Speedata, a chip startup competing with Nvidia, raises a $44M Series B

Speedata, a promising startup based in Tel Aviv, has successfully raised $44 million in a Series B funding round, boosting its total funding to $114 million. This round was spearheaded by established backers such as Walden Catalyst Ventures, 83North, Koch Disruptive Technologies, Pitango First, and Viola Ventures, alongside strategic investors including Lip-Bu Tan, CEO of Intel, and Eyal Waldman, co-founder and former CEO of Mellanox Technologies. The company is focusing on the development of an analytics processing unit (APU) that aims to enhance the performance of big data analytics and AI tasks. Unlike traditional graphics processing units (GPUs) that were adapted for these workloads, Speedata’s APU is specifically engineered to mitigate bottlenecks at the computing level. "For decades, data analytics have revolved around standard processing units, and GPUs have been pushed into this space by companies like Nvidia," stated Adi Gelvan, CEO of Speedata, in an interview with TechCrunch. "However, these are general-purpose processors, not chips built from the ground up for data analytics. Our APU is designed for efficient data processing, and one unit can outperform entire server racks, offering remarkable performance improvements." Founded in 2019 by a team of six, including pioneers of Coarse-Grained Reconfigurable Architecture (CGRA) technology, Speedata's founders aimed to solve a critical issue where complex analytics tasks often required multiple servers. They envisioned a dedicated processor that could handle these tasks more swiftly and with lower energy consumption. "This was our chance to leverage decades of research in silicon to transform data processing in the industry," Gelvan commented. Currently, Speedata’s APU targets Apache Spark workloads, with plans to support all major data analytics platforms in the future. Gelvan expressed ambitions for the APU to become the industry standard for data processing, similar to how GPUs became essential for AI training. Several major companies are currently testing the APU, although Speedata has chosen not to disclose their identities. The official unveiling of the APU is scheduled for the Databricks' Data & AI Summit in mid-June, where it will be showcased for the first time. In a notable demonstration, Speedata highlighted that its APU completed a pharmaceutical workload in just 19 minutes, a staggering improvement over the 90 hours required with a non-specialized processor, marking a 280-fold speed enhancement. The startup has also reached significant milestones since its last funding round, including finalizing the design and manufacturing of its inaugural APU, expected to be operational by late 2024. "We’ve transitioned from concept to testing on a field-programmable gate array (FPGA), and we are excited to introduce our working hardware. With a growing list of enterprise clients eager for our technology, we are prepared to scale our market operations," Gelvan added.