Together AI Raises $800M to Grow Its AI-Optimized Public Cloud

Together AI, a prominent cloud platform dedicated to optimizing the execution of open-source artificial intelligence models, has announced a substantial Series C funding round, securing $800 million from investors. This significant capital injection, led by Aramco Ventures with participation from Nvidia Corp., Vista Equity Partners, and General Catalyst, elevates the company's valuation to an impressive $8.3 billion. A cornerstone of Together AI's platform is its serverless inference service, designed to streamline the deployment of open-source AI models. This service liberates developers from the intricate task of configuring and managing graphics cards and network infrastructure, thereby simplifying the operational overhead associated with AI model deployment. The company boasts that its serverless environments achieve approximately double the performance of the fastest alternative solutions available in the market. Beyond its high-performance serverless offering, Together AI provides a suite of other inference services. These include two dedicated infrastructure options that offer enhanced reliability and customization capabilities, catering to specific enterprise requirements. Additionally, a Batch Inference service is available, which prioritizes cost-efficiency for models that do not demand immediate user interaction, potentially reducing prices by up to 50% for such workloads. The technological backbone of Together AI's platform relies on Nvidia chips and a proprietary software engine named ATLAS. This engine employs a machine learning technique known as speculative decoding to accelerate customer workloads. Speculative decoding works by integrating the main AI model with a lighter, secondary neural network. When a user input is received, the lighter algorithm rapidly generates a draft response. The primary model then reviews this draft, corrects any errors, and delivers the final output to the user, a process significantly faster than the main model generating the response independently. ATLAS further enhances this by automatically adapting the lightweight algorithm's configuration, addressing the common issue of fixed-configuration models becoming less accurate over time. The newly acquired capital will be strategically utilized to expand Together AI's infrastructure, with ambitious plans to increase its public cloud capacity by a factor of 50 over the next five years. The company also intends to further enhance its training and inference features, continuing its mission to provide a robust and efficient platform for the open-source AI community. Together AI's annual bookings reportedly exceeded $1.15 billion in the second quarter, and its platform is currently leveraged by thousands of organizations, including LG Inc.'s AI research lab, Cohere Inc., and the Mozilla Foundation.