AI is shifting from infrastructure investment toward real-world deployment, and whether capital expenditure can be sustained and how the supply chain should respond to shortages and price hikes have become shared concerns across the industry.
Recently, the 2026 Avnet Intelligent Technology Ecosystem Forum was held in Shenzhen, where Cai Zongfu, Vice President of Supplier and Product Management for Avnet China, and Tong Xinghua, Senior Sales Director for Avnet China South Region, spoke with Sina Tech.
Cai noted that Avnet continuously tracks the capital expenditure of major cloud service providers in China and the US, with a focus on its proportion of free cash flow, which remains healthy and investment is still accelerating. He believes that whoever has stronger deployment capabilities will reap returns sooner, and once a positive cycle forms, investment can be sustained.
Tong stated that the AI industry is still in the infrastructure construction phase, and capital investment has not yet fully translated into returns, which is a long-term process. From a demand perspective, customers' focus has shifted from finding chips and computing power to finding solutions, which also brings new business growth opportunities for Avnet.
The Second Half of AI Moves Toward Real-World Deployment
According to Avnet's latest survey, 56% of engineers globally work at companies that have begun delivering products with integrated AI technology to customers, a 33% increase from last year. Among them, China's AI application penetration rate is significantly higher than the global average.
When discussing the AI industry's boom cycle, Cai said that Avnet continuously tracks the capital expenditure of major cloud service providers in China and the US, but the company looks not only at the amount, but also at capex as a proportion of these providers' free cash flow for the quarter. Over the past two years, capex has continued to rise, but its proportion of free cash flow remains healthy, and major cloud service providers in both countries are still increasing investment with no sign of stopping, instead accelerating.
Cai also noted that companies invest capital to drive industry development and generate returns, and whoever has stronger deployment capabilities will reap returns sooner. Once a positive cycle forms, investment can continue; if returns fall short of expectations, cyclical changes may occur.
Regarding the sustainability of the industry's high growth rate, Tong said that capital expenditure is cyclical, and at this stage infrastructure is still being built, which is a long-term process similar to how the automotive and mobile phone industries took years to gradually standardize. Capital expenditure has not yet translated into returns, and the first half of AI is still underway. As AI capabilities are deployed in some application scenarios, the second half of AI has also begun.
Cai pointed out that the first half of AI mainly involved domestic and international cloud service providers as well as US tech giants and China's BAT and ByteDance investing in AI data centers, with applications including AI servers and switches. The second half is about deploying large model capabilities across industries and endpoints, such as edge AI, robots, embodied intelligence, factory automation, and smart manufacturing.
Cai added that from Avnet's perspective, it looks not only at the investment cycle but also at the return cycle. How to generate returns after massive capital input is largely answered in the second half of AI.
Only when scenario deployment achieves significant progress can the computing power invested earlier be fully utilized and further converted into quantifiable returns. Against this backdrop, Avnet's AI business has also seen significant expansion, becoming the company's second growth curve, with most of the growth coming from domestic customers, involving areas such as embodied robots, smart manufacturing, and physical AI. Currently, Avnet has established a dedicated AI business unit covering technology, solutions, sales, and technical support.
From the perspective of customer demand, Tong said that in the past many customers looked for chips and computing power, but now more customers are looking for solutions, focusing on whether the solution can meet product requirements, which has also created new business growth points for Avnet.
Domestic Chip Going Global Requires Long-Term Capabilities
Speaking about component shortages, Tong pointed out that when demand surges in a concentrated manner, shortages occur. Currently, there is high demand for AI computing power, memory, and power supplies, and supply is noticeably tight.
She emphasized that supply chain management can help customers reduce the risk of material shortages, and with proper planning, it can also reduce the risk of supply disruptions.
Cai added that Avnet typically does not rely on forecasts to help customers cope with product shortages and price fluctuations, because the semiconductor industry is highly globalized, and supply-demand changes in any region or market can affect the pricing system, making forecasting difficult.
He pointed out that Avnet relies more on supply chain management, tracking the lead time of each model at the system level and connecting with a large number of suppliers through EDI (Electronic Data Interchange). Once lead times lengthen or shortages appear, the system will promptly capture and notify customers, quickly schedule orders according to service agreements, and help lock in products. This is not about relying on forecasts, but on data and tools to reduce customer risk.
Regarding cooperation with domestic chips and going global, Tong said that domestic chip manufacturers have continued to increase their market share in recent years. To further expand share, especially to break through in the high-end market, customers value not only the performance of a single product but also a complete product roadmap, that is, whether there are upgraded products, whether annual planning is possible, and whether chips can be supplied sustainably.
For domestic chips to successfully go global, they cannot rely only on short-term tactics to win orders, but must integrate with local industry ecosystems and continuously create value for local customers.
As for whether domestic chip going global is driven by replacement demand, Cai believes that most customers do not prioritize geopolitics when choosing products, but place more emphasis on product performance and long-term supply capability. Only when these conditions are met can a product succeed. The semiconductor industry is highly globalized, and the supply chain shortages brought by this round of AI represent an important development opportunity for domestic semiconductor manufacturers.
It is understood that Avnet China has continuously expanded its domestic chip distribution business in recent years, and the annual growth rate of this business's revenue continues to accelerate. In terms of specific cooperation, Avnet has become the top-performing distributor for SG Micro Corp in China.
Comments