The article tackles the significant disparity between massive capital expenditures (capex) in AI and the relatively meager revenues generated by AI applications. The US economy is heavily investing in AI, with data center capex contributing 92% of GDP growth in early 2025. OpenAI plans to spend $1.4 trillion on infrastructure, raising concerns about whether this investment can deliver sufficient returns. While predictions indicate that OpenAI and Anthropic will generate $20 billion and $9 billion in annual revenues, respectively, these figures fall far short of justifying the trillions being spent. Since the launch of ChatGPT in late 2022, AI's allure has driven capex upward, but many AI applications remain in experimental phases, leading to inflated revenue figures. The quality of these revenues is questionable, especially as companies face challenges in delivering value from initial deployments. The article emphasizes that a substantial portion of AI startup revenues relies on venture capital, which is dwindling. As a result, many startups might cut back on cloud spending, impacting the broader ecosystem. The analysis also highlights the risk of an AI bubble, pointing to the growing debts of major tech companies due to AI investments. Public market trends show a re-evaluation of companies providing foundational AI services. While some SaaS companies are seeing growth thanks to AI, their overall revenue from AI remains a small fraction of total earnings. The piece concludes by estimating that AI revenues need to substantially increase to fill a projected $1.5 trillion gap relative to capex by 2028, suggesting a precarious situation for the industry if rapid growth doesn’t materialize.