AI Trends

European data center expansion meets protests, while China's western computing clusters accelerate deployment

36Kr · AIUpdated Nov 9 17:18

AI summary

Phoenix Net Finance reviewed the progress of protests against data center expansion in multiple European countries: during a conflict at a Microsoft construction site in Amsterdam, more than 150 people were detained; Google cleared over 500 hectares of forest land in Finland without completing an environmental impact assessment, and Finnish police have launched a criminal investigation. The article also compares this with Ulanqab in Inner Mongolia, China, where more than 80 data centers have been deployed and 109 computing power projects have been signed, arguing that the advance planning of the East Data West Computing project and the mechanism isolating residential electricity prices are the reasons for the different pace of deployment in the two places.

Why read this

AI score 60

The article compares the resistance encountered by European data centers with the deployment of computing clusters in western China, presenting the differences between the two approaches in terms of site selection, electricity prices, and water resource constraints.

Topics

Tags

Body · AI translation

On November 6, on the outskirts of Amsterdam, the construction site of a Microsoft data center was in chaos.

Hundreds of protesters broke through the fence and stormed into the construction site; some climbed onto crane jibs and hung banners, and police intervened on site, detaining more than 150 people.

This conflict is not an isolated incident.

Within just a few days, demonstrations were launched simultaneously in multiple European countries, the U.S. midterm elections pushed the pace of data center construction onto the campaign stage, and the capital market also began to re-examine the risks of computing power projects.

Meanwhile, on the grasslands and loess plateaus of western China, computing power clusters are continuing to be implemented at a speed far exceeding that of Europe and the United States.

01 Europe's computing power expansion hit by deforestation scandal

The conflict in Amsterdam opened the prelude to resistance in multiple European countries against the disorderly expansion of data centers.

The leading "Stop Dirty Data Centers" alliance linked up hundreds of social organizations across 15 European countries and organized street rallies in Dublin, Helsinki and other places.

The core demand of the participants is: before a complete regulatory system is established, suspend all new large-scale data center projects.

That this protest could gain momentum in such a short time can be traced back to the deforestation scandal exposed at Google's project in Finland a month earlier.

Pictured is the construction site of Google's AI data center in Muhos, Finland. Image source: AFP

In October this year, on the outskirts of the small town of Muhos in northern Finland, stretches of forest were bulldozed flat by construction machinery.

This messy construction site belongs to Google's AI infrastructure plan with an investment of up to 13 billion euros.

Under Finnish law, such large industrial projects must first complete a full environmental impact assessment before land clearing can begin on site.

But Google's Finnish subsidiary brought in logging equipment and bulldozers before the approval process was completed. In Muhos alone, 330 hectares of forest were cleared, and about 250,000 trees were felled; together with another project plot in Kajaani in the south, the two sites cleared more than 500 hectares of forest in total.

After receiving environmental complaints, the regulatory authorities verified the violations and directly issued a stop-work order in early October, requiring all construction operations that alter the terrain to be suspended.

Pictured is the cleared peat bog plot in Muhos, Finland (the site selected for Google's data center). Image source: The Guardian

By the end of October, Finnish police formally launched a criminal investigation, focusing on whether the damage to the water environment caused by construction drainage violated the Water Act. After the incident was exposed by the media, it quickly escalated, and a citizens' initiative in Finland calling for stronger regulation of data centers gathered 50,000 signatures in just three days.

Google quickly issued a public apology, saying that project execution had not met internal environmental standards, and at the same time proposed a restoration plan to replant 130 hectares of forest locally, but this statement failed to extinguish public anger.

Pictured is the scene of the press conference where Google announced its data center investment project in Finland. Image source: Google

Public dissatisfaction goes beyond a few patches of felled forest, because Google also signed a 22-year nuclear power purchase agreement, directly locking in half of the power generation of the Loviisa Nuclear Power Plant.

Low-carbon electricity in the Nordic region was already regarded as a scarce public resource. After multinational companies enclosed electricity on a large scale, the first thing local residents felt was upward pressure on electricity prices.

According to calculations by Nordea, if Google and other data centers are implemented at the current pace, Finland's wholesale electricity price will rise by about 31% by 2030, and a household using electric heating will pay nearly 400 euros more each year for this; Norway's regulatory agency NVE also warned that without new power sources, electricity prices in 2035 may be 35% to 55% higher than expected.

