Together AI raises $800M Series C, secures over 500 MW compute

Together AI announced on July 1, 2026, an $800 million Series C round from investors including Aramco Ventures, NVIDIA, Vista Equity, General Catalyst, Emergence Capital, SE Ventures, Pegatron, Salesforce Ventures, March Capital, DTCP Growth, Lux Capital, Geodesic, PSP Partners, and others. Alongside the equity, the company secured commitments for over 500 MW of compute capacity, to be capitalized independently by new investors to support expected compute growth in the coming years.

Production economics and open models

The company positions the round as a response to the shift of AI from occasional tools to core production infrastructure. As agents write code, resolve customer issues, and automate workflows, inference demand scales, and the cost structure of closed frontier LLMs becomes unsustainable in production. Together AI says companies building with open-weights models routinely achieve 6x to 20x lower costs while maintaining equal or better performance. As an example, Decagon reduced its inference costs sixfold after moving to Together AI.

Together AI has become the production platform for open and custom AI, combining frontier models with continuous innovation across the inference stack. The company cites open-weights models like DeepSeek, Nemotron, MiniMax, Kimi, and GLM as having closed the quality gap with proprietary frontier models while giving developers freedom to customize and fine-tune them.

Research-to-production pipeline

Together AI describes itself as a research-driven company, with a research-to-production pipeline that has accelerated over the past several quarters. The company announced FlashAttention-4 for NVIDIA Blackwell, Together Megakernel, and together.compile to bring kernel-level optimization to production workloads. It also expanded its post-training APIs to support tool calling, reasoning, and vision-language models. Together AI says it launched the fastest endpoints for leading frontier open-source models and has become one of the largest producers of AI tokens in the world.

The company's platform covers inference, training, and accelerated compute, offering choice, control, advantaged economics, and frontier performance for the full suite of generative AI production capabilities. Together AI says it is trusted by thousands of customers, including Cognition, Decagon, Eleven Labs, Cursor, and Suno.

For inference engineers, the round underscores the importance of kernel-level optimization and cost per token in production. The 500 MW compute commitment suggests a scale-up in serving capacity, likely for large-scale open-weights deployments. The focus on FlashAttention-4 and Megakernel indicates that software optimization remains a key lever for reducing inference costs, alongside hardware utilization. The company's emphasis on open models and customization points to a market where fine-tuning and tool calling are becoming standard requirements for production AI workloads.