Tencent Cloud lost a customer's entire data to a firmware bug, offered 1% of the claim

On July 20, 2018, the operating-system cloud disk of a single Tencent Cloud CVM rented by Beijing Qingbo Data Control Technology — the startup behind the WeChat media account Qianyan Shukong (前沿数控), a machinery-industry outlet — suffered filesystem metadata corruption. Tencent Cloud attributed it to a silent error caused by a firmware bug in the physical hard drive: data read back differed from what was written. Filesystem and vendor experts could not recover the data. After eight months on the platform, the customer lost everything, including backups.

The dispute went public on August 5 when the customer posted on Weibo. Tencent published its response: refund ¥3,569 of actual spend — the maximum its service agreement allowed — plus ¥132,900 in cash or credits, which it noted was 37 times everything the customer had ever paid. The claimed loss was ¥11.016 million. Tencent said the customer had also demanded investment and traffic support; the customer said it asked only after being told recovery was hopeless. The next day Tencent acknowledged employees had broken data-relocation rules and announced improvements — but no immediate solution.

Industry engineers picked apart both sides. The customer had bought a single CVM with only a system disk — no data disk, no load balancing, no off-cloud backup — and Tencent's heavily advertised 99.9999999% data reliability and three-replica storage applies to cloud data disks, not clearly to an OS disk. The customer's founder countered that Tencent marketed one-in-a-billion failure odds with triple replication, so he believed no further backup was needed. The episode turned into China cloud computing's textbook argument about what reliability numbers actually cover.

One media startup lost all data including backups; it claimed ¥11.0M (~$1.6M); Tencent offered a ¥3,569 refund plus ¥132,900 — about 37× the customer's cloud spend

Tencent ran data-relocation operations on drives with buggy firmware and later admitted employees had broken operating rules — the improper operations behind the silent error.

The advertised 99.9999999% reliability and three-replica storage did not cover a single CVM system disk, so the headline number created confidence the architecture could not back up.

The contract capped liability near fees paid — about 1% of the claimed loss — and the customer kept no independent backup, leaving no financial or data fallback.

The compensation plan arrived fourteen days after the loss and was read as dismissive, converting an operations failure into a public trust crisis for China's number-two public cloud.

Nine-nines reliability is a marketing number, not an architecture: one VM, one OS disk, no backups turns a freak failure into total loss. Check what the SLA covers — and what the contract pays.

Tencent took every drive with the buggy firmware offline, and its August 7 statement announced operational improvements while admitting rule-breaking during data relocation. The money gap — ¥136,469 offered against ¥11.016 million claimed — was never publicly reconciled. The case became China's cloud industry's two-line lesson: keep backups outside your provider no matter what the SLA says, and read reliability figures as covering specific products, not your architecture. Tencent Cloud kept its number-two position; the case still surfaces whenever providers market nine-nines.