Zhipu AI Completes 1GW Domestic-Chip Data Center, Shares Surge Nearly 30%

Zhipu AI's 1GW data center is a large-scale server facility powered entirely by domestically produced Chinese chips for training and iterating its GLM models.

The development of this data center underscores the strategic importance of domestic hardware in sustaining AI innovation within China. With international chip supplies facing ongoing challenges, local alternatives have gained prominence in enterprise planning. Zhipu AI's approach allows for uninterrupted scaling of computational resources dedicated to advanced model training. The integration of over 10,000 semiconductors highlights the manufacturing capabilities that have been cultivated domestically. This setup not only supports current GLM model iterations but also prepares for future expansions in AI applications across various industries. Enterprise users can expect more stable access to high-performance computing as a result of such investments.

In addition to the immediate benefits, the data center project signals a maturing phase in China's AI industry where infrastructure keeps pace with algorithmic advances. The ability to train and iterate models locally enhances data privacy and customization options for enterprise clients. This is particularly relevant for sectors handling sensitive information. The project's success could encourage more funding into domestic chip development and related technologies. Overall, it contributes to a more resilient AI supply chain within the country. The partial operations provide early indicators of the facility's effectiveness in supporting demanding workloads.

What background led to Zhipu AI pursuing a fully domestic data center?

Prior to this project, Chinese AI firms including Zhipu AI had relied heavily on imported chips for their training infrastructure. The shift to domestic solutions reflects a response to global market dynamics and policy directives aimed at self-sufficiency. The Chinese Academy of Sciences has played a role through its research arms, contributing to the ecosystem that enables such facilities. Knowledge Atlas Technology, the entity behind Zhipu AI and listed as 2513.HK, has positioned itself at the forefront of this transition. The acquisition of XCore Sigma further strengthens this position by bringing in specialized software expertise from the Institute of Computing Technology. This combination of hardware and software optimization is key to maximizing the performance of local chips in demanding AI workloads.

The timeline of events shows a coordinated effort to fill capability gaps in the domestic AI stack. Bloomberg has reported on the completion based on sources close to the company. This background sets the stage for understanding why the stock market reacted positively to the news. Investors see the data center as a validation of Zhipu AI's long-term strategy in the enterprise AI space. The partial operations indicate that the facility is already contributing to productive outcomes rather than remaining in a purely developmental phase. Such rapid progress is notable in the context of large-scale infrastructure projects that often face delays.

The role of the Chinese Academy of Sciences in nurturing companies like XCore Sigma cannot be overstated in this context. Their research has translated into practical applications that support the data center's operations. This collaboration between academic institutions and commercial entities exemplifies the integrated approach China is taking to AI development. Enterprise AI users gain from this synergy through more advanced and accessible tools. The strategy also mitigates the impact of any potential export controls on advanced technologies. As a result, Zhipu AI is better equipped to compete on the global stage with its GLM models.

What are the technical specifics of the 1GW data center?

The data center features a designed power capacity of 1GW, which is substantial even by global standards for AI training clusters. It is equipped entirely with Chinese-made AI chips, avoiding any dependence on foreign suppliers for the core hardware. More than 10,000 domestic semiconductors are deployed across the servers, enabling parallel processing for large language model training. The facility has entered partial operations, allowing for immediate use in GLM model development and refinement. This scale makes it the largest among those built by Chinese AI companies, according to available reports. The design prioritizes efficiency and compatibility with local chip architectures to achieve optimal performance.

Technical integration involves advanced cooling and power management systems tailored to the characteristics of domestic semiconductors. The partial operation phase allows engineers to test and fine-tune the setup before full-scale deployment. This methodical approach helps in identifying and resolving any issues specific to the domestic hardware ecosystem. The focus on 1GW capacity reflects ambitious plans for handling the compute-intensive tasks associated with frontier AI models. Enterprise customers stand to benefit from the increased capacity as Zhipu AI can offer more robust AI services. The technical achievement is a testament to the maturation of China's semiconductor industry in the AI domain.

Power management at this scale requires sophisticated engineering to maintain efficiency and reliability. The domestic chips have been optimized for the specific demands of AI training, including high throughput for matrix operations. This technical alignment reduces the need for extensive modifications during integration. The result is a more streamlined deployment process that accelerates the time to operational status. For enterprise applications, this means faster access to updated models and improved service levels. The 1GW capacity provides headroom for growth in computational needs over the coming years.

How does the acquisition of XCore Sigma contribute to the overall strategy?

The acquisition of XCore Sigma, also known as Zhongke Jiahe, brings in critical AI infrastructure software capabilities. Originating from the Compiler Laboratory of the Institute of Computing Technology at the Chinese Academy of Sciences, the company specializes in tools that enhance chip utilization. This move directly addresses the need to optimize the performance of domestic semiconductors within the new data center. By integrating this software expertise, Zhipu AI can achieve higher efficiency in its training processes for the GLM models. The acquisition fills a gap in the software layer that complements the hardware investment. Reports indicate that this dual approach of hardware and software is essential for competing effectively in the enterprise AI market.

XCore Sigma's technology is expected to improve resource allocation and reduce waste in the data center operations. This is particularly important given the high power demands of AI training. The origins in the Chinese Academy of Sciences ensure that the software is developed with deep understanding of local hardware constraints. The timing of the acquisition alongside the data center completion suggests a deliberate strategy to create a complete domestic stack. Enterprise AI deployments can leverage these optimizations for more cost-effective and reliable performance. The combined entity is better positioned to support a wide range of applications from natural language processing to other AI tasks.

