GPU Cryptojacking Expands Via SEO and AI Chatbots, Targeting High-Performance Silicon
A Microsoft-identified cryptojacking campaign is targeting high-performance GPU users via SEO poisoning and AI-generated recommendations. The malware uses DLL sideloading and process hollowing to deploy miners while actively evading detection by monitoring system load.
Microsoft has identified a persistent cryptojacking campaign active since at least March 2026 that specifically targets high-performance GPU users. According to Tom's Hardware, the operation leverages SEO poisoning and AI-generated software recommendations to distribute malware disguised as common PC utilities. By targeting hardware enthusiasts, gamers, and AI users, the attackers aim to compromise systems equipped with powerful discrete GPUs. The campaign utilizes a sophisticated infection chain that relies on DLL sideloading and process hollowing, allowing the malware to operate with high stealth while diverting compute cycles toward unauthorized cryptocurrency mining operations.
The distribution mechanism relies on luring users to attacker-controlled domains hosting malicious ZIP archives. These archives impersonate widely used diagnostic and utility software, including CrystalDiskInfo, HWMonitor, Display Driver Uninstaller, FurMark, K-Lite Codec Pack, and PDFgear. Microsoft noted that while traditional SEO poisoning directs traffic, some users were also led to these malicious portals via recommendations from large language model-based assistants. This indicates an emerging attack surface where AI-assisted discovery systems can inadvertently amplify the reach of malicious infrastructure, which currently spans over 150 domains, many utilizing the Dynu dynamic DNS service.
Once a victim executes the bundled installer, the malware exploits a standard Windows DLL sideloading technique. By placing a malicious autorun.dll in the same directory as a legitimate executable, the system automatically loads the malicious code without requiring further exploits. The attackers then deploy ScreenConnect, a legitimate remote-management tool, to establish persistent access. From there, the malware utilizes SimpleRunPE.exe to perform process hollowing, injecting mining code into trusted Microsoft-signed .NET utilities such as MSBuild.exe, InstallUtil.exe, and RegAsm.exe, effectively masking the malicious activity within standard system processes.
The mining payloads—specifically lolMiner, gminer, and SRBMiner-MULTI—are not embedded directly but are dynamically downloaded based on the system's specific GPU and CPU configuration. To maintain persistence and evade user detection, the malware implements rigorous anti-analysis checks. It scans for virtual machine artifacts, debugging tools like x64dbg and dnSpy, and forensic utilities such as Wireshark or ProcMon, terminating itself if these are detected. Furthermore, the malware monitors GPU utilization and system idle time, automatically suspending mining operations during heavy gaming or rendering workloads to prevent frame-rate drops or abnormal fan noise that would alert the user.
For engineers and system administrators, this campaign highlights the limitations of relying on traditional perimeter defenses when the software supply chain is compromised at the point of discovery. The abuse of legitimate tools like ScreenConnect and signed .NET binaries demonstrates a shift toward living-off-the-land techniques that bypass signature-based detection. As AI-assisted search becomes more prevalent, the integrity of information sources becomes a critical component of the security stack. Protecting high-performance silicon requires moving beyond simple malware scanning to include robust verification of software sources and continuous monitoring for anomalous GPU utilization patterns that suggest unauthorized resource diversion.