Moreh Secures $22 Million Series B Round Led by AMD and KT to Enhance AI Infrastructure
Advanced Micro Devices (AMD) and South Korean telecom giant KT are among the key investors in Moreh, an innovative startup focused on developing AI software that enhances and creates optimized AI models. The company, with offices in Santa Clara and Seoul, has successfully completed a $22 million Series B funding round, raising its total funding to $30 million.
Moreh's flagship product, MoAI, is designed to streamline AI processes and is comparable to Nvidia's CUDA. However, it boasts compatibility with popular machine learning frameworks such as Meta’s PyTorch and Google’s TensorFlow, as well as previously Nvidia-restricted applications and AI models.
The startup asserts that while current AI software caters to small-scale models using a limited number of GPUs, it falls short when it comes to larger AI infrastructures. A recent report highlighted that as “AI rapidly matures at the enterprise level,” existing IT infrastructure and data architectures are often “unfit” for training advanced AI models.
Moreh's AI solutions are tailored for a diverse range of users, including AI developers, data center operators, and AI chip manufacturers, enabling them to construct a versatile AI infrastructure. This flexibility is increasingly vital, particularly amid a global GPU shortage, as explained by Gangwon Jo, co-founder and CEO of Moreh.
In just three years, Moreh's technology allows users to run AI models—including large language models (LLMs) like GPT-3 and T5—on GPUs and other AI chips, such as neural processing units (NPUs), without necessitating code modifications. NPUs are specialized chips designed to accelerate machine learning algorithms, while GPUs handle supercomputing tasks, AI, and extensive data analysis, where precision is vital.
Since initiating collaboration in 2021, KT has worked closely with Moreh to develop a cost-effective and scalable AI infrastructure powered by AMD GPUs, utilizing the MoAI software. KT claims that Moreh’s technology has outperformed Nvidia’s DGX in developing LLMs, showcasing superior performance speed and GPU memory capacity. Moreh reports that AMD’s MI250 Instinct accelerator, when paired with the MoAI platform, achieved 116% greater GPU throughput than Nvidia’s A100. Furthermore, Moreh states that AI developers can cut down the total time needed to start training large AI models by approximately ten percent with its software.
Recently, Moreh completed training a Korean-language-based LLM with an impressive 211 billion parameters, slated for release as open source later this year. Since generating revenue in 2021, the startup aims to reach an estimated $30 million by the end of 2023.
“The AI software ecosystem supporting AMD AI hardware is rapidly expanding, offering more choices for data scientists and AI users as they develop the models and solutions that will propel this industry's growth,” stated Brad McCredie, AMD’s corporate vice president of data center GPU and accelerated processing.
The new funding will be allocated to research and development, product expansion, and personnel growth, with Moreh currently employing 70 staff members. South Korean venture capital firms Smilegate Investment and Forest Partners also participated in this Series B funding round.
Ranking
