveniently accessible by transportation, and with land costs far lower than those in Silicon Valley; engineers and technicians formerly employed in steel plants form a skilled workforce for data center operations; Carnegie Mellon University, a world-leading institution in AI research, has transformed from an academic ivory tower into a driver of industrial transformation.
The development of AI itself requires interdisciplinary collaboration, emphasizing knowledge sharing among technology, teams, and organizations, and necessitating close coordination among businesses, educational institutions, and government agencies. In this process, Pennsylvania’s advantages become even more pronounced: policymakers and industry leaders have keenly recognized a historic alignment opportunity between the steady, baseload power demands of AI data centers and Pennsylvania’s substantial power generation capacity, with the state’s industrial-era infrastructure providing a unique physical platform for this alignment.
This means that Pennsylvania not only possesses resources but also has the ability to redefine their value and transform them into a competitive advantage. Compared to other states in the Rust Belt, Pennsylvania can proactively connect the technological logic of the AI industry with its regional resource strengths—this exemplifies the first dimension of dynamic capability: sensing—the organization’s continuous scanning of the external environment to identify technological inflection points, market restructuring, and emerging demands, thereby uncovering truly significant trends.
Identifying the hidden alignment between "legacy industrial infrastructure" and "new computing demands" is the true starting point of the Rust Belt's rebirth.
A genuine transformation that comprehensively updates systems, spaces, and identities.
Recognizing an opportunity is the first step; what truly tests your ability is how you turn that opportunity into reality.
In September 2023, Pennsylvania Governor Josh Shapiro signed an executive order establishing the Generative AI Oversight Committee to oversee AI policies and implementation within government. This was among the earliest initiatives in U.S. states to elevate AI governance to the executive level.
In January 2024, Pennsylvania partnered with OpenAI and Carnegie Mellon University to launch the nation’s first state-government-level ChatGPT pilot program. The results showed that participants saved an average of eight hours per week—providing strong evidence of the practical value of AI tools and sending a clear message to the public that Pennsylvania is embracing AI.
In April 2026, Shapiro announced that the use of generative AI had expanded to over 3,000 state employees across 35 agencies, with more individuals signing up for training.
On the tax and fiscal side, Pennsylvania enacted the Computer Data Center Equipment Program as early as 2016, offering a capped sales tax rebate for eligible computer data center equipment. In 2021, the tax exemption was further expanded by removing the cap, allowing qualifying data centers to be exempted directly from the 6% state sales tax when purchasing servers, cooling systems, and software.
In response to local residents' concerns about potential electricity price increases and environmental impacts from large-scale data center construction, Shapiro unveiled a set of certification standards for digital infrastructure in May 2026, called GRID (Governor's Responsible Infrastructure Development). These standards require data centers to meet four criteria—energy self-sufficiency (without displacing residential power use), community transparency, local hiring, and environmental protection—to qualify for financial support and gain access to an expedited approval process.
The synergy among policies provides Pennsylvania with a sustainable institutional environment that balances diverse interests, serving as a toolkit for capturing AI opportunities.
A more tangible transformation occurs at the physical level. Vast tracts of abandoned industrial land, once considered “liabilities” in urban renewal due to severe contamination, outdated facilities, and high renovation costs, have suddenly become “strategic assets” in the context of the AI era—they already have grid connectivity, well-developed transportation infrastructure, are located away from residential areas, and offer land costs far lower than those of newly built campuses. Converting existing industrial infrastructure into data centers significantly reduces investment costs and timeframes, while also minimizing environmental impacts associated with constructing new infrastructure.
In 2025, Amazon announced an investment of $20 billion to build two data center campuses in Pennsylvania, one of which is located within the Fairless Hills and Keystone Trade Center logistics park, formerly the site of a U.S. Steel plant.
Another data center is built next to the Susquehanna nuclear power plant in northeastern Pennsylvania, challenging the traditional internet investment logic of lightweight asset deployment.
Tech companies aim to connect energy-intensive data centers directly to power plants, bypassing congested grids and shortening development timelines by several years. Talen Energy, a major shareholder in the Susquehanna nuclear plant, has a dedicated power supply agreement with Amazon for a data center adjacent to the plant, ultimately providing 960 megawatts—equivalent to 40% of the output of one of the largest nuclear plants in the U.S.—sufficient to power over 500,000 homes. However, this “behind-the-meter” connection, which bypasses the public grid, has been blocked by the Federal Energy Regulatory Commission (FERC) on procedural grounds, as universal regulations for co-locating large loads with generation facilities have yet to be established.
