Codeium is an AI-powered coding assistant that serves suggestions in the context of an app’s entire codebase, supporting around 70 programming languages and integrations with Microsoft Visual Studio and JetBrains. The company offers a generous free tier and provides self-hosted and hybrid deployment options for enterprises wary of exposing proprietary code. Codeium says it never trains its proprietary model on user data, encrypts all data transmission, and filters out non-permissively licensed code with post-generation attribution filtering and logging. The startup reports more than 700,000 users and over 1,000 enterprise customers (including Anduril, Zillow and Dell), and revenue reached eight figures this year. Founded in 2021 by Varun Mohan and Douglas Chen, Codeium began as Exafunction and pivoted to generative coding in 2022. The company plans to ramp up R&D and growth and expand headcount from about 80 to 120 by 2025. Codeium provides a generative AI-powered coding toolkit that leverages proprietary code-biased large language models to reduce inefficiencies and optimize developer productivity. The product supports over 70 programming languages and integrates with more than 40 IDEs, including Visual Studio Code, the JetBrains suite, Eclipse, and Jupyter Notebooks, offering in-editor autocomplete, chat, and context-aware capabilities. Its models write over 44% of newly committed code for a 300k+ individual user base. Enterprise features include SOC2 Type 2-compliant SaaS or self-hosted deployments (VPC or on-prem), assurances that customer code is never saved or used to train the public system, SCM integration for repository context and optional local model fine-tuning, and real-time analytics dashboards. Customers named in the article include Anduril, Clearwater Analytics, and several Fortune 500 companies, and Codeium is working with partners such as Atlassian. The company plans to use the new funding to grow engineering and sales teams and to expand systems and solutions targeted at enterprise software development needs. Exafunction builds a platform to abstract away the complexity of using accelerated hardware (GPUs and other accelerators) to train and run deep-learning workloads. Customers can use Exafunction as a managed service or deploy its software in a Kubernetes cluster; the technology dynamically allocates resources and can move computation onto cost-effective hardware such as spot instances. At a high level the company leverages virtualization to run AI workloads with limited hardware availability, aiming to increase utilization and lower infrastructure costs. Exafunction says it already manages GPUs for sophisticated autonomous-vehicle companies and cutting-edge computer-vision organizations. The founders plan to use new capital to expand the team, deepen the product, and optimize runtimes for latency-sensitive applications like autonomous driving and large-scale video inference. The company emerged from stealth with $28 million in venture capital backing, positioning its subscription-based platform as an alternative to higher-cost cloud-only approaches.

Total raised

$243M

Funding rounds

4

Latest round

Series C

Latest activity

Aug 2024