Together AI Raises $800 Million at an $8.3 Billion Valuation as Open-Source Inference Demand Surges

Together AI Raises $800 Million at an $8.3 Billion Valuation as Open-Source Inference Demand Surges

The Vault — AI Edition · Business · July 24, 2026

Sixteen months ago, the AI infrastructure startup Together AI was worth $3.3 billion. On July 1, the San Francisco company announced it had raised an $800 million Series C at an $8.3 billion post-money valuation — roughly a 2.5x leap in about a year and a half, and a wager that the cheapest way to run artificial intelligence is increasingly the open-source way.

The round was led by Aramco Ventures, the venture arm of Saudi Arabia’s state oil company, with participation from Vista Equity Partners, General Catalyst, Emergence Capital, NVIDIA, March Capital, Pegatron and SentinelOne’s S Ventures. It ranks among the largest financings yet for a so-called neocloud — the fast-growing crop of specialized providers that rent out NVIDIA GPU clusters and inference software to companies that would rather not build their own data centers or pay premium prices for closed frontier models.

Intelligence as infrastructure

Together AI’s pitch is blunt: frontier-model pricing can eat a startup’s entire margin, and open-weight models now do much of the same work for a fraction of the cost. The company operates cloud infrastructure through which enterprises can train and run hundreds of open models — DeepSeek, Nemotron, MiniMax and Kimi among them — via an OpenAI-compatible interface, backed by an in-house research lab focused on inference efficiency.

“Intelligence is becoming a foundational resource for the modern economy, every bit as essential as electricity, bandwidth or capital,” said Vipul Ved Prakash, co-founder and CEO of Together AI. “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 is a repeat founder who sold his previous startup, social-media search platform Topsy, to Apple in 2013 for a reported $200 million-plus. He co-founded Together AI in 2022 with Stanford professor Percy Liang and ETH Zürich and University of Chicago associate professor Ce Zhang.

The economics driving the boom

The numbers behind the raise explain the investor enthusiasm. Together AI says its annual bookings crossed $1.15 billion last quarter, and that it now serves thousands of paying customers, including AI heavyweights Cursor, Cognition and Decagon. Customers report cost savings of 6x to 60x versus closed-model pricing for equal or better performance; the company points to Decagon, which it says cut inference costs sixfold after moving to Together.

That is happening as demand for open models accelerates industry-wide. Usage of open-source models tripled over the past twelve months, according to research from AI gateway OpenRouter that Together cited, while McKinsey has found nearly three-quarters of organizations expect to increase their use of open-source AI. The financing, Together said, will fund new products and a dramatic expansion of capacity — an infrastructure footprint the company expects to grow roughly 50-fold over the next five years.

Investors framed the shift as structural rather than a passing preference. “The shift toward open source isn’t a niche preference anymore, it’s becoming the default for any company that wants to scale AI without losing its margin,” said Joe Floyd, General Partner at Emergence Capital. “Together has the team, the infrastructure, and the customer base to lead that shift, and the growth speaks for itself.”

For lead investor Aramco Ventures, the appeal is the sheer scale of the buildout. “Building AI infrastructure over the next decade will be the biggest infrastructure project in human history,” said Abhishek Shukla, Managing Director of Prosperity7 Ventures US, part of Aramco Ventures. “Together has built the platform that makes open source models genuinely usable at enterprise scale.”

A crowded, cash-soaked neocloud race

Together AI’s raise lands in the most heavily capitalized moment the neocloud sector has seen. The largest player, CoreWeave, went public in 2025 and now commands a market value in the tens of billions. Inference-focused rivals such as Fireworks AI and Baseten are chasing the same enterprise developers with similar promises of speed and lower cost, while newer entrants keep drawing checks: Upscale AI added an extension last month that pushed its total funding to $500 million at a $2 billion valuation, and AMD-focused TensorWave raised a $350 million Series B at a $1.55 billion valuation.

Together’s differentiation is its bet that model choice, not just raw GPUs, is the moat. By pairing a large open-model catalog with proprietary inference optimization, it positions itself less as a commodity GPU landlord and more as the default runtime for companies that want frontier-class results without frontier-class bills — a segment where efficiency compounds directly into gross margin.

What to watch

The obvious question is whether Together can convert $1.15 billion in bookings and a 50x expansion plan into durable economics as GPU supply loosens and rivals slash prices. A neocloud’s margins hinge on utilization and power, and even backers acknowledge the tie to energy: Schneider Electric, an investor through its SE Ventures fund, cast the deal as part of a convergence between AI and energy infrastructure.

Also worth tracking is the Saudi angle. Aramco Ventures leading the round deepens Gulf sovereign capital’s footprint in Western AI infrastructure at a moment of intensifying scrutiny over who owns the compute layer. And with The Information having reported in the spring that Together was seeking $1 billion at a $7.5 billion valuation, the final terms — less money, higher price — suggest the company negotiated from strength. Whether open-source inference remains the cheaper default as closed-model providers cut their own prices will determine if $8.3 billion looks conservative or frothy a year from now.