Together AI Raises $800 Million at an $8.3 Billion Valuation to Scale Open-Model Infrastructure

Together AI, the four-year-old startup betting that the future of artificial intelligence belongs to open-source models running on rented GPU clusters, said Tuesday it had raised $800 million in a Series C round that values the company at $8.3 billion. The financing more than doubles the valuation the company commanded just 16 months ago and cements its status as one of the most richly capitalized players in the fast-growing category of AI-specialized cloud providers known as neoclouds.

The round was led by Aramco Ventures, the venture arm of the Saudi oil giant, with participation from Vista Equity Partners, General Catalyst, Emergence Capital, Nvidia, March Capital, Pegatron, and SentinelOne's S Ventures, among others. Together AI last raised a $305 million Series B at a $3.3 billion valuation in early 2025. The company has now raised roughly $1.33 billion in total since its founding in 2022.

The Bet on Open Models

Together AI rents access to Nvidia GPU clusters and sells the software layer that lets developers train, fine-tune, and run open-weight models such as those in the Llama, DeepSeek, and Qwen families. The pitch is straightforward: capable open-source models now rival closed systems from the likes of OpenAI and Anthropic on many tasks, and they can be run for a fraction of the cost.

That thesis is showing up in the numbers. The company said annual bookings crossed $1.15 billion as of its most recent quarter, up sharply as enterprises shift workloads away from premium closed-model APIs. Together AI now serves thousands of paying customers, including coding-tool maker Cursor, AI software firm Cognition, and customer-service startup Decagon.

"Intelligence is becoming a foundational resource for the modern economy, every bit as essential as electricity, bandwidth or capital," said Vipul Ved Prakash, Together AI's co-founder and chief executive. "Our mission is to ensure that intelligence is abundant, not expensive. The future of AI won't be owned by a few companies. It will be built by millions of developers and businesses, and open-source models are making that possible."

Prakash, who founded the company alongside Stanford researchers Chris Re, Percy Liang, Ce Zhang, and Tri Dao, framed the moment as a familiar one in technology history. "History shows that the biggest technology shifts are won by open ecosystems that make innovation cheaper, faster and available to everyone," he said. "We believe AI will follow the same path."

Scaling the Iron

The capital is aimed squarely at compute. Together AI said it has secured commitments for more than 500 megawatts of compute capacity, much of it to be capitalized independently by investors rather than sitting directly on Together's balance sheet, a financing structure that has become common among neoclouds racing to deploy expensive Nvidia hardware. The company expects its infrastructure footprint to grow roughly 50-fold over the next five years.

Those ambitions are already translating into deals. In early June, Rumble Inc. signed a multi-year agreement worth approximately $270 million with Together AI to deploy dedicated Nvidia HGX B300 GPU cloud capacity, a transaction disclosed in a securities filing. Nvidia's presence on the cap table underscores how tightly the neocloud model is bound to the chipmaker's supply, and how motivated Nvidia is to seed customers who will absorb its latest Blackwell-generation silicon.

Why It Matters

The neocloud category exists because the economics of AI infrastructure have outrun what most enterprises, and even many hyperscalers, want to build themselves. Training and serving large models requires enormous, tightly networked GPU clusters that are costly to buy, hard to source, and quick to depreciate. Neoclouds like Together AI, CoreWeave, and Lambda specialize in assembling that iron and renting it out, absorbing the capital risk in exchange for recurring usage revenue.

Together AI's wager is more specific than pure infrastructure rental. It is betting that open-weight models will become the default substrate for enterprise AI, and that a company controlling both the hardware and the optimization software sitting on top of it can capture margin at both layers. If open models continue to close the quality gap with closed systems, the addressable market is enormous, because price, not capability, becomes the deciding factor for most business workloads.

The risks are equally clear. Neocloud economics depend on keeping expensive GPUs utilized, and a glut of capacity or a slowdown in demand could squeeze margins fast. The heavy reliance on Nvidia hardware and on off-balance-sheet financing vehicles introduces leverage that looks fine in a boom and dangerous in a downturn. And the participation of a Saudi sovereign-linked investor as lead will draw scrutiny given the geopolitical sensitivities around who funds and ultimately influences frontier AI infrastructure.

What to Watch

The near-term question is whether Together AI can convert its $1.15 billion in bookings into durable, profitable revenue as it scales capacity 50-fold. Watch for how quickly the 500 megawatts of committed compute comes online, whether large customers like Cursor and Cognition deepen their commitments, and whether the open-model quality gap keeps narrowing. If it does, Together AI's $8.3 billion valuation may look conservative. If closed models pull decisively ahead again, or if GPU supply loosens and pricing collapses, the neocloud boom could prove far more cyclical than its backers hope.

---

Sources: Business Wire, TechCrunch, The New York Times, Together AI blog