Big Tech is paying for the AI boom, and chipmakers are cashing in: Chart of the Day

Big Tech hyperscalers are spending heavily on AI infrastructure, causing their free cash flow to turn negative, while chipmakers like Nvidia and Micron are cashing in with rising free cash flow, according to Bank of America. The risk is that AI revenue may take longer to materialize than the spending, with falling token prices and Chinese models adding pressure.

Wall Street knows Big Tech can spend on artificial intelligence. The harder question is how quickly that spending returns cash to those companies. That is where the story changes. The AI boom is not just a growth story anymore; it is a money-in, money-out story. Hyperscalers — Amazon AMZN https://finance.yahoo.com/quote/AMZN/ , Alphabet GOOG https://finance.yahoo.com/quote/GOOG/ , GOOGL https://finance.yahoo.com/quote/GOOGL/ , Meta META https://finance.yahoo.com/quote/META/ , Microsoft MSFT https://finance.yahoo.com/quote/MSFT/ , and Oracle ORCL https://finance.yahoo.com/quote/ORCL/ — are writing the checks for chips, data centers, and power. Chipmakers are getting paid first https://finance.yahoo.com/markets/article/the-ai-boom-now-has-a-price-tag--and-micron-just-sent-the-bill-chart-of-the-day-100000526.html , and they are expected to keep more cash after their own bills are paid. Bank of America Global Research calls it a "generational transfer in free cash flow," and the chart below shows why. Free cash flow https://www.investopedia.com/terms/f/freecashflow.asp is the money a company has left after running the business and paying for major investments. In BofA's chart, that number is moving in opposite directions for Big Tech and chipmakers. For the hyperscaler basket, it is falling into negative territory, while the semiconductor basket — Nvidia NVDA https://finance.yahoo.com/quote/NVDA/ , Micron MU https://finance.yahoo.com/quote/MU/ , Broadcom AVGO https://finance.yahoo.com/quote/AVGO/ , and Applied Materials AMAT https://finance.yahoo.com/quote/AMAT/ — keeps climbing. That does not make the AI boom any less real. It makes the payoff harder to pin down. Chip demand can remain strong while hyperscaler cash flow weakens because one side is selling the infrastructure and the other side is paying to build it. BofA said "Magnificent Seven" hyperscalers have spent $234 billion in capital expenditures this year, while their stocks are basically flat in 2026. Apollo chief economist Torsten Sløk frames the question https://finance.yahoo.com/markets/article/apollos-sl%C3%B8k-the-market-faces-big-risks-if-hyperscalers-ai-profits-get-delayed-161708773.html this way: "But what if the payoff takes longer than consensus assumes?" Disclosure: Yahoo is a portfolio company of funds managed by affiliates of Apollo Global Management. His answer starts with two pressure points. Token prices are still falling, which means AI usage can grow without producing as much revenue per unit of use. Chinese models are also gaining on the US's models, adding pressure on American platforms trying to turn AI adoption into high-margin revenue. The risk is timing. The spending happens now. The cost of chips, servers, and data centers keeps showing up over time. But the revenue and cash flow may take longer to arrive. Sløk's chart of Chinese vs. US model token usage sharpens the point. Among the top 20 AI models, Chinese models handled more tokens than US models in both May and June, and the gap widened. Chinese model usage rose from 46 trillion tokens in May to 98 trillion in June, while US model usage rose from 37 trillion to 53 trillion. There is still a clear upside case if AI usage keeps growing, customers pay for better tools, and Big Tech turns today's spending into higher revenue over time. In that scenario, the current cash-flow hit is the cost of building the next platform, not a warning sign.