Together AI

Together AI is a San Francisco-based GPU cloud and full-stack AI platform founded in 2022 by Vipul Ved Prakash (CEO), Ce Zhang (CTO), Chris Ré, Tri Dao (Chief Scientist), and Percy Liang. Its positioning sits between the pure-metal neoclouds (CoreWeave, Crusoe, Lambda) and the model-API vendors: customers can rent bare GPU clusters by the hour, buy dedicated single-tenant inference endpoints, or hit a serverless per-token API against a catalog of open-source and partner models. That three-lane packaging, backed by systems research the founders brought from Stanford (FlashAttention, Hyena, RedPajama), is the company's main distinctive versus a rack-focused operator.

Commercially the company is inference-and-fine-tuning first, with a training-cluster side business. Published pricing puts on-demand H100 clusters at $3.99 per GPU hour, dedicated HGX H100 endpoints at $5.49 per hour, and B200 dedicated at $8.99 per hour, with reserved terms of 7 to 180-plus days pulling H100 rates down toward roughly $3.19. Customers named on the site include Salesforce, Cohere, ElevenLabs, Cursor, Decagon, and DeepMind-adjacent teams; a dedicated GPU cluster program for Y Combinator startups is one of the more visible distribution moves. Nvidia is the anchor silicon partner (Together is listed as an Nvidia Preferred Partner) with early B200 availability being the current headline.

Capital-wise Together raised a $305M Series B in early 2025 (Salesforce Ventures, General Catalyst, others), followed by an $800M Series C the company announced in 2026 to expand open-source AI infrastructure. Prior backers across the stack include NEA, Kleiner Perkins, Coatue, Lux Capital, and Nvidia's venture arm. The company has not publicly disclosed a total operational GPU count or MW footprint; it operates through colocation partners rather than owned campuses.