Stanford adjunct professor and former startup founder Zain Asgar has successfully raised $80 million in Series A funding for his new venture, Gimlet Labs, with the backing of Menlo Ventures. Gimlet Labs claims it has developed an innovative “multi-silicon inference cloud.” This cutting-edge software allows AI workloads to be executed concurrently across various hardware types, including conventional CPUs, AI-optimised GPUs, and high-memory systems.
Asgar explained that their technology efficiently utilises the diverse hardware available, enabling different computational tasks, such as inference, decoding, and network calls, which are bound by varying resource requirements. Current chip technologies cannot perform all tasks ideally, which presents a gap that Gimlet Labs aims to bridge with its software solution.
The potential of AI workloads is largely untapped, with Asgar indicating that existing applications only utilise their deployed hardware 15% to 30% of the time. He highlighted the alarming waste of resources, estimating that hundreds of billions of dollars are squandered on idle infrastructure. To combat this inefficiency, Gimlet Labs aspires to enhance AI workloads’ performance by a factor of ten.
With a focus on orchestration software, Gediminas Labs aims to disassemble and distribute workloads across various chip architectures, optimising the use of each system. The company claims its solution can accelerate AI inference by three to ten times, all while maintaining the same cost and power usage. Furthermore, its software can adapt the underlying models to run across different computational architectures, selecting the most suitable chip for each segment.
Gimlet Labs has already forged partnerships with notable chip manufacturers, including NVIDIA, AMD, Intel, ARM, Cerebras, and d-Matrix, to incorporate their technologies into its offerings. While the product is not tailored for general AI app developers, it primarily targets major AI model labs and data centres.
Since its public launch in October, Gimlet Labs reported impressive early revenues of at least $10 million and has seen its customer base double in just four months. The company, however, has chosen not to disclose specific client identities, although it counts a significant model creator and a large cloud computing entity among its clients.
Asgar and his co-founders, including Michelle Nguyen, Omid Azizi, and Natalie Serrino, previously collaborated at Pixie—a startup that developed a prominent open-source observability tool for Kubernetes, later acquired by New Relic. The team gained traction after Asgar’s chance meeting with investor Tim Tully, leading to significant interest from venture capitalists swiftly following the initial funding round, which became oversubscribed.
To date, Gimlet Labs has raised a total of $92 million, bringing in contributions from notable angel investors, including Sequoia’s Bill Coughran and Stanford Professor Nick McKeown. Currently, the startup has a team of 30 employees and is well-positioned for growth in the AI sector.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
