For years, employees have been passing server rooms in the halls of the National Supercomputing Center in Tianjin, a vast state complex located just south of Beijing that contains some of China’s most potent computing hardware. Over 6,000 institutional clients nationwide use computing services from the center, which opened as China’s first significant supercomputing hub in 2009. Its devices are used by universities to run climate models. Flight simulators are performed by aerospace engineers. The kinds of computations that only supercomputers can perform are carried out by defense contractors. The setup appeared simple for a while: strong machinery, cautious access, delicate work. After then, an unidentified actor discovered a vulnerability in a VPN domain and used it to covertly transfer data for six months.
On February 6, 2026, an account going by the name FlamingChina appeared on Telegram and shared samples of what it said was a dataset that had been taken from the Tianjin plant. The gang claimed to have more than 10 petabytes of data, or about 10 million gigabytes, or roughly 10,000 times the storage capacity of a high-end consumer laptop. The samples included technical diagrams, animated simulations, documents labeled “secret” in Chinese, and visual representations of military systems, such as explosives and missiles. The group also asserted that the National University of Defense Technology, the Aviation Industry Corporation of China, and the Commercial Aircraft Corporation of China were among the clients from whom the stolen material originated. The cost was $3,000 for a peek of the whole file index. The asking fee, paid in bitcoin, was in the hundreds of thousands for full access.
Key Information: China Supercomputer Breach (2026)
| Field | Details |
|---|---|
| Target Facility | National Supercomputing Center (NSCC), Tianjin, China |
| Facility Founded | 2009 — China’s first major supercomputing hub |
| Number of Institutional Clients | 6,000+ (universities, state enterprises, defense contractors) |
| Alleged Attacker | Unknown group operating as “FlamingChina” |
| Data Volume Claimed | 10+ petabytes (~10 million gigabytes) |
| Breach First Made Public | February 6, 2026 — Telegram post with sample dataset |
| Entry Method Claimed | Compromised VPN domain + botnet for extraction |
| Duration of Intrusion | ~6 months before detection |
| Alleged Stolen Content | Missile schematics, aerospace simulations, radar tests, classified weapons data, fusion simulations |
| Data Classification | Samples reportedly marked “secret” in Chinese |
| Price Being Asked | Preview: ~$3,000 (10 Monero); Full access: hundreds of thousands (cryptocurrency) |
| Key Expert Review | Dakota Cary, SentinelOne — samples “exactly what I would expect to see” |
| Chinese Government Response | No official confirmation or denial as of April 2026 |
| Aftermath | Several senior Chinese defense officials reportedly removed after breach emerged |
After reviewing some of the leaked content, some cybersecurity specialists concluded that it was reliable. After reviewing the samples, Dakota Cary, a SentinelOne consultant who specializes in China’s technology industry, told CNN that they were “exactly what I would expect to see from the supercomputing center.” The diversity of the samples, which included military systems, bioinformatics, fusion simulation, and aerospace, represented the facility’s vast spectrum of institutional clients.
According to Cary, the attack strategy is more about architecture than technological expertise. After gaining access via the hacked VPN, the attackers set up what seemed to have been a botnet, a network of automated programs that surreptitiously gathered data over several months while rerouting it through several servers to evade discovery. “You can think of it as having a bunch of different servers that you have access to and you’re pulling data through this hole in the security,” he said to CNN.
If verified at the purported size, the intrusion would rank among the biggest intelligence breaches linked to China in recent memory. It is difficult to completely absorb the scale. If 10 petabytes of data were extracted over a six-month period without a security reaction, there may be a significant design error, inadequate network segmentation, or both.
According to one expert cited in the incident’s coverage, the system might not have strong barriers separating its various network segments, which would allow an attacker to gain access to far more than a conventional breach would. In a more direct online statement, another observer stated, “The Great Firewall may be solid like a pot lid, but everything else is like a sieve.” It did a good job of capturing the atmosphere, whether or not that is totally fair.

It’s worthwhile to sit with this particular irony. Western nations have long accused China of supporting state-sponsored cyber operations against government networks, international infrastructure, and intellectual property. The majority of such charges have been refuted by the nation. The nation that based a large portion of its geopolitical argument on information security now appears to be the victim of a breach that it is unable to formally confirm or deny.
Chinese authorities had not made any public remarks regarding the purported intrusion as of April 2026. Reports surfaced shortly after the leaked data attracted more attention that a number of senior defense program executives, including one tied to the J-20 fighter program, had been dismissed from their positions; however, the connection to the hack was not formally verified.
The amount of information this anecdote conveys about the nature of shared infrastructure is difficult to ignore. The number of firms that use the Tianjin facility determines its worth and attractiveness as a target. A single exploited entry point allows an attacker to simultaneously access the work of 6,000 clients sharing the same computing environment.
The facility was particularly dangerous when it failed because of the security model that made it effective. China is not the only country that can learn that lesson. There are centralized, high-performance computing environments in the US, Europe, and other places, and they all have to make the same compromise. The actual amount of data collected is still up for debate, and the breach has yet not been fully independently validated. The existence of the samples, their examination by professionals, and their authentic appearance are undeniable.