Speedata, a Tel Aviv-based startup creating an analytics processing unit (APU) designed to speed up massive knowledge analytic and AI workloads, has raised a $44M Sequence B funding spherical, bringing its whole capital raised to $114M.
The Sequence B spherical was led by its current traders, together with Walden Catalyst Ventures, 83North, Koch Disruptive Applied sciences, Pitango First, and Viola Ventures, in addition to strategic traders, together with Lip-Bu Tan, CEO of Intel and Managing Companion at Walden Catalyst Ventures, and Eyal Waldman, Co-Founder and former CEO of Mellanox Applied sciences.
The APU structure focuses on addressing the precise bottlenecks of analytics on the computing stage, not like graphics processing units (GPUs), which have been initially designed for graphics and later modified for AI and data-related duties, based on the startup.
“For many years, knowledge analytics have relied on commonplace processing models, and extra not too long ago, corporations like Nvidia have invested in pushing GPUs for analytics workloads,” Adi Gelvan, CEO of Speedata, mentioned in an interview with TechCrunch. “However these are both general-purpose processors or processors designed for different workloads, not chips constructed from the bottom up for knowledge analytics. Our APU is purpose-built for knowledge processing and a single APU can substitute racks of servers, delivering dramatically higher efficiency.”
Speedata was based in 2019 by six founders, a few of whom have been the primary researchers to develop Coarse-Grained Reconfigurable Structure (CGRA) expertise. The founders collaborated with ASIC design consultants to handle a elementary drawback: knowledge analytics have been being carried out by general-purpose processors. If the workloads grew too complicated, they may must faucet into tons of of servers. The founders believed that they may develop a single devoted processor to perform the duty quicker utilizing much less vitality.
“We noticed this as a possibility to place our a long time of analysis in silicon into remodeling how the trade processes knowledge,” Gelvan mentioned.
Its APU at the moment targets Apache Spark workloads, however its roadmap contains supporting each main knowledge analytics platform, based on the corporate CEO.
“We intention at turning into the usual processor for knowledge processing—simply as GPUs turned the default for AI coaching, we would like APUs to be the default for knowledge analytics throughout each database and analytics platform,” Gelvan advised TechCrunch.
The startup says it has quite a few massive corporations testing its APU, although it declined to call them. The official product launch is ready for the Databricks’ Knowledge & AI Summit within the second week of June. Gelvan mentioned that they are going to publicly showcase its APU for the primary time on the occasion.
Speedata claims a selected case the place its APU accomplished a pharmaceutical workload in 19 minutes, which was considerably quicker than the 90 hours it took when utilizing a non-specialized processing unit, leading to a 280x velocity enchancment.
The startup mentioned it has achieved a number of milestones since its final fundraising, together with finalizing the design and manufacturing of its first APU in late 2024.
“We’ve moved from idea to testing on a field-programmable gate array (FPGA), and now we’re proud to say we have now working {hardware} that we’re at the moment launching. We have already got a rising pipeline of enterprise clients eagerly ready for this expertise and have been able to scale our go-to-market operations,” Gelvan, mentioned.