Moreh Secures $22M Series B Funding with Support from AMD and Korean Telco KT

0:00AMD and Korean telecommunication giant KT have both doubled down on their investment into AI software and they've backed a new company called Marea. This is a really promising startup that is specializing in optimizing and creating AI models. So the Santa Clara and also Sal based startup recently raised$22 million in their series B and that is lifting its total investments to around $30 million so far. So what sets Marea apart I think is its innovative AI software named MoAI. So this flagship product stands out as a bit of a rival to NVIDIA's CUDA or CUDA. However, it has some compatibility with some popular machine learning frameworks like Google's TensorFlow and Meta's PyTorch.

0:47So more notably, I think it supports applications and AI models that were kind of traditionally restricted to NVIDIA, and now they're able to kind of work with them as well. So there is a perception, I think, right now in the tech community that while the existing AI software might be fit for, you know, small scale AI models requiring only a few GPUs, it doesn't really get the job done when faced with larger AI infrastructure. So as enterprises rapidly adopt AI, they often find that their IT infrastructure and data architectures aren't really primed for training AI models. I think this is kind of a sentiment that is echoed in a lot of different industry reports that came out recently.

1:30So Gang Wang Zhou, who's the co-founder and CEO of Maria, elaborated on kind of their AI solutions. He had a discussion recently with someone and he was kind of mentioning how essentially it facilitates AI developers, data center managers, and AI chip manufacturers to design a more adaptable AI infrastructure. This comes as a vital innovation, particularly in light of the ongoing GPU shortage worldwide. So it's definitely something that's needed right now. Joe also was kind of underscoring the software's ability to effortlessly power GPUs and other AI chips like neural processing units or, you know, NPUs to manage AI models, including some, you know, massive AI models like GPT-3 and T5.

2:13And they do all of this without, you know, the need for code altercations or alterations, sorry. So, yeah, I think it's fairly impressive. So KT, that's the Korean telecommunications company, and I think their association with Marea has been going on since around 2021. They already resulted in the creation of a cost-efficient AI infrastructure powered by AMD GPUs and MoAI software. KT essentially got Marea's technology. They highlighted that its efficiency in developing large language models and other AI models outstripped NVIDIA's DGX. And that was both in speed and also GPU memory capacity. So definitely an impressive piece of software.

2:57Maria also kind of showcased that AMD's MI250 Instinct Accelerator integrated with the MoAI platform. And they kind of demonstrated a, I think it was like a 116 % surge in GPU throughput in contrast to NVIDIA's A100. So definitely impressive, right? They're making some big plays in the space. I think they were also kind of showcasing their power. Maria recently completed the training of a Korean language-based large language model, which had around 211 billion parameters. They plan to roll out this model as soon as an open source model in the coming months, which will be really cool. I'm super happy to see that they're kind of committed to that whole vision of open source.

3:43Having started generating revenue just last year, Marea is, I think it's pretty ambitious, but essentially they're targeting a revenue of around$30 million by the end of this year. And I think as shared by Joe in a recent conversation, this is kind of like what their forecast is gonna be going forward. So Brad McCready from AMD also was really optimistic about the whole AI software landscape. And he said, quote, The AI software ecosystem supporting AMD AI hardware continues to expand, offering choices for data scientists and other AI action aficionados as they develop the AI models and solutions that will spearhead the industry's continued growth.

4:25This is I think this is so interesting, really. You're going to see AMD making these kind of plays and making these kind of investments into software that works with their with their chips as they really try to fight. They have very strong competition right now from NVIDIA, who's just by and large kind of run away with the industry. So I think right now a way that AMD can try to catch up is by essentially kind of investing in technology that is complementary and helps them to outcompete NVIDIA. So I think that's what we're seeing with this investment, particularly. um the recently secured funds that they just got essentially they're going to use them for research product expansion and augmenting uh their headcount right hiring more employees they currently have around 70 employees i think it's noteworthy that alongside new investors um their south korean vc firms like smilegate investments and force partners who had previously invested in maria also pitched in during the series b round so that's always a really good sign when um you you know, an investor, a venture capital firm or someone else invest in one round and they do a follow on investment in the next round.

5:26It really kind of shows the company's growing and those investors view the next investment as a way to, you know, continue to make returns. They don't think that the company's capped out or isn't going to continue making money. So definitely a good sign. And this is going to be a very interesting company to follow in the future.