Another nuclear power plant along the Susquehanna River is even more representative. With financial support from Microsoft, Constellation Energy, the largest nuclear operator in the United States, is restarting the Three Mile Island nuclear plant, and Microsoft has secured exclusive rights to purchase all of the plant’s electricity for the next 20 years after it resumes operations. This marks the first time in U.S. history that a permanently shut down nuclear reactor has been restarted, and the first time an entire commercial nuclear power plant’s output has been allocated to a single customer. Three Mile Island is expected to resume operations in 2028.
Microsoft’s choice of Three Mile Island was no accident. AI data centers have rigid demands for stable, clean, large-scale baseload power, which intermittent solar and wind energy cannot meet to ensure “always-on” electricity. In an era defined by AI, companies are willing to pay for electricity—this is, in essence, a market-driven transformation of the energy structure.
From policy support to the integration of resources from various stakeholders, Pennsylvania demonstrates two additional dimensions of dynamic capability: seizing and transforming—the ability to convert perceived opportunities into concrete actions and systematically reconfigure organizational (or regional) asset allocations, business processes, and institutional frameworks to adapt to a new competitive environment.
Thus, the outline of the smart belt gradually becomes clear.
National-level dynamic capabilities make way for AI in U.S. energy policy
Looking deeper, Pennsylvania’s transformation from the “Rust Belt” to the “Smart Belt” is not only a demonstration of local government’s dynamic capabilities but also reflects a broader awakening of dynamic capabilities at the national level across the United States.
A most direct manifestation is the shift in U.S. energy policy. The restart of the Three Mile Island nuclear plant is not merely a state-level decision but has received approval from the federal Nuclear Regulatory Commission and policy endorsement from the Department of Energy. A series of executive orders issued by the Trump administration in 2025 effectively "unshackled" the nation’s energy strategy: streamlining nuclear licensing procedures, expanding natural gas extraction permits, and prioritizing power access for data centers. Although this policy shift—from “climate first” to “AI first”—has sparked controversy among environmental groups, it demonstrates the U.S. government’s ability to rapidly reallocate resource priorities in response to intensifying global AI competition.
Therefore, a nation’s dynamic capabilities are not only reflected in macroeconomic policy adjustments but also in institutional designs that create space for state and local governments and businesses to act. The federal government’s role is not to replace the perception and capture efforts of state governments and enterprises, but rather to amplify their dynamic capabilities through regulatory relief, financial support, and strategic coordination. In this sense, Pennsylvania’s $90 billion investment is both a product of Pennsylvania’s own dynamic capabilities and a manifestation of America’s national dynamic capabilities at the local level.
In organizations of any size, there exist "sleeping assets"—whether in technology, space, or talent. The key to dynamic capability lies not in what you possess, but in your ability to identify new sources of value for these assets within the context of technological change and market restructuring.
In this rapidly changing world, breaking free from inertia, sensing shifts, seizing opportunities, and having the courage to rebuild your systemic capabilities are the competitive advantages for the future.
Reference materials
[1] Helfat, C. E., & Peteraf, M. A. (2015). Managerial cognitive capabilities and the microfoundations of dynamic capabilities. Strategic Management Journal, 36(6), 831–850.
[2] Anthony Van Agtmael, Fred Bakker. Smart Transformation: The Economic Miracle from the Rust Belt to the Smart Belt [M]. Xu Yizhou, trans. Beijing: CITIC Press, 2017.
Trace it back to its source
Dictionary of Management
To better understand complex business phenomena, we need to clarify fundamental concepts. Regarding the content of this article, we found the following relevant definitions in the Dictionary of Management, for your reference:
Dynamic Capabilities Theory
Dynamic capability theory
The theory that emphasizes exploring the definition of the firm and the sources of its competitive advantage from the perspective of dynamic capabilities is an evolution of the “resource-based view” of the firm, developed in the early 1990s in the United States. Key figures include David J. Teece (b. 1948) and his works, “Concepts of Dynamic Capabilities, Resources, and Strategy” (1990) and “Dynamic Capabilities and Strategic Management” (1997).
The main points are: a firm is a dynamic system composed of processes, routines, and resources; its competitive advantage arises from effectively leveraging its management and organizational processes, shaped by the strategic positioning of its assets and capabilities; and its long-term competitiveness depends on its "dynamic capabilities."
所谓“动态能力”是指企业持续更新自身能力以适应变化的商业环境的能力；企业能否持续创造竞争优势，很大程度上取决于其现有的能力基础。