Together AI announced an $800 million Series C on July 1, 2026 at an $8.3 billion post-money valuation, more than doubling the San Francisco company's prior mark. The round was led by Aramco Ventures, with Vista Equity Partners, General Catalyst, Emergence Capital, Nvidia, March Capital, Pegatron and S Ventures (SentinelOne) also participating.
Together AI's business is training and serving open-source models — DeepSeek, Nemotron, MiniMax, Kimi and others — at a fraction of the cost of closed frontier systems, often with comparable or better task-specific performance. The company disclosed that annual bookings crossed $1.15 billion last quarter, driven by open-source model usage tripling across the industry over the past twelve months as enterprises look to cut inference costs without sacrificing capability. Customers now include Cursor, Cognition and Decagon.
The backdrop is a broader shift in enterprise AI economics. As frontier labs like OpenAI and Anthropic push premium pricing on flagship models, a growing share of production AI workloads is migrating to open-weight models that can be fine-tuned and hosted more cheaply — and neoclouds like Together AI, Baseten, Groq and Modal are the primary beneficiaries. Nvidia's direct investment is notable: rather than just selling GPUs into the buildout, it is now taking equity stakes in the software layer that determines how efficiently those GPUs get utilized.
Comparable deals put the round in context. CoreWeave went public in 2025 and now trades at a multiple many times its last private valuation; Lambda and Crusoe have both raised nine-figure rounds this year at valuations well below Together AI's new mark, reflecting Together's differentiated open-model specialization versus pure GPU-rental competitors. The $8.3B valuation puts Together AI in the same tier as Anthropic's early growth-stage marks, on a fraction of the revenue — a signal of how aggressively investors are pricing infrastructure picks-and-shovels exposure to the AI buildout.
For founders and growth investors, the implication is that the open-versus-closed-model divide is now a fundable, durable axis of the market rather than a temporary cost-arbitrage play. Portfolio companies building anywhere in the inference or fine-tuning stack should expect open-weight adoption to keep accelerating as enterprises optimize AI spend, and LPs evaluating infrastructure exposure now have a real public-adjacent comp (CoreWeave) plus a fast-growing private one (Together AI) to underwrite valuations against.
What to watch: whether Together AI's roughly 50x planned capacity expansion over five years gets financed through additional equity or debt, how quickly open-source model quality closes the gap with frontier closed models, and whether Aramco Ventures leading a US AI-infra round draws additional sovereign-capital interest into the sector.