Electrons and Intelligence: The Structural Reshaping of the Global Economy

The structural drivers of economic expansion are undergoing a fundamental reset. For decades, housing, manufacturing, and consumer spending served as the primary bellwethers. Today, a new dual engine is taking the lead: the physical infrastructure of artificial intelligence and the software intelligence it runs on. This investment boom is not just a market theme; it is becoming a direct, measurable force in the economy, reshaping growth trajectories and challenging the very frameworks used to measure them.

The most concrete evidence lies in capital expenditure. In the first half of 2025, AI-related capital spending contributed 1.1 percentage points to GDP growth, outpacing the U.S. consumer as an engine of expansion. This surge is concentrated in hardware, where investment in computers and related equipment jumped 41% year-over-year. The scale is staggering, driven by the hyperscalers-Meta, Alphabet, Microsoft, Amazon, and Oracle-projected to allocate $342 billion to capex in 2025, a 62% increase from the prior year. This initial phase has focused on servers and networking, with data center construction hitting a record $40 billion annual rate in June.

Yet this is only the first act. The next phase is a transition to supporting infrastructure, a shift that will determine the longevity of this growth wave. The buildout is now targeting the power plants and grid upgrades needed to fuel the AI boom, a process that can take years to plan and permit. This creates a structural gap: the investment is real and accelerating, but its full economic impact on GDP is delayed and potentially muted. Not every AI dollar translates directly to domestic output, as much spending goes toward imported technology goods, and data centers themselves employ relatively few workers, limiting wage-driven consumption multipliers.

This boom is part of a broader, deeper trend. The source of modern growth is increasingly intangible capital-software, organizational complements, and now, AI models themselves. Official GDP data, however, struggle to capture this reality. Research shows that failure to incorporate intangible capital leads to a systematic underestimate of economic activity. In the AI era, this problem is acute. Under current accounting rules, much of the investment in AI models, datasets, and internal deployment systems is treated as an operating expense, not as capital investment. This creates a classic J-curve dynamic: upfront costs depress measured productivity, while the output gains from AI integration materialize only after firms have redesigned workflows and retrained staff. The result is that the true scale of productive capacity being built is invisible in official statistics.

The bottom line is a new growth paradigm. The economy is being reshaped by a dual investment boom, but its measurement lags behind its reality. The physical infrastructure of electrons and the software intelligence they carry are the twin engines, yet traditional GDP calculations are poorly equipped to measure their full impact. This creates both a risk of underestimating growth potential and a vulnerability if policy is based on incomplete data. The structural shift is clear; the challenge now is to build measurement frameworks that can keep pace.

The Measurement Gap and Its Consequences

The structural reshaping of the economy by AI creates a parallel challenge: our ability to see it. Official GDP calculations, built for an era of tangible goods, are fundamentally ill-equipped to track this new paradigm. The result is a persistent measurement gap that risks distorting both policy and investment decisions.

The core issue is the treatment of intangible capital. As the knowledge economy has grown, so has the share of investment in software, organizational complements, and now, AI models themselves. Research shows that failure to incorporate intangible capital leads to a systematic underestimate of GDP. In the AI era, this problem is acute. Under current accounting rules, the vast majority of investment in AI models, datasets, and internal deployment systems is treated as an operating expense, not as capital investment. This creates a classic J-curve dynamic: upfront costs depress measured productivity in the short run, while the output gains from AI integration materialize only after firms have redesigned workflows and retrained staff. The capital stock is understated throughout this adjustment period, making growth accounting appear weaker than it truly is.

This gap extends beyond capital to the value being created. The scalable nature of digital services and models often bypasses traditional market transactions. A company's proprietary AI model or a free, bundled tool like ChatGPT generates immense value without a clear price tag in official statistics. This "missing value" is a blind spot in economic measurement, leaving over $3 trillion in intangible capital stock invisible in official data. The consequence is a distorted view of productivity and growth, where the true scale of productive capacity being built remains hidden.

This uncertainty fuels a critical debate. With AI adoption now widespread-55% of people and 37% of workers in the U.S. used generative AI tools as of August 2025-there is an ongoing discussion about whether the promised productivity gains are already being realized. The measurement gap makes this hard to answer definitively. Are we seeing the payoff, or are we still in the costly, expensing phase of the J-curve? Without clear data, policymakers struggle to gauge the economy's true health and allocate resources effectively. Investors, in turn, face difficulty assessing the real returns on massive capex commitments.

The path forward requires new frameworks. A near-term solution is a Generative AI Intensity Index that combines statistical surveys with provider telemetry to track adoption. In the medium term, national accounts themselves must be reformed to separately identify AI-related capital, services, and labor reallocation. Until then, the measurement gap will persist, creating a lag between the real, structural reshaping of the economy and the data used to manage it.

