Together AI raised $800 million in a Series C round on July 1, 2026, lifting the "neocloud" to an $8.3 billion valuation as enterprises move open-source AI models into production. The round was led by Aramco Ventures, with chipmaker Nvidia, Vista Equity Partners and General Catalyst among the participants.
A cloud built for open models
Together AI's platform offers serverless inference, dedicated infrastructure and batch inference for open-source models, running on Nvidia GPUs. At its core is a proprietary engine called ATLAS, which uses adaptive speculative decoding — a lightweight model drafts responses that the main model then verifies and corrects — and which the company says can speed some inference workloads by up to 400%.
The business behind the round
The company disclosed that annual bookings surpassed $1.15 billion in the second quarter of 2026. It says customers save between 6x and 20x on inference costs compared with closed-model APIs, citing the customer-service firm Decagon as achieving a sixfold reduction after migrating to its platform.
Where the money goes
Together AI plans to use the capital to expand its products as it pushes to become a leading inference provider, and to scale its capacity dramatically — a footprint it expects to grow roughly 50-fold over the next five years. Alongside the equity, investors have committed more than 500 megawatts of compute capacity, to be financed independently of the round.
Open versus closed
The raise lands as open-weight models from DeepSeek, Alibaba's Qwen, Meta's Llama and Mistral narrow the gap with closed APIs, making a cloud specialized in running open models more valuable to enterprises that want control over cost and deployment. Together AI's pitch is that cheaper inference and the freedom to run models where a customer chooses will keep pulling that work its way.