Text / Yang Jianyong

If you ask who makes the most money in the era of AI large models, the answer is undoubtedly chip and computing power server manufacturers. In the generative AI era, training and iterating models creates strong demand for computing power. In particular, Agent applications are accelerating their penetration into various industries, further increasing Token usage and continuously expanding the overall computing power market—allowing chip and computing power server manufacturers to reap enormous profits.

Harvesting Computing Power Dividends: Both Revenue and Profit Hit New Highs

Foxconn Industrial Internet is a typical manufacturer deeply benefiting from the computing power market, with revenue scale and profit quality continuing to grow rapidly. In the first three quarters of 2026, revenue reached 941.414 billion yuan, up 55.88% year on year; net profit was 40.4 billion yuan, up 79.55% year on year. Among this, GPU and ASIC computing power were the core engines driving overall revenue growth.

Specifically, AI-ASIC server shipments increased 1.5 times year on year, and AI-GPU rack shipments rose 70% year on year; at the same time, a new generation of AI racks has begun shipping. These figures demonstrate core competitiveness in the AI computing infrastructure field.

At the same time, demand for large model training and inference is growing rapidly. Global data center network architectures are accelerating iterative upgrades toward ultra-large scale, high bandwidth, and low latency. The high-speed switch market continues to grow, and the company's market share remains firmly among the industry leaders. In the third quarter, shipments of 800G and above switches increased 80% year on year.

It is worth noting that global cloud giants continue to increase investment in AI infrastructure, and demand for high-end computing power equipment continues to rise—especially unprecedented demand for computing power services such as AI GPU racks and AI servers. Computing power services have thus become Foxconn Industrial Internet's largest business segment, transforming this traditional contract manufacturer into a leading computing power enterprise.

Revenue Nearing One Trillion, Market Value Down 770 Billion

Demand for AI computing power continues to surge, bringing a positive impact on the performance of companies in the AI industry chain. As a supplier of core hardware such as AI servers and high-speed switches, Foxconn Industrial Internet has become a hot AI computing power company, with its market value once reaching 1.67 trillion yuan in June this year. Unfortunately, it has now fallen below the one trillion mark, with a current market value of 892.9 billion yuan.

In just five months, Foxconn Industrial Internet's market value has wiped out 777.1 billion yuan, a decline of 46.5%. Market fluctuations are affected by multiple factors—not only the overall market, but also profit-taking after earlier gains. More fundamentally, the company only earns slim contract manufacturing profits in the AI industry chain.

In the AI large model wave, Foxconn Industrial Internet has a huge revenue volume, but its profit side forms a sharp contrast. Net profit in the first three quarters exceeded 40 billion yuan, and profitability continues to improve, but the overall net profit margin is only 4.29%, highlighting the profit ceiling of the contract manufacturing model.

By contrast, upstream manufacturers that master core technologies occupy the bulk of profits. Take Nvidia as an example: its net profit margin exceeds 60%. While harvesting the dividends of global AI large models, it has leapfrogged into one of the most profitable technology companies in the world.

Even though Foxconn Industrial Internet's revenue has reached an astonishing 940 billion yuan, it is still difficult to conceal the reality of a low profit margin—like an insurmountable gap. Its essence is the result of its business model: it can only earn hard-earned contract manufacturing fees and achieve scale effects through small profits but high sales volume.

Final Thoughts

Since entering the large model era, major global cloud service providers have continued to increase AI spending, demonstrating the long-term growth potential of AI infrastructure. Demand for core hardware such as AI servers and high-speed switches is increasing day by day, and overall shipments continue to show strong demand. For computing power manufacturers, they are expected to benefit from the growth in computing power demand, which will in turn have a positive impact on performance.

Yang Jianyong is a Forbes China contributor. The views expressed are personal. He focuses on in-depth observation of technology fields such as AI large models, artificial intelligence, industrial robots, IoT, cloud computing, and smart hardware.