Upstream NVIDIA and Taiwan Semiconductor are making a fortune, while downstream faces high debt and low profits – the current state of the AI industry chain

The wave of artificial intelligence (AI) is creating a distinct industrial landscape. Chip and server manufacturers upstream in the supply chain are enjoying substantial profits, while many downstream cloud service providers and application developers are mired in high costs and low profits, with rising debt levels ringing alarm bells for the industry. The latest example comes from software giant Oracle. According to The Information, the company's rapidly growing AI cloud rental business is eroding its substantial profits. Financial reports show that in the three months ending in August, Oracle's gross margin for renting NVIDIA chip servers was only 14%, far below the company's overall gross margin of about 70%. This news caused Oracle's stock price to plunge by 5% at one point, dragging down the overall market. Oracle's predicament is not an isolated case. Due to the high prices of NVIDIA chips, all cloud service companies providing AI computing power rentals are facing profit pressures. Meanwhile, Goldman Sachs and JP Morgan both issued warnings this week, pointing out that the debt levels of tech companies funding the AI boom have crossed important benchmarks, and they are increasingly turning to the credit markets to support high development costs. This series of dynamics clearly indicates that, although the AI story remains captivating, the path to commercialization and profitability is longer and more arduous than expected for most participants. For investors and companies, the current core question is: Are the returns from AI services truly worth their high costs? The Feast of Hardware Giants In the current AI race, the biggest winners are those companies providing the infrastructure for AI. From chip manufacturer NVIDIA and chip foundry Taiwan Semiconductor to server supplier Dell, these upstream companies are converting massive market demand into tangible profits. Dell's performance is a testament to this. The company has doubled its annual revenue growth forecast to between 7% and 9%, while also doubling its profit growth forecast to at least 15%. Dell clearly states that this is due to strong demand for servers equipped with AI chips. These hardware giants hold the "ticket to entry" in the AI industry and are capturing the vast majority of value at this stage. The "Increased Revenue but No Increased Profit" Dilemma for Cloud Vendors In stark contrast to the upstream companies reaping rewards, cloud service providers in the midstream of the supply chain are facing the awkward situation of "increased revenue but no increased profit." Although they have achieved rapid revenue growth by providing AI computing power, their profitability is severely squeezed. Oracle's case is highly representative. Although the company expects its AI cloud rental revenue to grow significantly by 2030 and potentially become a major source of income, this is likely at the expense of the company's overall profit margin. The root of the problem lies in the high costs of NVIDIA chips, which leave cloud vendors with thin margins in their computing power rental business. For the cloud service industry, which is accustomed to high profit margins, the profitability brought by AI business has fallen far short of expectations. High Costs and Profitability Challenges For downstream AI model developers and application vendors, the profit outlook is equally uncertain. There is currently almost no evidence that selling AI applications to businesses or individuals is a profitable business The development and operational costs of AI models are a heavy burden borne by all AI applications. For example, applications like programming assistants do not have high profit margins. Although these costs have decreased over time, the challenges of commercialization remain immense. Many large publicly listed companies are tight-lipped about the specific revenues from their AI applications, let alone profits. Furthermore, enterprise customers are cautious about whether to pay high prices for AI services. Participants in industry conferences predict that widespread adoption of AI technology by enterprises may still take several years. Debt Alarm Sounds While the profit outlook remains unclear, financial risks in the AI sector are accumulating. Top Wall Street investment banks have already noticed this trend and begun to issue warnings. Goldman Sachs and JP Morgan both pointed out this week that the debt levels of tech companies diving into the AI wave are soaring sharply. To cover the high computing power costs required for AI development, these companies are increasingly turning to the credit markets. This phenomenon marks a warning sign, indicating that during the massive investment cycle in AI, the financial leverage of some companies is being rapidly amplified, adding new uncertainties to the future market