The 2026 Sustainable Global Leaders Conference and the Second Green Industry and Sustainable Consumption Expo 2026 Sustainable Global Leaders Conference was held from October 20 to 22 at the Huangpu Expo Park in Shanghai.
Michele Wucker, author of The Gray Rhino: How to Recognize and Act on the Obvious Dangers We Ignore and an expert in global risk and economic policy, attended and spoke.
At the forum, moderator Huang Yanming, Director of the Orient Securities Research Institute, raised a core investment question. He pointed out that there is significant uncertainty in the construction of computing power centers, that asset depreciation and residual value assessment are difficult, and that the actual usable life of computing power assets is hard to predict. From the perspective of financial sustainability, he asked how overseas markets assess the return on investment, depreciation residual value, and stranded asset risk of computing power assets; what financing and contractual mechanisms can share risk and prevent excessive computing power construction; and what differences exist between overseas models and domestic intelligent computing center construction.
Michele Wucker said that if the perfect answer to the valuation of computing power assets had already been grasped, all investors would reap handsome returns. At present, there is a large amount of duplicated investment in the global AI industry, and large AI companies are competing for data resources. From an investor's perspective, the efficiency of computing power deployment is more critical, and companies should be more prudent and more selective in data screening and analysis.
She noted that the U.S. market has already begun discussing the challenge of utilizing idle computing power. The operating logic of the power grid can be borrowed to build a computing power scheduling network. The power grid uses price incentives to encourage users to shift electricity use to off-peak hours; AI computing infrastructure can likewise adopt a similar approach to achieve stable all-weather supply of computing power. At the same time, electricity is the core constraint on computing power expansion, and large amounts of capital continue to pour into the power and data center tracks.
In her view, the pace of large-scale data center construction is expected to slow in the future, and companies should be more prudent in the data introduction stage. Many companies are currently advancing sovereign AI construction and choosing localized model deployment to achieve independent control of data and ensure data reliability. But hyperscale data centers also carry hidden debt risks, and project liabilities will be transmitted to upstream and downstream players such as chip manufacturers, engineering construction parties, and contractors. Once AI demand declines and data center revenue contracts, it will deal a blow to the entire industrial chain, and some suppliers have already shown default risk, making the design of financing mechanisms especially critical.
Michele Wucker mentioned that many computing power projects currently rely on private credit fundraising, and the transparency of funds is relatively low. Improving financing transparency is an important lever for defusing risk. She used an overseas saying to describe the current investment model of the computing power industry: "throw spaghetti at the wall and see which strand sticks," meaning first carry out large-scale construction on a broad front, then wait for the market to verify the feasibility of the projects.
She cautioned that computing infrastructure investment must not expand blindly, that a risk-sharing mechanism needs to be established, and that the chain financial shocks brought by stranded assets must be guarded against. Through means such as computing power scheduling networks and optimizing financing structures, the supply of computing power should be balanced against real market demand.
Comments