Ukrainian Forces Bomb Yandex Data Centers, X Platform's Russian Bot Traffic Drops to 0.4%
After three consecutive days of drone strikes on three Yandex data centers in Russia, Cloudflare monitoring shows that bot traffic from the Yandex network (AS13238) has plummeted to 0.4% of its peak over the past two weeks.
On October 8, drones hit the data center in Sasovo, Ryazan Oblast, causing a fire and complete shutdown of the facility. This site housed two of Yandex's three supercomputers (Lyapunov and Chervonenkis), used for training the YandexGPT large model, with chips based on NVIDIA A100. The company confirmed that core services were disrupted, and some users could not access cloud backups.
On October 9, the largest data center in Kaluga was attacked, resulting in multiple hardware modules being completely disabled. The facility has a designed capacity of 63 megawatts and can accommodate over 3,800 racks, covering approximately 130,000 square meters. Yandex stated that some infrastructure was taken out of service, and shares of the Moscow Exchange company fell by about 4% at one point.
In the early hours of October 11, the 50-megawatt data center in Vladimir (designed to accommodate 2,880 racks) was completely shut down after being hit by drones. Over 80 Yandex Cloud services were interrupted, including Compute Cloud, object storage, managed Kubernetes, PostgreSQL, ClickHouse, YandexGPT API, Alice voice assistant, music, and smart home services. The third supercomputer, Galushkin, is located here.
These strikes are described as a response to previous Russian attacks on Ukrainian data centers. Yandex has removed some data center locations from its maps and stated that it is working around the clock to repair, but the service restoration time is difficult to predict. NetBlocks data shows that Yandex Cloud network traffic has dropped to about 30% of normal levels.
From a market mechanism perspective, the event directly cut off the availability of Yandex's cloud infrastructure, leading to a sharp decline in automated traffic relying on its servers (including platform comments and crawlers). Beneficiaries include ordinary user experiences and content distribution on social platforms like X, while those under pressure are Russian automated operators and cloud customers relying on Yandex's computing power and network exit.
Source: Public Information
ABAB AI Insight
Since the 2010s, Yandex has invested heavily in building its own data centers and AI computing power. In 2021, it pushed the Chervonenkis supercomputer into the top 20 of the global TOP500 and trained its own large models using NVIDIA A100 clusters, attempting to maintain a domestic cloud and search monopoly under sanctions. The company has repeatedly emphasized that data centers are "critical infrastructure" and disclosed in 2021 that Sasovo hosts two core supercomputers.
In terms of capital pathways, Yandex has highly concentrated its cloud and AI training in a few large domestic parks (Sasovo, Kaluga, Vladimir), while selling computing, storage, and YandexGPT API to enterprise customers. This concentrated deployment reduces cross-region redundancy costs but turns physical nodes into single points of failure. After Ukraine conducted precise strikes using FP-1 drones, the company was forced to advise customers to migrate data, exposing the cloud business's direct dependence on physical availability.
Similar cases include Amazon and Google being forced to advise clients to migrate to other locations after attacks on data centers in the Middle East in the 2020s, as well as Chinese companies' excessive reliance on single parks when expanding overseas. Currently, Yandex is in a stage of "post-expansion centralized control": cloud revenue and AI model training are highly bound to a few domestic nodes, and physical strikes can simultaneously weaken search, cloud, and automated traffic capabilities.
Structurally, this belongs to the reconstruction of the industrial chain and the transfer of pricing power. After physical nodes are directly destroyed, the availability of cloud services and automated exit traffic declines simultaneously, and the domestic computing power and network exits originally controlled by Yandex are forced to be redistributed. The mechanism is that modern information warfare has regarded data centers as targets that can be attacked on par with oil refineries, and centralized computing architectures lose geographical depth protection after the precision of drones improves.
ABAB News · Cognitive Laws
Concentrated computing power equals a single point target
When a physical node is cut off, digital traffic stops
Automated advantages are built on infrastructure that has not been hit