On April 1, St. Louis Fed economists Miguel Faria-e-Castro and Serdar Ozkan published a blog post with a restrained title and striking conclusion: AI optimism itself is an inflationary driver. It is not because electricity prices have risen, or because of chip shortages, but because everyone believes AI will make the future better—this belief leads them to spend more money now.

On the same day, Fortune revealed an experiment by Deutsche Bank: they had three AI models assess "the impact of AI on inflation". The conclusion was that even AI itself believes it is driving up prices.

Posts about soaring prices in the United States are abundant on social media.

These two events combined point to an uncomfortable cycle: the more investment in AI, the higher the inflation, the further interest rate cuts are, and the higher the financing costs—yet investment continues to accelerate.

An Unstoppable Arms Race

First, let's look at the money. According to the financial reports of various companies, the combined capital expenditure of Amazon, Microsoft, Google, and Meta in 2023 is about $152 billion. By 2024, this figure jumps to $251 billion, an increase of 65%. For the entirety of 2025, it settles at $416 billion, another increase of 66%.

The corporate guidance for 2026 is even more aggressive. According to Wolf Street’s summary, Amazon guides $200 billion, Google guides $175 billion to $185 billion, Microsoft guides $145 billion to $150 billion, and Meta guides $135 billion. Together, the four companies total about $663 billion. If Oracle's $42 billion is included, the five companies approach a total of $700 billion.

In four years, the capital expenditures of these four companies have quadrupled. This growth rate has no precedent in the history of American enterprises. According to Fortune, this scale has already surpassed Sweden's entire annual GDP.

A Data Center Consuming as Much Power as an Entire State

Most of this money is flowing into data centers. The biggest bottleneck for data centers is not land, but electricity. According to EIA data, Vermont's total electricity consumption for the year is about 5,364 GWh, averaging a load of 0.61 GW. Rhode Island is slightly higher, at about 0.83 GW.

Now let's see what the data centers are doing. According to company announcements, the Stargate project by OpenAI in collaboration with Oracle and SoftBank has a total planned power capacity of 10 GW, equivalent to the total electricity consumption of 16 Vermonts. Meta's Hyperion complex in Louisiana is planned for 5 GW, with an investment of $27 billion. Elon Musk's xAI in Memphis, Tennessee has expanded to 2 GW at Colossus, deploying 555,000 NVIDIA GPUs at a cost of about $18 billion, according to Introl. Amazon and Anthropic's Project Rainier in Indiana is planned for 2.2 GW.

According to S&P Global data, U.S. data centers are projected to consume a total of 183 TWh in 2024, accounting for over 4% of national electricity consumption. By 2030, this number is expected to triple.

These power demands are not a far-off story planned for the future; they are already straining the existing power grid. According to a CBRE report, the vacancy rate of North American data centers dropped from 3.3% in the first half of 2023 to 1.6% in the first half of 2025, the lowest recorded level. According to Cushman & Wakefield data, the vacancy rate slightly rebounds to 3.5% in the second half of 2025, but that is only due to a large amount of new capacity being delivered—absolute levels remain historically low, and meaningful supply relief is unlikely to occur before 2030.

Even AI Says It Is Driving Up Inflation

These investments are pushing demand, raising electricity prices, and dragging down chip supply shortages, but there is an even more covert channel for inflation.

According to a Fortune report on April 1, a team led by Deutsche Bank chief U.S. economist Matthew Luzzetti conducted an experiment: they had Deutsche Bank's proprietary model dbLumina, Anthropic's Claude, and OpenAI's ChatGPT-5.2 each evaluate "the probability of AI driving up inflation in the next year."

The results: dbLumina gave a 40% probability, Claude gave 25%, and ChatGPT-5.2 gave 20%. The three models unanimously assessed the probability of "AI significantly lowering inflation" at just 5%.

The inflationary drivers cited by the three models were highly consistent: large-scale expansion of data centers, soaring semiconductor demand, and rapid growth in power consumption from AI workloads—all of which are demand-pull price pressures.

This stands in stark contrast to the consensus among some Wall Street investors. The Deutsche Bank team wrote in their research report: "Will AI become a major deflationary force? Even AI itself doesn’t think so."

Over a five-year horizon, the models do indeed turn towards greater deflationary possibilities. However, the probability of "AI triggering massive deflation" remains suppressed in the tail risk area.

Optimism Itself Is Inflationary

The St. Louis Fed paper provides a theoretical framework to explain all this.

Faria-e-Castro and Ozkan used standard macroeconomic models to define the AI investment boom as a "news shock". According to the Fed blog post, the model's logic is: when households see AI described as a revolutionary technology, they anticipate future income increases and begin to increase consumption in advance. Businesses anticipate productivity gains and increase investments. The two effects combine, leading demand to quickly exceed supply. The paper states: "These forces together create a wave of inflationary surges in aggregate demand—this is a core characteristic of the early stage of a news shock."

The model provides two paths. If AI does indeed bring about a leap in productivity, short-term inflation will be absorbed by long-term output growth, allowing the economy to enter a virtuous cycle. But if productivity does not materialize—the paper uses the term "persistent low growth and stubbornly high inflation", which is stagflation.

According to the data cited in the Fed blog post, the annualized growth rate of total factor productivity (TFP) in the U.S. since the launch of ChatGPT is 1.11%, lower than the historical average of 1.23%. So far, AI has not left a mark on productivity data.

Meanwhile, according to BLS data, the U.S. CPI year-on-year in February 2026 is 2.4%, and the core CPI is 2.5%, which has not yet returned to the Federal Reserve's 2% target. The Fed's March dot plot shows a median interest rate forecast of 3.4% by the end of the year, pointing to only one interest rate cut this year.

$700 billion is flowing into AI infrastructure. Whether this money is a cause of inflation or a prelude to a productivity revolution depends on a question that no one can answer yet: will the models running in these data centers make the economy more efficient?

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。