Moreover, the tangible benefits the project can leave to the local area are very limited. Most of the construction jobs during the construction period are temporary labor, and after formal operation there are only a handful of long-term operation and maintenance positions.

Computing power services, however, target global customers, and profits ultimately flow back to the Silicon Valley headquarters. Finland pays the costs of forests, electricity and the environment, but the industrial value it can get back is pitifully thin.

This dissatisfaction has already spread across Europe, forcing countries to tighten regulation quickly.

Amsterdam in the Netherlands has drawn a red line, suspending acceptance of new large data center applications before 2035; North Holland province limits all projects to designated industrial clusters, with drinking water safety as a hard prerequisite for review. Italy and Portugal have successively raised the threshold for environmental approval, requiring new computing power projects to match a corresponding proportion of renewable energy and to issue a complete long-term impact assessment on water resources.

Pictured is Yassamin Ansari calling for a pause on advanced AI development. Image source: X (formerly Twitter)

Looking at the United States on the other side of the Atlantic, in the just-concluded midterm elections, multiple Democratic candidates running on "restricting the disorderly expansion of data centers" won successfully.

This directly led to the crash of U.S. technology stocks on Friday of this week: the Nasdaq plunged 5.5% in a single day, the Philadelphia Semiconductor Index fell more than 12%, the market value of S&P 500 constituent stocks evaporated by more than 3 trillion U.S. dollars in total, and data center operators and computing power industry chain stocks generally recorded double-digit declines.

Technology giants continue to announce hundreds of billions in computing power capital expenditure, but due to continuously tightening regulation, not many projects can be put into operation and delivered on schedule. The runaway expansion of computing power has thus stalled at the threshold of implementation.

02 Computing power construction in full swing in a small city in Inner Mongolia

Unlike Europe and the United States, on the grasslands of Inner Mongolia in China, the tower cranes building data centers rarely stop.

Pictured is a data center construction site in Ulanqab. Image source: QuantumBit

In the Chahar High-tech Development Zone of Ulanqab, computing power projects are being rolled out across the board, in full swing.

Engineering vehicles shuttle through the park day and night, vehicles with out-of-town license plates can be seen everywhere, outside technical personnel and migrant workers continue to pour in, and rents in surrounding townships have risen accordingly.

The government reception exhibition hall receives more than a dozen visiting delegations on average per day, driving the rapid growth of retail businesses such as restaurants along the streets.

This grassland city, once known for its potato industry, has now landed more than 80 data centers, signed 109 computing power projects, and has an intelligent computing power share of over 90%.

Huawei, Alibaba, ByteDance and other companies have successively put down roots here, and the industry directly calls this place the "Token Capital."

Both are computing infrastructure with high power and high water consumption—so why hasn't China seen the kind of implementation deadlock that Europe and the US have?

The answer lies in the advance planning of top-level design.

In February 2022, the National Development and Reform Commission and three other departments officially launched the "East Data, West Computing" project, laying out 8 computing power hub nodes and 10 data center clusters across the country. The core logic is to guide the East's vigorous computing demand in an orderly way to the West, where land and energy resources are more abundant, and to complete the coordinated layout of the computing power industry from the perspective of national territorial space.

Ulanqab is a typical sample of this logic.

As the core carrying area of the Inner Mongolia hub, all computing power projects are gathered within the scope of the Chahar High-tech Development Zone. The park site is far from the main urban area and residential settlements, with unified designation of construction land, supporting power grids and water supply facilities, without touching the ecological red lines or competing with the people for land.

The city's computing power expansion has clear boundaries and planning, and will not be scattered around communities and natural woodlands as in Europe and the US.

China's electricity pricing mechanism also cuts off the path by which computing power costs are transmitted to residents' electricity costs.

Pictured are a wind farm and a wind-solar energy storage station in Ulanqab, Inner Mongolia. Image source: Xinhua News Agency

First, Ulanqab is rich in wind and photovoltaic resources, and the industrial-to-household electricity price is stable at 0.32-0.36 yuan/kWh, nearly half lower than that of first-tier cities in the East. This is the core driving force for computing power companies to cluster and land here.

More critically, according to China's electricity price management rules, residential electricity consumption follows the government-set catalog price, which has long been rigid and stable.

Data centers fall under the category of large industrial electricity consumption, all of which is conducted through market-based transactions, with costs borne by enterprises and power generators themselves, completely isolated from residential electricity prices. Unlike in European and American markets, the electricity costs of computing power expansion will not be passed on to ordinary households' electricity bills.

