Together AI raises $800M at $8.3B valuation to expand its neocloud for open-source AI models
The AMW Read
Novelty 2: moves Together AI into top-tier neocloud status with $8.3B valuation; Significance 2: validates neocloud market at segment-level scale, with impacts on cloud AI competition.
Together AI raises $800M at $8.3B valuation to expand its neocloud for open-source AI models
Together AI, a neocloud provider renting Nvidia GPU clusters for open-source AI models, announced an $800 million Series C at an $8.3 billion valuation. The round was led by Aramco Ventures, with participation from Vista Equity Partners, General Catalyst, Emergence Capital, Nvidia, and others. The company claims annual bookings exceeding $1.15 billion, and counts Cursor, Cognition, and Decagon among thousands of paying customers. The raise follows a $305 million Series B 16 months ago and comes as enterprise adoption of open-source models via neoclouds triples year-over-year.
Why it matters: This round crystallizes the neocloud-as-hyperscaler-alternative pattern. As enterprises seek cost-effective alternatives to proprietary frontier models, neoclouds like Together AI capture the fastest-ARR-ramp in infrastructure, rivaling traditional cloud GPU margins. The participation of Nvidia and sovereign capital (Aramco Ventures) signals both compute supply-chain alignment and geopolitical diversification of AI infrastructure. Together AI’s growth validates that open-weight models are not just a developer curiosity but an enterprise procurement category, potentially reshaping the cloud AI revenue split.
Grounded expert take: This funding is a structural signal that the capital-compression arc in AI infrastructure is accelerating. Together AI’s $8.3B valuation — up 2.5x from its Series B — shows VCs are betting that the neocloud layer will disintermediate the big three cloud providers in AI workloads. However, as more neoclouds raise massive rounds (Upscale AI, TensorWave), the segment risks a capital-glut that could compress margins. The key moat will be customer stickiness through fine-tuning and managed inference, not just raw GPU access.
