The demand for computing power to run AI models is skyrocketing. Companies in this sector face two main hurdles: sourcing the necessary chips and deploying them into revenue-generating data centers. General Compute, a new inference-focused neoclaud startup, is offering a unique solution. The company raised $15 million in a funding round led by FUSE VC, valuing it at $60 million, according to TechCrunch reports.
While demand for GPU chips is high, experts note they are not the most efficient solution for running (inference) pre-trained AI models. Alongside giants like Nvidia, Groq, and Cerebras, General Compute has focused on SambaNova chips, backed by Intel. These specialized chips are significantly faster than GPUs, capable of generating 600-700 tokens per second.
General Compute has already ordered $300 million worth of SN50 chips and announced it will be the first neoclaud provider to deploy them. The advantage of these chips is that they are air-cooled rather than water-cooled and consume less energy. This allows them to be installed in existing data centers without new infrastructure, or even in facilities previously used by cryptocurrency miners.
The company launched its cloud service last week and claims to demonstrate the highest speed for running the open-source MiniMax 2.7 model. Investors are comparing the partnership between SambaNova and General Compute to the successful relationship between Nvidia and CoreWeave.
Read “Zamin” on Telegram!
