Jie Zhou on data centers and digital policy

Growing up in a rapidly modernizing China, Zhou was preoccupied by a particular question: why do some countries grow rich, while others don’t? Knowing how the governments of the region’s economic “Tigers” (Hong Kong, Singapore, Taiwan, and South Korea) found prosperity through good industrial policy, Zhou gravitated toward understanding how such policies can be a lever for developing economies.

“Most of the literature is about manufacturing, but what about industrial policy in the digital economy, given that Google is blocked in China, yet powerful corresponding apps serve the whole market?”

- Jie Zhou

This brought her to study economics at MIT and intern at a tech firm in China, which deepened her curiosity about the economics of data and infrastructure in the digital age. Her experiences led her to a more specific intuition: in the digital age, something is fundamentally different about how the economy organizes and expands. Data centers are both everywhere and somewhere specific, impacted by a mix of energy costs, regulatory forces, and geopolitical pressures that existing industrial policy frameworks have been slow to map. Now at Yale, Zhou has developed that intuition into a research project: how does industrial policy adapt when the drivers of the economy are no longer physical, but digital?

To understand Zhou’s research, it helps to start with what makes data economically unusual. Unlike physical goods, using data doesn't use it up. In fact, more data often makes data more valuable. That characteristic gives the digital economy powerful economies of scale and makes it fundamentally different from the manufacturing sectors that dominate most industrial policy thinking. “When people think about data,” Zhou notes, “this kind of characteristic really makes this industry different.”

When drivers of the economy shift from being physically manufactured goods to digital ones, it changes the assumptions that underlie those economic systems. Debates over digital security, sovereignty, and environmental costs have grown as data centers are established across the globe. These are the issues that brought Zhou to study trade policies and the digital economy, particularly with the growth of data infrastructure and regulation.

To understand where firms locate centers and why, Zhou and her collaborators constructed a large-scale dataset from the ground up, pulling from three key streams of information. On the supply side, they scraped pricing information directly from data center providers. On the demand side, they used IP address data to map user locations and estimate latency – the transmission time between a user and the server storing their data. They also compiled policy data on data localization laws across countries to measure regulatory environments.

The result of their work is a framework centered on three key factors: data centers, the firms that use them, and the end users those firms serve. Firms weigh costs – energy, water, infrastructure – against latency requirements and, increasingly, regulatory constraints, as they decide where to locate their operations. “Recently, what we see as an increasing concern for firms is actually regulation," Zhou says. "Localization matters.”

Some theoretical approaches, including equilibrium models, have attempted to simulate optimal location choices by formalizing all of these trade-offs simultaneously, but as Zhou notes, they have proven difficult to implement in practice. This is part of the complexity of studying an industry that is still rapidly evolving.

Several patterns have emerged from Zhou’s initial analysis, some expected, some surprising. 

First, she notes that firms are increasingly choosing not to build their own data centers at all, relying instead on third-party providers. This suggests that many firms are currently treating data storage as a service rather than a core competency. This may have significant implications, as having only a handful of major suppliers raises questions about dependency and the stakes of data localization policy. 

Second, and perhaps most striking, real-world patterns in construction also reflect geopolitical forces, not merely economic considerations. Singapore, for instance, has emerged as a neutral ground for firms caught between US-China relations. On paper, a small, land-constrained city-state with few natural resources may not seem like the obvious location for a hub.

“What we see in the data is that Chinese and US firms, because of the tension between the two countries, are using Singapore as the middle ground.”

- Jie Zhou

Other regions, such as Northern Europe, have become attractive to data center builders, offering ample land and favorable energy considerations. Even Switzerland is now attempting to market itself as a politically neutral home for data, expanding its reputation for discretion as a selling point for data storage. 

Third, policy is becoming an increasingly important driver of firm behavior. As Zhou notes, the global boom in AI has prompted countries to prioritize constructing their own data infrastructure. This boom has also prompted a threefold increase in data localization policies in recent years, and as they proliferate, firms are making location decisions that may no longer be driven by cost or efficiency alone. 

These findings point to a fundamental tension at the heart of the data economy. On the one hand, data exhibits strong economies of scale. The unlimited replicability of data creates powerful incentives for sharing and consolidation, offering gains beyond the kind the manufacturing economy could produce, and that could, in principle, benefit all. On the other hand, countries are increasingly treating data as a national asset, erecting barriers that may be economically inefficient and environmentally costly. If every country builds its own isolated data infrastructure to store its own copies of identical data, it risks massive electricity and environmental resources waste. “How should we build data alliances?” Zhou asks. “And will it be beneficial for social welfare – really saving costs in terms of production, but also beyond that?”

The question for policymakers, then, is whether data should be treated as a global resource or a national asset – and whether new forms of international coordination, “data alliances,” as Zhou calls them, might offer a new path forward.

Zhou is looking to take her research further in two directions. The first involves building an even richer dataset to better understand how data accumulates over time. The second is a deeper theoretical question: what are the consequences of having too much data? In particular, she is interested in whether large datasets and the algorithms built on them might introduce new forms of market inefficiency, with implications well beyond data centers. 

In many ways, the challenge Zhou’s research confronts is not the problem of scarcity that has preoccupied other economists. “The challenge is not scarcity, but abundance,” she suggests. As the digital economy continues to expand, understanding where data lives, who controls it, and how it is used will only become more urgent.