Startup Modular has secured $100 million from General Catalyst, GV, SV Angel, Greylock, and Factory, taking the total raised funds to $130 Million. The funds will be utilized to expand their AI platform, enhance hardware support, and develop their programming language, Mojo.
Chris Lattner and Tim Davis co-founded Modular in 20222. They believed complex tech setups impeded AI. Modular simplifies large-scale AI by improving model performance on CPUs and, soon, GPUs. Additionally, Modular is cost-effective. The company has collaborated with popular frameworks like TensorFlow and PyTorch, boosting speeds up to 7.5 times faster than native setups.
Mojo is a key product of Modular. It’s a programming language blending Python’s ease with caching, adaptive compilation, and metaprogramming. Mojo will be accessible to everyone next month. Moreover, Modular simplifies the current AI complexity by tackling fragmentation between software and hardware. Despite being ambitious, the roughly 70-person company’s goals are feasible.
As AI demand stresses sustainability limits, decreasing the computing needs is important. Current generative AI models are 10 to 100 times larger than older ones, posing challenges for existing cloud infrastructure. AI demands have caused Microsoft to experience server hardware shortages, which will cause service disruptions.
The rising demand for AI hardware such as GPUs has propelled Nvidia’s market capitalization to $1 trillion, but the company has completely sold out its leading AI chips until 2024. Over 50% of top companies’ AI leaders encounter obstacles in implementing cutting-edge AI tools, based on a 2023 S&P Global poll.
AI applications include more than just high-performance acceleration; they involve data tasks like preprocessing and networking. Modular’s Mojo aims to integrate these tasks in one technology base without compromising on performance or scalability. Modular has gained approximately 120,000 developers and received interest from leading tech companies since its May product keynote. Modular’s approach resolves the complexity in software layers tied to specific hardware.