The dual advantages of electricity costs and pricing mechanisms support the rapid rise of multiple computing power hubs in the West.

As the core carrying area of the Ningxia hub of "East Data, West Computing," Zhongwei has likewise attracted 26 data center enterprises including Amazon, Meiliyun and CCCC to land here, built 9 data industry parks, and its intelligent computing scale ranks among the top three of the country's top ten clusters.

Pictured are the Yellow River channel and riparian wetlands in Zhongwei, Ningxia. Image source: social media screenshot

But Zhongwei, located in the arid zone of the Northwest and drawing water from the Yellow River, faces water resources as a hard constraint that computing power expansion cannot avoid.

Local industrial water intake is under the dual rigid control of the "1987 Water Allocation Scheme" and the Yellow River Protection Law, and during the 14th Five-Year Plan period the industrial water quota is only 56 million cubic meters.

Traditional data centers rely on evaporative cooling towers for heat dissipation, and evaporation losses, drift and blowdown all continuously consume water resources. The faster the computing scale expands, the more prominent the water supply pressure becomes.

However, these pressures are absorbed within the industry through technology selection and standard constraints.

Pictured is the indirect evaporative cooling air-conditioning unit in the data room building of the Meiliyun Zhongwei Data Center. Image source: Ministry of Industry and Information Technology

For example, Phase I of the Meiliyun Zhongwei Data Center adopts an air-cooling scheme with indirect evaporative cooling units for refrigeration, requiring no traditional water systems such as chillers and condenser towers. For the vast majority of the year it relies on natural cold sources, and only during the hot summer period does it use compressor refrigeration to supplement cooling capacity, significantly reducing evaporation water consumption.

WUE measures the water consumption corresponding to each kilowatt-hour of electricity consumed by a data center. The lower the value, the less water resources are consumed per unit of computing power, and it is the core indicator for judging the water-saving capability of computing power projects.

Pictured is the WUE water efficiency grading standard table for data centers. Image source: National Standards Information Public Service Platform

On December 1, 2025, the national standard "Resource Utilization of Data Centers Part 6: Water Usage Efficiency" was officially implemented. Domestic data center water efficiency is thus divided into four levels: Level 1 is the optimal level, requiring WUE ≤ 0.8 L/kWh; Level 2 is 0.8–1.2 L/kWh; Level 3 is 1.2–2.0 L/kWh; and anything above 2.0 is a level requiring immediate optimization. Water saving has changed from voluntary corporate optimization to a hard yardstick for the entire industry.

Pictured is the CICC Data Zhongwei Zero-Carbon Computing Power Base. Image source: official website of the Zhongwei government of Ningxia

The CICC Data Zhongwei Zero-Carbon Computing Power Base adopts a refrigeration scheme combining liquid cooling and air cooling, controlling water resource utilization efficiency WUE below 0.6, far better than the industry average.

Relying on these iterations of technology routes and upfront access standards, the western computing power hubs have found a feasible implementation path within the hard constraints of water resources.

And the layout of local industries has not stopped at collecting cabinet rents and earning electricity price spreads.

At the end of 2025, Ulanqab officially issued the "Ulanqab Three-Year Action Plan for Building the 'Token Capital' (2026–2028)," clarifying the three-stage leap path of "Grassland Cloud Valley - Intelligent Computing City - Token Capital," attempting to transform computing power advantages into higher value-added industrial value.

According to the plan, the locality will no longer only output basic computing power, but extend downstream to full-chain services such as token production, measurement and evaluation, and security compliance, while cultivating business forms such as data annotation and model fine-tuning, promoting the sedimentation of computing power value locally.

The contiguous stalling encountered by European and American computing power projects is the pure market-based expansion path hitting the boundaries of community interests and resource carrying capacity.

The profit-driven site selection logic of European and American tech giants naturally tends toward suburbs with mature power grids and proximity to users, ultimately forming an irreconcilable mismatch with local livelihood and ecological demands.

The steady progress of China's computing power construction relies on a whole set of institutional advantages, and has walked out an adapted development path within the resource endowments of the West.

As wind turbines on the grasslands and loess plateaus run in sync with servers, and computing power flows eastward along optical cables, western China is growing from a traditional energy exporter into a new hinterland of the digital economy.

This article is from the WeChat public account "Phoenix Finance," author: Phoenix AI Research Institute, published by 36Kr with authorization.

Source

36Kr · AI