What was the stock market reaction to the data center news?

On July 21, 2026, shares of Zhipu AI experienced a significant increase, surging nearly 30% intraday. The morning session saw gains of 25.32%, reflecting strong investor confidence in the company's direction. This reaction was triggered by reports of the data center completion and the related acquisition. The surge highlights the market's recognition of the strategic value in achieving compute independence. Bloomberg's coverage contributed to the visibility of the achievement among global investors. Such movements in stock price can influence broader sentiment toward Chinese AI stocks in the enterprise sector.

The intraday high reached during the session underscores the positive perception of the milestone. Investors are likely factoring in the long-term competitive advantages this infrastructure provides. The acquisition aspect adds another layer of value by enhancing the operational capabilities. This market response is consistent with how other tech milestones have been received in the past. The nearly 30% gain serves as a tangible indicator of the importance attached to domestic AI advancements. Further monitoring of the stock will reveal if the gains are sustained as more details emerge.

What implications does this have for enterprise AI stakeholders?

For enterprise customers in China and beyond, the data center opens up new possibilities for accessing advanced AI capabilities without reliance on external supply chains. The increased capacity supports more extensive training and fine-tuning of models tailored to specific business needs. Zhipu AI's GLM models stand to benefit from the enhanced compute resources, potentially leading to improved performance in applications such as data analysis and automation. The domestic focus may also appeal to organizations prioritizing data sovereignty and security. This development could accelerate the adoption of AI solutions in sectors like finance, manufacturing, and healthcare within China.

Stakeholders in the global AI market may view this as a signal of intensifying competition in the hardware space. Nvidia and other international suppliers could face increased pressure to adapt their strategies. The milestone reinforces the trend toward regionalization of AI infrastructure. Enterprise decision makers are now presented with viable domestic alternatives that match or exceed previous options in scale. The partial operations mean that benefits are not deferred but can be realized in the near term. Overall, the project contributes to a more diversified AI ecosystem.

Market analysts are closely watching how this development influences the competitive dynamics in the AI hardware market. The challenge to established players like Nvidia is implicit in the successful deployment of alternative infrastructure. Enterprise stakeholders may diversify their AI partnerships to include domestic providers for certain use cases. This could lead to a more balanced global market with multiple technology standards. The stock surge reflects not only the immediate news but also expectations of future growth. Zhipu AI's position in the enterprise AI space is strengthened by these developments.

Key metrics of the Zhipu AI data center project

Aspect	Details	Attributed Source
Power Capacity	1GW	Seoul Economic Daily
Number of Semiconductors	More than 10,000 domestic	Seoul Economic Daily
Stock Performance	Nearly 30% intraday surge on July 21, 2026	BigGo Finance
Household Equivalent	Approximately 750,000	KuCoin
Acquisition Target	XCore Sigma from Chinese Academy of Sciences	BigGo Finance

What expert or report-based reactions have been noted?

Reports from financial and economic outlets have emphasized the significance of the domestic chip utilization. The completion is seen as filling two critical capability gaps in hardware and software. Bloomberg sources have highlighted the 1GW capacity as a major achievement. This has fueled discussions on the viability of China's AI ambitions independent of traditional suppliers. The stock surge is interpreted as market validation of these efforts. Such reactions underscore the project's role in the ongoing evolution of the enterprise AI landscape.

According to sources cited by Bloomberg, Zhipu AI recently completed construction of a large-scale data center with 1GW of designed power capacity, fully equipped with China's domestically produced AI chips.Bloomberg via BigGo Finance report

The partial operation status indicates practical progress rather than just announcements. Analysts are likely to monitor how the facility performs in real-world training scenarios. The largest status among Chinese AI companies adds to the prestige of the accomplishment. This could inspire similar projects from competitors in the sector. The overall sentiment in the reports is one of cautious optimism regarding the future of domestic AI infrastructure. Enterprise AI strategies may increasingly incorporate such local options.

Looking ahead, the full operation of the data center will likely reveal additional insights into the performance of domestic AI chips at scale. This data will be valuable for further refinements and for the broader industry. The acquisition's impact on software optimization will be a key area of focus. Enterprise AI strategies may evolve to take advantage of these new capabilities. The milestone is part of a larger trend toward localized AI infrastructure that prioritizes resilience and innovation. Continued progress in this area will shape the future of AI adoption in enterprise settings.

What developments can be expected next for Zhipu AI?

With the data center in partial operation, full activation is anticipated to further boost model development capabilities. The integration of XCore Sigma's software is expected to yield immediate improvements in chip efficiency. Zhipu AI may expand the facility or develop additional sites to meet growing demand for its services. The stock performance provides capital for further investments in the AI ecosystem. Continued focus on domestic technologies will likely remain a priority. This positions the company to play a larger role in China's enterprise AI market.

Future iterations of the GLM models could demonstrate enhanced capabilities due to the new resources. Partnerships with other domestic entities might emerge to leverage the infrastructure. The milestone sets a benchmark for other AI firms to follow in building self-sufficient operations. Monitoring regulatory and market responses will be important for understanding the broader impact. The project contributes to the narrative of technological advancement in the region. Overall, the trajectory points toward sustained growth in domestic AI compute capacity.

1. Completion of the 1GW data center construction using domestic chips.
2. Initiation of partial operations for GLM model training.
3. Acquisition of XCore Sigma to optimize chip performance.
4. Report of nearly 30% stock surge on July 21, 2026.
5. Utilization of more than 10,000 domestic semiconductors in the facility.