Geopolitical Competition: The Infrastructure vs. Intelligence Bet

The global race for AI supremacy is no longer a contest of pure software prowess. It has evolved into a strategic wager on the form of investment, with the United States and China making fundamentally different bets that will shape their economic competitiveness for decades. The U.S. is making a leveraged wager on AI model development, while China is strategically investing in the physical capacity to power and deploy these models at scale. This divergence creates a critical vulnerability for the American strategy.

On one side, the U.S. leads in top-tier model capabilities. The defining question of 2025, as one analysis noted, was whether intelligence alone would be enough. Silicon Valley's answer was a resounding yes, betting everything on building the most capable models. This focus is reflected in private investment, where U.S. firms poured $109.1 billion into AI in 2024, dwarfing China's $9.3 billion. Yet this leadership is not guaranteed. China is rapidly closing the performance gap, indicating a competitive race where the U.S. advantage is eroding. The strategic bet, however, is shifting.

China's approach is to treat intelligence as a commodity and focus on the infrastructure to make it useful. This means dominating the manufacturing of the physical components that convert electricity into motion and work-the so-called "Electric Stack" of batteries, magnets, power electronics, and embedded compute. This industrial ecosystem gives China a critical edge in deploying AI at scale across physical systems, from electric vehicles to robotics. The U.S. strategy, by contrast, risks ceding much of this physical capacity, creating a dependency on imported components for its own AI ambitions.

Policy actions in Washington are attempting to secure the U.S. technological lead, but they are largely defensive. The White House has issued a series of executive orders aimed at removing barriers to AI leadership. These include accelerating federal permitting for data center infrastructure and promoting the export of American AI technologies. While these moves aim to speed up the build-out of the physical stack and capture global market share, they are reactive to a competitive landscape where China is already ahead in manufacturing that stack. The January 2025 order explicitly revoked previous actions that were seen as paralyzing the industry, signaling a clear intent to foster innovation and deployment.

The bottom line is a structural mismatch. The U.S. is betting on the intelligence layer, while China is building the infrastructure to power and embed it. This creates a classic "leverage vs. capacity" trade-off. The U.S. may lead in the most advanced models, but China's dominance in the physical stack ensures it can deploy those models at scale and integrate them into the global economy. For American firms and policymakers, the challenge is to bridge this gap, ensuring that the country's leadership in AI software is matched by the industrial capacity to realize its economic potential.

Forward Scenarios and Key Catalysts

The dual-engine growth model faces a fork in the road. The path forward hinges on a handful of critical catalysts and risks, with three plausible scenarios emerging from recent analysis. The baseline outlook assumes sustained AI investment, but the real test will be whether this momentum can be maintained through a period of elevated uncertainty.

Deloitte's outlook outlines three distinct trajectories. The baseline scenario reflects the current consensus, modeling continued AI-driven expansion. However, the primary risk is a sudden pullback in investment, which would be modeled as a downside scenario. Such a reversal could slow GDP growth and increase unemployment, as the economy's reliance on this capex boom becomes more apparent. The upside case, conversely, depends on a faster-than-expected payoff from AI productivity gains, which would accelerate the J-curve recovery. The key variables driving these divergences are not just AI spending itself, but the broader policy environment. Scenarios assume higher tariffs persist, with the average effective rate rising to 15% by early 2026, and that net migration remains lower than anticipated, both of which would weigh on economic output.

A critical catalyst for the entire setup is the transition in Federal Reserve leadership. As the central bank's next chair takes the helm, their approach to data-driven policy will be put to the test. The debate over AI productivity gains is already intersecting with labor market dynamics, prompting the Fed to balance inflation pressures with emerging structural shifts. The upcoming chair transition could create a period of elevated uncertainty, emphasizing the need for clear communication to stabilize markets. The Fed's stance will be crucial in determining whether monetary policy supports or hinders the investment cycle.

For investors, the sustainability of the growth engine depends on monitoring complementary investments. The initial capex surge in hardware is now giving way to a need for parallel spending in software, research and development, and new data centers. The pace of AI adoption across different sectors, which is already affecting GDP numbers, will be a key indicator. As of August 2025, 55% of people and 37% of workers in the U.S. used generative AI tools, but the economic impact will deepen only with further integration. A new measurement framework is essential to track this evolution. The proposed Generative AI Intensity Index offers a near-term tool to combine surveys with provider telemetry, providing a more timely gauge of adoption than lagging official statistics.

The bottom line is one of structural opportunity shadowed by policy risk. The U.S. has made a leveraged bet on AI intelligence, but its success depends on maintaining investment momentum and navigating a complex policy landscape. The scenarios show that the economy's growth trajectory is no longer set by consumer spending alone, but by the interplay of technology investment, immigration, trade, and the next chapter of Fed leadership.