Together AI, which helps pre-train and fine-tune open source foundation models more efficiently, raised a $102.5M Series A led by Kleiner Perkins
I spoke with Together AI CEO Vipul Ved Prakash — Together AI, a startup that's helping companies pre-train and fine-tune open source foundation models …
Context & Ripple Effects
This Series A backed Together AI's initial pitch: making pre-training and fine-tuning of open-source foundation models more efficient. The company subsequently broadened into developer access to Nvidia training capacity through a $106M round at a $1.25B valuation, showing how model tooling and compute access became intertwined in its offering.
Later financings—from a $305M raise for AI computing access to an $800M round—place this early investment in a wider capital-intensive buildout around serving open-source-model users.
First-order effects
- Together AI gains capital to develop and deliver its pre-training and fine-tuning services, while Kleiner Perkins becomes the lead institutional backer of the company.
- Companies using open-source foundation models gain another funded provider focused on reducing the operational burden of adapting those models.
Second-order effects
- As Together AI expands from model workflows toward compute access, rivals serving developers face pressure to package infrastructure and model customization more tightly rather than offer either in isolation.
- The funding helps establish a path in which demand for open-source-model services translates into demand for the underlying accelerator capacity needed to train and fine-tune them.
Third-order effects
- If this pattern persists, AI infrastructure platforms may increasingly compete on an integrated stack—compute access, model adaptation, and deployment support—rather than on raw infrastructure alone.
- The later escalation in Together AI's funding suggests that providers in this segment can become progressively more capital-dependent as customer demand shifts from experimentation to sustained model training and serving.
The trend: This is an early example of AI compute commercialization: venture-backed platforms turning open-source model adoption into recurring infrastructure and customization demand.