DeepSeek, the Chinese AI startup famous for building efficient models that cost far less to train than competitors, is now making a dramatic infrastructure investment that signals a fundamental shift in its strategy. The company plans to construct a gigawatt-scale data center in Ulanqab, Inner Mongolia, with at least 1 gigawatt (GW) of computing capacity, according to reporting from Bloomberg. Part of the facility could become operational by late 2027 or early 2028, backed by a $7.4 billion funding round from investors including Tencent and CATL.

What Does a 1 GW Data Center Actually Mean?

One gigawatt refers to the maximum electrical capacity available to power computing equipment, cooling systems, networking hardware, and supporting infrastructure. A fully developed 1 GW AI campus would belong to a new generation of facilities designed around increasingly dense clusters of advanced processors. However, not all of that power flows directly into AI chips. Cooling, networking, power conversion, and backup systems will consume part of the available electricity.

To put this in perspective, a gigawatt-scale facility represents a massive commitment to physical computing infrastructure. DeepSeek's plan shows that even companies built on the promise of efficiency cannot escape the fundamental need for enormous computing resources to train larger models, serve more users, and reduce dependence on external computing providers.

Why Is DeepSeek Building in Inner Mongolia?

The location choice reveals strategic thinking about cost and logistics. Ulanqab, approximately 350 kilometers northwest of Beijing, has become one of China's most important computing and data center locations. The city's cold climate allows operators to use outside air for cooling during much of the year, reducing electricity consumption by conventional cooling equipment. Huawei's Ulanqab data center, for example, achieves annual power usage effectiveness of 1.32, with monthly efficiency reaching approximately 1.22 under full natural cooling conditions.

The region also offers substantial wind and solar resources. In July 2025, China connected a green-powered computing base in Ulanqab to the grid, combining 200 megawatts of wind power, 100 megawatts of solar capacity, and 45 megawatts of energy storage. The facility was designed to generate approximately 848 million kilowatt-hours of green electricity annually and serve a computing base containing 25,000 server cabinets.

Ulanqab remains close enough to Beijing to serve companies operating in the Beijing, Tianjin, and Hebei economic region. Apple, Alibaba, and Huawei have already developed data center infrastructure in the area, creating an established computing cluster rather than an isolated facility in a new market.

How Is DeepSeek Structuring the Project?

DeepSeek is using a hybrid development model that combines owned and leased capacity. The company reportedly plans to build part of the facility itself while leasing additional capacity from existing data center operators. This approach offers strategic advantages on multiple fronts:

• Control and Security: Building its own facility gives DeepSeek greater control over hardware, networking, security, and model development operations.
• Speed to Capacity: Leasing capacity from existing operators allows the company to access computing resources before the entire campus is complete, avoiding delays.
• Flexibility: The hybrid model reduces risk by spreading infrastructure investment across owned and leased assets rather than betting entirely on a single facility.

This strategy matters because chip availability is no longer the only constraint on AI expansion. Grid connections, transmission lines, permits, and construction schedules can delay projects even when companies have sufficient financing.

What Hardware Will Power the Data Center?

The processors that will power the project have not been confirmed. DeepSeek could use Nvidia hardware, Huawei processors, its own future chips, or a combination of different systems. This question carries significant political weight.

A senior US official told Reuters in February 2026 that DeepSeek had trained an advanced model using Nvidia Blackwell processors despite US export restrictions. The official claimed that the Blackwell chips were likely clustered at a data center in Inner Mongolia. DeepSeek has not publicly confirmed how the chips were obtained or whether the allegation is accurate.

The case intensified concerns that advanced Nvidia chips could reach Chinese AI companies through intermediaries, third countries, or data center arrangements outside established export channels. The final processor mix for DeepSeek's Ulanqab facility remains unclear, though Nvidia chips continue to set the standard for many frontier AI workloads, while Huawei has become China's leading domestic alternative.

DeepSeek is also taking steps to diversify its supply chain. Reuters reported in July 2026 that the company is developing its own inference chip, designed to run trained models and generate responses for users. This represents a different type of processor than the more powerful chips required to train a frontier model from scratch.

What Does This Shift Reveal About AI Economics?

DeepSeek's infrastructure pivot represents a major strategic shift for a company that became famous for reducing the computing costs associated with frontier AI development. The company's R1 model, for example, gained attention for achieving strong performance with lower training costs than competitors. Yet the new data center plan shows that efficiency gains do not eliminate the need for massive infrastructure investment.

This reflects a broader industry trend. As computing becomes more efficient, total demand for infrastructure actually increases because more users can access the technology, more companies can afford to build AI systems, and more applications become economically viable. The result is paradoxical: efficiency drives growth, which drives infrastructure spending.

DeepSeek's $7.4 billion funding round and gigawatt-scale data center plan demonstrate that the company is preparing for a future where it controls more of its own computing destiny. By building owned infrastructure alongside leased capacity, DeepSeek can reduce its dependence on external providers, improve security and control, and scale more rapidly as demand for its models grows.

" }