Modular Raises $100M to Develop AI Models, Aims to Challenge NVIDIA

Modular, a startup for simplifying the development and optimisation of AI systems, secured a substantial $100 million in funding through a round led by General Catalyst, with participation from Google Ventures, SV Angel, Greylock, and Factory. 

The infusion brings Modular's total raised capital to $130 million for the startup which was co-founded in 2022 by Chris Lattner, a former Google employee, and Tim Davis, a colleague from Google's research division. The funds will be directed towards key initiatives, including product expansion, hardware support, and the enhancement of its programming language, Mojo, as emphasised by CEO Chris Lattner.

“This new funding will enable us to scale to the incredible customer demand we are seeing, continue to hire world-class talent and scale the release of our AI Engine and Mojo. We’re incredibly excited for the future,” Lattner said

Lattner and Davis shared the conviction that the potential of AI was being hindered by intricate and fragmented technical infrastructures. Modular's foundation was thus laid with a mission to simplify the complexities of creating and managing AI systems at a larger scale.

Modular introduces an engine designed to enhance the inferencing performance of AI models on CPUs, and soon on GPUs as well. This initiative aligns with the increasing demand for AI capabilities, which has put pressure on GPU supply, affecting companies from Microsoft to smaller AI startups. Nvidia, a dominant GPU supplier, also controls Cuda, a prominent software for crafting machine-learning applications, which is exclusive to Nvidia chips. In contrast, Modular's software strives to streamline AI developers' ability to train and run their models on chips produced by various companies, such as AMD, Intel, and Google.

Modular's flagship products encompass its engine for improved AI model performance and the Mojo programming language, designed to combine the usability of Python with features like caching and adaptive compilation techniques. These initiatives aim to tackle the issues of complexity and inefficiency commonly faced by developers in the AI landscape.

While challenges related to complexity and demand are palpable, Modular's endeavours exhibit considerable ambition. Lattner acknowledged that AI's compute power requirements are becoming unsustainable, leading to compute capacity shortages in certain instances. Modular seeks to address this issue, making AI technology more accessible, affordable, and sustainable for enterprises beyond just the large tech companies.

“We have been able to achieve tremendous momentum in only 20 months,” said Tim Davis, Modular co-founder and President. 

“The financing will allow us to accelerate our momentum even more, scaling to meet the incredible demand we have seen since our launch. We now have a community of more than 120K+ developers, including many of the world's leading technology companies, and 10K’s of enterprises that are excited to deploy Modular infrastructure,” he added.

Challenges 

However, there's a potential challenge in driving the widespread adoption of Mojo, given Python's entrenched status in the machine-learning community. Lattner believes in the potential of Mojo's unique benefits, asserting that AI applications entail more than just high-performance acceleration; they encompass end-to-end data processes. He believes that Mojo can unify these processes, enhancing performance and scalability.

Modular's endeavours have garnered attention and support. The company's community has expanded rapidly, and leading tech companies have already embraced its infrastructure. With an ongoing commitment to simplify and revolutionize the AI landscape, Modular's journey is off to an auspicious start.