While Nvidia’s popular data center GPUs are making some of these security advancements possible, chip designers like Intel, AMD and Qualcomm are also changing up the security game with the introduction of the NPU, short for neural processing unit, in PCs.
The rise of AI chips in data centers and PCs is changing the way businesses can protect against cyberattacks and data breaches by introducing new security capabilities such as large-scale digital fingerprinting or rack-scale confidential computing.
While Nvidia’s popular data center GPUs are making some of these capabilities possible, semiconductor companies like Intel, AMD and Qualcomm are also changing up the security game with the introduction of the NPU, short for neural processing unit, in PCs.
[Related: Nvidia Touts New Storage Platform, Confidential Computing For Vera Rubin NVL72 Server Rack]
Designed to efficiently handle AI and machine learning workloads like matrix multiplication and vector operations, the NPU can run security workloads that previously took up a large chunk of CPU resources to improve system performance and even a laptop’s battery. In some cases, the NPU can enable capabilities that weren’t previously practical on a CPU.
As part of CRN’s AI Security Week 2026, here are five ways AI chips such as GPUs and NPUs are accelerating security advancements in data centers and client computing environments.
Confidential Computing For Heavy AI Workloads
Nvidia has been implementing confidential computing capabilities in its data center GPUs starting with the Hopper generation to protect AI workloads from unauthorized access. The company’s Blackwell GPUs are the first to support TEE-IO, which extends the Trusted Execution Environment of capable CPUs to peripheral devices such as GPUs. With Nvidia’s upcoming Vera Rubin NVL72 rack-scale platform, the company said it will have the ability to create a unified security domain that spans all 72 Rubin GPUs, 36 Vera CPUs and interconnects to protect massive AI workloads running on the entire rack.
Enabling Digital Fingerprinting Across The Data Center
Nvidia said its Morpheus software development kit takes advantage of the company’s GPUs to detect deviations in behaviors of every user, service, account and system across a data center, with the goal of stopping cyberattacks as early as possible. This is made possible with Morpheus digital fingerprinting AI workflow, which uses Nvidia’s BlueField DPUs to collect data across the data center and GPUs to analyze the data for anomalies. Nvidia said this enables 100 percent visibility across the data center and speeds up the detection of cyber threats from weeks to minutes.
Freeing Up CPU Headroom For Behavioral Detection
Acronis said its Cyber Protect Cloud software can analyze behavioral patterns on PCs to detect ransomware, zero-day exploits and other kinds of advanced cyberattacks without tanking system performance or a laptop’s battery by using the NPU in Intel’s Core Ultra chips. The software has been designed to utilize Intel’s OpenVino software to “offload heavy AI tasks such as behavioral heuristics and anomaly scoring” to the NPU, according to the company. This can free up CPU resources by up to 92 percent, Acronis said.
Enabling Better Protection Against Phishing
Bufferzone said it can provide better protection against phishing attacks than the Google Chrome and Microsoft Edge web browsers by using the NPU or GPU in the latest generations of PC processors from Intel, AMD and Qualcomm. The anti-phishing capability is part of the company’s Safe Workspace platform and acts as a lightweight browser extension that analyzes web pages “using deep-learning AI models to identify phishing attempts as they occur,” according to the company. By using the NPU or GPU, Bufferzone said its anti-phishing capability never sends data to the cloud and has 70 percent less latency than cloud-based approaches such as Google’s or Microsoft’s.
Powering Deepfake Detection
McAfee is taking advantage of the NPU in the most recent generation of PC processors from Intel, AMD and Qualcomm to detect deepfake videos by checking for AI-generated audio. According to McAfee, its Deepfake Detector feature uses advanced AI detection techniques to determine when the audio stream of a video playing in a web browser was generated by AI. The company said this feature requires an NPU with at least 40 trillion operations per second (TOPS) to ensure the PC continues to run smoothly.
