Quick Facts
- Together AI closed an $800 million Series C on July 1, 2026, at an $8.3 billion valuation, led by Aramco Ventures with participation from NVIDIA, Salesforce Ventures, and Vista Equity Partners.
- Annual bookings crossed $1.15 billion last quarter, and customers including Decagon report cutting inference costs sixfold after switching from closed-model APIs.
- A new IBM Institute for Business Value study finds 71% of enterprises say switching their primary AI vendor would be difficult, and 91% say they don’t fully understand their AI vendor dependencies.
Together AI has closed an $800 million Series C at an $8.3 billion valuation, signaling that open-weight AI inference has moved from an experimental option to production infrastructure for enterprise customers.
The round was led by Aramco Ventures, the venture arm of Saudi Arabia’s state oil company. NVIDIA, Vista Equity Partners, General Catalyst, Emergence Capital, Schneider Electric’s SE Ventures, Salesforce Ventures, and SentinelOne’s S Ventures also participated.
Together AI’s previous round was $305 million at a $3.3 billion valuation roughly 16 months ago. The company’s valuation more than doubled in that period.
What Together AI Actually Does
Together AI does not build foundation models. It builds cloud infrastructure that lets enterprises run open-weight models from other developers, including DeepSeek, Nemotron, MiniMax, Kimi, and GLM, on NVIDIA GPU clusters through an OpenAI-compatible API.
The company has secured commitments for over 500 megawatts of compute capacity from investors, funding roughly 50-fold capacity growth over five years. That scale of supply guarantee was previously available only from hyperscalers like AWS.
CEO Vipul Ved Prakash says the goal is to make intelligence “abundant, not expensive.” Open-weight model usage on Together AI’s platform tripled over the past 12 months.
Cost and Control Drive Enterprise Adoption
Customers building on open models through Together AI report cost reductions of six to 20 times versus equivalent closed-model APIs. Decagon, an enterprise AI company and named Together AI customer, cut its inference costs sixfold after switching.
Current pricing for the flagship DeepSeek V3.1 model runs $0.60 per million input tokens and $1.70 per million output tokens. The cheapest hosted model, GPT-OSS 20B, starts at $0.05 per million input tokens.
Cost is not the only driver. According to a survey by MIT Sloan Management Review, the top reason enterprises in regulated industries adopt open-source models is data sovereignty, not cost. In a separate 2026 survey, 75% of enterprises said they plan to restrict external tools like ChatGPT over data leakage concerns.
Enterprise Lock-In Risk Growing
The IBM Institute for Business Value study puts a number on the problem. Seventy-one percent of respondents say switching their primary AI vendor or model would be difficult. Sixty-eight percent say meeting data residency and sovereignty requirements across geographies is challenging.
Ninety-one percent say they don’t fully understand their organization’s dependencies across AI vendors, models, and infrastructure. That lack of visibility makes switching harder and raises the stakes for the initial infrastructure choice.
EDB CEO Kevin Dallas put it plainly at the RAISE Summit: “Data is really a new currency; it’s the IP for many companies.”
A Crowded Race for Inference Infrastructure
Together AI is not alone in attracting capital. In June 2026, Groq raised $650 million to rebuild as an AI inference cloud. RunPod closed a $100 million round at a $1 billion valuation. Fireworks AI was in talks to raise at a $15 billion valuation, nearly four times higher than seven months prior. Baseten reported roughly a threefold jump in annualized revenue in a single quarter and was valued at up to $13 billion.
The pattern is consistent: the market is betting that as open-weight model quality converges with proprietary frontier models, value in the AI stack migrates toward whoever can serve those models most cheaply and reliably at scale. Together AI, with its $1.15 billion in annual bookings and locked-in compute capacity, is positioning itself to be that provider for large enterprise workloads.
Read more: Open-weight AI models drive shift to data control at RAISE Summit