Advanced Micro Devices (AMD) and Korean telco KT have joined forces to invest $22 million in Moreh, an AI software developer that specializes in optimizing and creating AI models. With this latest Series B funding round, Moreh’s total raised capital now stands at $30 million. The startup, headquartered in Santa Clara and Seoul, aims to address the inadequacies of existing AI software tools by creating a more flexible AI infrastructure.
Key Takeaway
Moreh, an AI software developer, has secured $22 million in a Series B funding round led by AMD and KT. The startup’s flagship software, MoAI, offers compatibility with popular machine learning frameworks, enabling cross-platform usage. Moreh’s AI solutions allow for the building of flexible AI infrastructure, tackling the limitations of existing software. With its expertise in GPU and AI chip operation, Moreh provides accelerated performance and efficiency gains for AI developers, while reducing the time needed to train large AI models.
MoAI Software: Revolutionizing AI Infrastructure
Moreh’s flagship AI software, known as the MoAI, provides a similar functionality to Nvidia’s CUDA but operates seamlessly with popular machine learning frameworks like Meta’s PyTorch and Google’s TensorFlow. By making AI models compatible with these existing frameworks, Moreh enables previously Nvidia-exclusive applications and AI models to be run on other platforms as well.
According to Moreh, existing AI software solutions are well-suited for small-scale AI models that utilize only a few GPUs. However, as AI technology continues to evolve, larger-scale infrastructure is required to meet the growing demand. Many companies’ IT infrastructure and data architectures are currently ill-equipped to handle the training of such models.
With its AI solutions, Moreh empowers AI developers, data center operators, and AI chip makers to build more flexible AI infrastructure. This is particularly crucial considering the global scarcity of GPUs. Gangwon Jo, co-founder and CEO at Moreh, emphasized the importance of flexibility in AI infrastructure and how it can benefit the entire ecosystem.
Revolutionizing AI Model Training
One of Moreh’s key advantages is its ability to enable GPUs and other AI chips, such as neural processing units (NPUs), to efficiently operate AI models without requiring any code changes. Moreh’s AI software supports large language models (LLMs) like GPT-3 and T5, enabling seamless operation and accelerated performance.
In fact, Moreh claims that its offering has outperformed Nvidia’s DGX in terms of performance speed and GPU memory capacity. Their AI accelerator powered by the MoAI platform demonstrated 116% higher GPU throughput compared to Nvidia’s A100. Additionally, Moreh’s software significantly reduces the total time required to initiate training for large AI models, providing considerable efficiency gains for AI developers.
Positive Results and Future Plans
Moreh has already started generating revenue and aims to reach approximately $30 million by the end of 2023, according to Jo. Their partnership with KT, which began in 2021, has proven to be successful, showcasing the effectiveness of their technology in developing LLM and other AI models.
Recently, Moreh completed the training of a Korean-language-based LLM with 211 billion parameters. They plan to release this model as open-source later in the year, further contributing to the AI community.
The funding received in the Series B round will be utilized for research and development, product expansion, and the hiring of additional staff. Currently employing 70 individuals, Moreh aims to continue its growth trajectory in the AI software market. South Korean VC firms Smilegate Investment and Forest Partners, existing investors in Moreh, also participated in this funding round.
