Nvidia chip gem Groq swipes Meta network chief Ola Kasali

Groq, an AI chip firm, hired Ola Kasali as head of network and data center engineering after nearly 12 years at Meta, where she rose from lead network engineer in 2017 to manager for network deployment and support in 2021. The hire follows Nvidia's virtual acqui-hire of Groq, which included licensing Groq's technology for a reported $20 billion and absorbing its leadership, and comes as Groq raised $350 million in August at a $3.5 billion valuation to expand its GroqCloud inference offering.

Nvidia-adjacent chip firm Groq hired Ola Kasali as head of network and data center engineering. Kasali joins the AI chipmakers after almost 12 years at Meta, where she joined as lead network engineer in 2017 before becoming manager for network deployment and support in 2021. Writing on LinkedIn https://www.sdxcentral.com/news/nvidias-ultra-low-latency-ai-inference-lpx-racks-hit-full-production/ , Groq’s new hire said: “As AI capabilities grow, so does the demand for inference and the infrastructure required to deliver it at scale. Groq’s first act was pioneering the LPU language processing unit chip, purpose-built for AI inference. In its second act, Groq is positioned to build the world’s leading AI inference cloud, delivering the performance, reliability, and global scale required to serve the next generation of AI workloads. “I’m looking forward to helping scale the network and datacenter infrastructure powering it, while building an exceptional team of infrastructure talent.” The hire comes as Groq continues to grow despite its virtual acqui-hire by Nvidia at the start of the year. That deal saw Nvidia license Groq's tech for a reported $20 billion https://www.datacenterdynamics.com/en/news/nvidia-to-license-tech-from-ai-inference-chip-company-groq-hire-its-leadership/ , as well as absorb Groq's then CEO, president, and other talent to advance its LPU footprint. The fruits of that deal came in March with the debut of the Nvidia Groq 3 LPU https://www.sdxcentral.com/news/nvidia-bets-big-on-bandwidth-with-groq-3-lpu-to-complement-gpus/ , designed to turbo-charge ultra-low-latency AI inference workloads. Nvidia's LPU-based rack-scale platform is in full production https://www.sdxcentral.com/news/nvidias-ultra-low-latency-ai-inference-lpx-racks-hit-full-production/ ahead of an anticipated launch later this year. Groq offers its static random access memory SRAM inference chips from 13 data centers across North America, Europe, the Middle East, and Asia Pacific. In August, the firm raised $350 million https://www.datacenterdynamics.com/en/news/groq-raises-350m-with-nvidia-participation-planned/ , at a $3.5 billion valuation, with plans to scale from 54 megawatts MW of compute to over 200MW in 2027 as it expands its GroqCloud inference offering. Nvidia was named as a planned participant in the fundraiser, with Groq declaring the boost would support medium and large clusters of Nvidia-accelerated computing for training and inference. Kasali, who also touts experience at Freddie Mac and Sprint prior to the latter's T-Mobile US merger https://www.sdxcentral.com/news/daily-roundup-judge-oks-t-mobile-sprint-merger/ , will look after a networking estate that includes Cisco Nexus 9000 switches. In promotional material released https://www.cisco.com/c/en/us/products/collateral/networking/cloud-networking-switches/nexus-9000-switches/nexus-9000-ai-networking-aag.pdf this year, Kasali’s predecessor Cameron Ferdinands commented that Cisco's wares allowed it “to move massive volumes of data with unparalleled speed and low latency that is foundational to unlocking the full potential of Groq’s innovative fast inferencing solutions.” Interestingly, Ferdinands has served as Nvidia’s head of network & datacenter engineering since February. With Kasali assuming the same title at Groq, it can be assumed that Groq still has some degree of autonomy and separation from Jensen Huang and company.