Hacker claims massive breach of China's Tianjin supercomputer

If you picture a state supercomputing center as a digital Fort Knox, it turns out the intruder didn't blow open the front vault. They reportedly slipped through an unlocked side window and spent six months quietly hauling everything out.

A hacker operating under the handle FlamingChina allegedly breached China's National Supercomputing Center (NSCC) in Tianjin, claiming to have stolen over 10 petabytes of classified defense records, missile schematics, and aerospace research. Cyber experts at SentinelOne who reviewed leaked samples described the files as credible, and the perpetrator is now attempting to sell full access for hundreds of thousands of dollars in cryptocurrency.

Siphoning 10 petabytes under the radar

The Tianjin supercomputing hub opened in 2009 and currently provides infrastructure for more than 6,000 clients, including top defense agencies, the Aviation Industry Corporation of China, and the National University of Defense Technology.

According to cybersecurity researcher Marc Hofer, who reached the alleged hacker via Telegram, the initial intrusion was surprisingly basic: access was gained through a compromised VPN domain. Once inside, the attacker deployed a botnet to slowly pull data across multiple servers simultaneously.

This distributed exfiltration took around six months to move 10 petabytes. As SentinelOne consultant Dakota Cary pointed out, the technique was not technically mind-blowing, but its architecture was effective. Pulling tiny streams across dozens of endpoints is far less likely to trip automated alarms than dumping huge files to a single IP address.

Samples posted on Telegram on February 6 featured Chinese files stamped "secret," alongside missile schematics, simulation animations, and technical blueprints. Cary noted the materials align exactly with the kind of heavy computational workloads customers outsource to a centralized facility.

Why it matters

We think this breach highlights a massive, overlooked vulnerability in the global race for centralized computing power.

Supercomputing centers and giant AI clusters are treated as shared digital utilities. Thousands of organizations rely on them because building and maintaining private high-performance infrastructure is wildly expensive. But that shared convenience creates a single point of failure. A single compromised VPN connection inside one facility can expose sensitive work belonging to thousands of unrelated tenants.

For enterprise builders and research labs, this should be an immediate wake-up call to audit data segregation. Handing proprietary simulation data or foundational model weights to third-party supercomputing hubs requires zero-trust verification and local encryption, not blind faith in the provider's perimeter defense.

Furthermore, 10 petabytes is far too large for an ordinary cybercriminal to parse. As researchers noted, foreign intelligence agencies are virtually the only buyers with the computational capacity to make sense of that much raw telemetry and engineering data. This incident will likely force state-backed computing facilities worldwide to aggressively clamp down on remote access and rethink how multi-tenant infrastructure handles classified workloads.