🔊CEOInterviews

Sachin Katti on agentic AI

From Intel Technology Tour 2025: Panther Lake is Built for Agentic AI | Talking Tech | Intel Technology · · Intel Technology

“The way I think about it is agentic AI is really composed of three things. First you have a model — today that's typically a large language model — they reason and reflect; agents interact with the environment and tools; and lastly they also use quite a bit of memory, both short-term and long-term.”

Sachin Katti
Senior Vice President, Chief Technology and AI Officer & GM of Network and Edge Group, Intel
agentic AIlarge language modelsmemoryAI architecture

On , Sachin Katti, Senior Vice President, Chief Technology and AI Officer & GM of Network and Edge Group at Intel, spoke about agentic AI during Intel Technology Tour 2025: Panther Lake is Built for Agentic AI | Talking Tech | Intel Technology on Intel Technology.

Intel Technology Tour 2025: Panther Lake is Built for Agentic AI | Talking Tech | Intel Technology
Watch on YouTube
Intel Technology Tour 2025: Panther Lake is Built for Agentic AI | Talking Tech | Intel Technology
Intel Technology
Watch on YouTube
You've probably heard of agentic AI, but do you understand how it works? In the latest chapter of our coverage of Intel Technology ...
Sachin Katti

About Sachin Katti

Senior Vice President, Chief Technology and AI Officer & GM of Network and Edge Group · Intel

At the Intel Technology Tour 2025, Sachin Katti described Panther Lake as being "designed for" agentic AI, stating that it delivers up to 180 TOPS of performance, divided between the CPU (10), MPU (50), and GPU (120). He characterized agentic AI as consisting of large language models that reason and reflect, agents that interact with tools, and memory usage. Katti said that the OpenVINO stack had improved, making it "twice as fast to first token, six times better on memory consumption, and about twice as fast on tokens per second" compared to the 2024 stack. He also noted that Intel has pre-converted and pre-quantized more than 300 models, which he said are available on Hugging Face as "Intel-optimized models." At Mobile World Congress 2025, Katti discussed the Xeon 6 family of processors, stating that the Xeon 6 SoC with P-cores delivers a 2.4x increase in capacity and a 70% decrease in power consumption for vRAN and cloud RAN workloads. He cited partners including Nokia, Samsung, and Ericsson, who he said reported improvements in power reduction, capacity, and performance per watt. Katti argued that AI should not be treated as a separate workload and that customers cannot afford to deploy a new architecture for AI, emphasizing that Intel's approach is to run AI inference on the same SoC that runs the network. He described the telecom industry as being on the "cusp of reaping all the benefits of automation because of the advent of AI," and said that the industry is shifting toward questions of how to monetize the network.

Profile compiled from Sachin Katti's verified public interviews and appearances. See all quotes & transcripts →

More from Sachin Katti Intel (INTC) Full Transcript Explore All Executives