Skip to content
News Funding & M&A

Together AI raises $800 million Series C at $8.3 billion valuation to scale open-model inference

· by Pondero Newsdesk

The short version

Together AI closed an $800 million Series C on July 1, 2026, pushing its valuation to $8.3 billion and securing commitments for over 500 megawatts of compute capacity to serve enterprises running open-source models.

Together AI raises $800 million Series C at $8.3 billion valuation to scale open-model inference

Open-model inference startup Together AI closed an $800 million Series C on July 1, 2026, at an $8.3 billion post-money valuation. That figure is 2.5x its $3.3 billion Series B valuation from roughly 16 months earlier, per TechCrunch, and signals that enterprise demand for cheaper, open-weight AI inference is outpacing the pace of most AI funding rounds this year.

What happened

Aramco Ventures led the round, with Nvidia, Vista Equity Partners, General Catalyst, Emergence Capital, Salesforce Ventures, March Capital, Pegatron, Lux Capital, SentinelOne's S Ventures, and others participating, per Together AI's announcement.

The company reported annual bookings crossed $1.15 billion in its last quarter per TechCrunch, serving thousands of paying customers. Named customers include Cursor, Cognition, Decagon, Eleven Labs, and Suno. Investors committed over 500 megawatts of compute capacity as part of the deal, which Together AI says supports roughly a 50-fold capacity expansion over five years, per its blog post.

The company runs serverless inference on open-weight models including DeepSeek, Llama 4, Qwen, and Kimi. Its value proposition is cost: per the Together AI blog, enterprises using open-weight models through its platform achieve 6x to 20x lower inference costs compared to closed frontier model APIs, though that is a vendor self-claim rather than a third-party benchmark.

Why it matters

The revenue numbers set this round apart from typical AI infrastructure bets. A $1.15 billion annualized bookings figure, posted before a major new round closes, shows real enterprise contracts rather than speculative capacity reservations. That is meaningful for any AI tool operator deciding whether to run open or closed models at scale: the neocloud market is maturing fast enough that operators like Together AI now have pricing power and supply commitments that closed API providers do not.

Cursor appearing as a named customer is the most concrete signal for developer-tool operators. Cursor runs AI inference across millions of code completions. The fact that it routes at least some of that load through Together AI suggests the latency and reliability profile of open-model neoclouds has crossed a threshold where product companies feel comfortable depending on them. Cognition (Devin's parent) and Decagon (AI customer-service) round out the named list, all high-inference, cost-sensitive workloads.

Aramco Ventures leading the round adds a strategic dimension beyond pure VC returns. Saudi Aramco has its own AI infrastructure build-out ambitions, and a 500MW compute commitment from investors suggests at least some of that capacity will be built close to Aramco's operations.

Context

Together AI's prior rounds were a $102.5 million Series A in 2023 (led by Kleiner Perkins) and a $305 million Series B at $3.3 billion in February 2025. The valuation trajectory from $3.3 billion to $8.3 billion in 16 months reflects the broader neocloud funding wave: Upscale AI raised $500 million and TensorWave raised $350 million in June 2026 alone, per TechCrunch.

Together AI also published a benchmark on July 1 claiming its inference stack delivers 31% more tokens per second than the next-fastest open-source engine for production coding agent workloads, per its blog. That claim is self-reported and has not been independently validated.

What to watch next

The 500MW compute commitment will take years to deploy. Near-term, watch whether Cursor's changelog credits Together AI explicitly as an inference provider, which would confirm depth of integration. Also watch for Together AI to name which open models are seeing the most enterprise adoption: that data would help operators decide which open-weight models are genuinely production-grade at scale, not just benchmark leaders.

Sources