AI datacenter investments are now showing up in US GDP data.
According to new analysis from Stripe — sourced from the Bureau of Economic Analysis and the Congressional Budget Office — computer and software demand accounted for 46% of real potential US GDP growth in 2025. That is not just a record high; it is a figure that dwarfs every previous year going back to the late 1960s, including the peak of the dot-com boom.
The chart, highlighted by venture firm a16z, tells a striking story. For decades, tech’s share of US GDP growth oscillated in a band roughly between 10% and 35%, spiking during the PC era and the internet build-out of the 1990s, crashing into negative territory during the dot-com bust of 2001–2002, and then recovering modestly through the 2010s. But the 2025 reading of 46% represents a discontinuity unlike anything seen before — comparable in magnitude to the dot-com collapse, but flipped entirely in the positive direction.
What Is Driving This Surge
The answer, in large part, is AI infrastructure. Massive capital expenditure on data centers, GPUs, and the networking equipment required to train and run large language models has translated into real, measurable investment in the US economy. This is no longer speculative — it is showing up in official government statistics.
Companies like Microsoft, Google, Amazon, and Meta have collectively pledged hundreds of billions of dollars in AI-related capital expenditure over the coming years. That spending creates demand for semiconductors, electrical infrastructure, cooling systems, and real estate — all of which register as economic output. OpenAI, which recently announced plans to expand its physical footprint, is emblematic of a broader pattern: AI companies are moving from pure software plays to capital-intensive physical infrastructure builders.
A Structural Shift, Not a Blip
One of the most important questions analysts are now asking is whether 2025 marks a one-time spike or the beginning of a sustained structural shift. The evidence increasingly suggests the latter. AI model development is computationally intensive in ways that prior software cycles were not. Training frontier models requires enormous, purpose-built clusters of chips running continuously for months. Inference — actually running these models at scale for millions of users — requires persistent, always-on infrastructure.
The idea generation and productivity gains unlocked by AI tools are also accelerating software development itself, which feeds back into more demand for compute. We are seeing a reinforcing loop: better AI tools → faster software creation → more applications → more compute demand → more investment.
The Dotcom Parallel — And Why It May Not Apply
A16z’s commentary on the chart draws an explicit comparison to the dot-com era, describing 2025’s jump as “a Dotcom crash amount of year/year discontinuity, but in the positive direction.” That framing is useful but should be handled carefully.
The dot-com bubble was characterized by massive investment in infrastructure that was, in many cases, economically premature — fiber laid before demand existed, companies funded before viable business models emerged. The question for AI is whether the same dynamic is at play. The counterargument is that AI models are already demonstrably capable of performing economically valuable tasks at scale, from software development to customer service to scientific research. Demand for AI services, measured in user growth and enterprise adoption, is real — Gemini’s web traffic, for instance, grew 643% year-over-year in February 2026, while ChatGPT’s grew 37% in the same period.
That said, the risk of overbuilding cannot be dismissed. If AI revenue growth fails to keep pace with infrastructure investment, some portion of current capex could prove to have been misallocated.
The Bottom Line
The 46% figure is not a rounding error or a statistical quirk. It reflects a genuine, unprecedented surge in technology-driven economic activity, powered by the AI investment cycle that began in earnest in 2023 and accelerated sharply through 2024 and 2025. Whether this proves to be the foundation of a durable new growth era or the leading edge of a capital cycle correction remains to be seen. But for now, the data is unambiguous: the AI boom is the single largest driver of US economic growth on record.