Topics
Latest
AI
Amazon
Image Credits:Jonathan Kitchen / Getty Images
Apps
Biotech & Health
Climate
Image Credits:Jonathan Kitchen / Getty Images
Cloud Computing
DoC
Crypto
Enterprise
EVs
Fintech
Fundraising
Gadgets
gage
Government & Policy
ironware
Layoffs
Media & Entertainment
Meta
Microsoft
privateness
Robotics
Security
Social
distance
Startups
TikTok
transport
Venture
More from TechCrunch
Events
Startup Battlefield
StrictlyVC
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
Advanced Micro Devices ( AMD ) and Korean phone service KT are among the investors ofMoreh , which builds an AI software tool that optimize and creates AI models . The Santa Clara- and Seoul - based inauguration said it has raise a $ 22 million Series B round , bring its aggregate produce to $ 30 million .
Moreh says its flagship AI package , called the MoAI , is standardised toNvidia ’s CUDAbut is compatible with existing machine learning framework like Meta ’s PyTorch , Google ’s TensorFlow , and applications and AI modelling that antecedently could only fly the coop on Nvidia .
The startup believe existing AI software system is suited for minuscule - exfoliation AI models using only a few GPUs ( graphics processing units ) but is inadequate for more monumental AI base . According toa late news report , as “ AI is rapidly maturing at initiative scale , ” companies ’ IT base and data architectures are “ unfit ” for direct AI model .
Moreh ’s AI solutions appropriate users , including AI developers , data center operator and AI chip Godhead , to build more conciliatory AI infrastructure , of import given the general shortfall of GPUs globally , Gangwon Jo , co - beginner and CEO at Moreh , differentiate TechCrunch .
The three - year - honest-to-goodness startup ’s AI software package enable GPUs and other AI chips , such as nervous processing unit ( NPUs ) , to operate AI theoretical account , including large language model ( LLMs ) like GPT-3 and T5 , without any computer code variety , Jo explained . ( NPU is a specialized buffalo chip designed to quicken machine learning algorithmic rule , while GPU is used for supercomputing , AI , motorcar encyclopedism , and extensive data depth psychology where numerical preciseness is required . )
KT has been working with Moreh since 2021 to build a cost - good , scaleable AI infrastructure powered by AMD GPUs and MoAI software .
KT claims that Moreh ’s oblation prove its technology works and that its carrying into action in developing LLM and other AI framework surpassedNvidia ’s DGXin term of performance speed and GPU memory board electrical capacity . AMD ’s MI250 Instinct throttle powered by the MoAI political program designate 116 % mellow GPU throughput than Nvidia ’s A100 , allot to Moreh , which also said AI developers could reduce a tenth of the total time required to initiate training for tumid AI models via its software .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
Most lately , the startup completed the preparation of Korean - language - based LLM with 211 billion parametric quantity and will put out the model as open source later this twelvemonth .
Moreh started return gross in 2021 and direct to arrive at approximately $ 30 million by the destruction of 2023 , Jo told TechCrunch .
“ The AI software ecosystem supporting AMD AI ironware continues to turn , provide choice for data scientists and other users of AI as they progress the AI models and solution that will labor the proceed growth of this diligence , ” Brad McCredie , corporate frailty president of datum center GPU and quicken processing at AMD , said in a statement .
Moreh will apply the money for research and development , product expansion and hiring additional staff , which now has 70 employees . South Korean VC firms Smilegate Investment and Forest Partners , Moreh ’s existing investor , also participate in the Series B circle .
OctoML raises $ 85 M for it for its machine learning acceleration weapons platform
