- Together AI raised an $800 million Series C at an $8.3 billion valuation, more than doubling the $3.3 billion mark it set just 16 months earlier.
- Aramco Ventures led the round, joined by NVIDIA, Vista Equity Partners, General Catalyst, Salesforce Ventures, and Emergence Capital.
- Enterprises running open-weight models on Together report inference costs 6x to 20x lower than closed frontier LLMs at equal or better quality.
- Annual bookings crossed $1.15 billion last quarter as industry-wide use of open-source models tripled in twelve months.
The most valuable question in enterprise AI right now is not which model is smartest. It is which model a company can actually afford to run millions of times a day. Together AI just raised $800 million betting that the answer is increasingly an open-source one, and some of the biggest names in tech and energy wrote the checks to prove it.
The $800M Bet on Open-Source Economics
On July 1, 2026, Together AI announced an $800 million Series C at an $8.3 billion post-money valuation. The round was led by Aramco Ventures, with participation from NVIDIA, Vista Equity Partners, General Catalyst, Emergence Capital, Salesforce Ventures, March Capital, Pegatron, SentinelOne’s S Ventures, Lux Capital, and others. Founded in 2022, the company rents out NVIDIA GPU clusters and specialized AI infrastructure — a category investors now call the “neocloud.”
A valuation that more than doubled in 16 months
The leap is steep. Together AI last raised a $305 million Series B at a $3.3 billion valuation in early 2025, and started life with a $102.5 million Series A led by Kleiner Perkins in 2023. Notably, earlier reporting suggested the company had been seeking $1 billion at a $7.5 billion valuation — meaning it ultimately took less cash but secured a higher price, a rare signal of leverage in a market where many startups are quietly accepting flat or down rounds.
Alongside the equity, Together secured commitments for more than 500 megawatts of compute capacity to be financed independently by its investors — a structure that offloads the enormous capital cost of data-center buildout onto backers while letting the company scale its footprint aggressively.
Business Insight — Aramco Ventures leading an AI infrastructure round is the story within the story. Oil-and-gas capital is chasing compute the way it once chased refining capacity, treating GPU clusters as the next long-duration, high-demand asset class. When energy money underwrites AI data centers, the line between the power grid and the model layer effectively disappears.
Why Closed-Model Pricing Breaks at Scale
Together AI’s pitch rests on a simple economic observation: the cost structure of closed frontier models that looks manageable in a prototype often becomes unsustainable in production. As companies move from generating a handful of demo responses to deploying agents that write code, resolve support tickets, and process documents around the clock, inference bills compound faster than budgets. The result, in the company’s words, is that firms are forced to “ration intelligence” precisely when demand for it is accelerating.
Open weights have closed the quality gap
The argument only works because open models have caught up. Open-weight families such as DeepSeek, Nemotron, MiniMax, Kimi, and GLM now rival proprietary frontier models on quality while giving developers the freedom to fine-tune and customize. Companies building on these models routinely report 6x to 20x lower costs at equal or better performance. Customer-support automation firm Decagon, for example, cut its inference costs sixfold after moving to Together.
Crucially, Together frames the economics as a full-stack problem, not a model problem. Cost is determined by kernels, compilers, inference systems, and hardware utilization together — which is why the company invests heavily in research like FlashAttention-4 for NVIDIA Blackwell and its own Megakernel and together.compile optimizations. Its customer roster now includes fast-growing AI companies like Cursor, Cognition, Decagon, ElevenLabs, and Suno.
Business Insight — The competitive frontier is shifting from “who has the best model” to “who runs it cheapest per token.” For enterprise buyers, that reframes the vendor decision: the model is becoming a commodity input, and the durable moat is inference efficiency. Any CTO forecasting AI spend should be modeling cost-per-token trajectories, not just benchmark scores.
The Neocloud Land Grab
Together AI is not raising in a vacuum. Neoclouds — specialized providers that rent AI-optimized GPU capacity rather than general-purpose cloud — have become one of the hottest categories for venture investment. In the month before Together’s raise, Upscale AI added a $190 million Series A extension that pushed its total funding to $500 million at a $2 billion valuation, and TensorWave, which builds AMD-powered GPU clusters, closed a $350 million Series B at a $1.55 billion valuation.
A bet on abundance over scarcity
The thesis binding these deals together is that AI inference is becoming core production infrastructure — the electricity of the software economy — and that whoever supplies it most efficiently captures durable demand. Together’s stated ambition is to make “intelligence abundant, not expensive,” and its bookings trajectory suggests the market is validating that framing: annual bookings crossed $1.15 billion last quarter, and the company now ranks among the largest producers of AI tokens in the world.
For enterprises, the practical takeaway is optionality. A viable, well-capitalized open-model supply chain gives buyers real negotiating leverage against closed-model providers and a hedge against per-token price increases. That is a structural change in the AI procurement landscape, not a temporary discount war.
Business Insight — The heavy concentration of capital in neoclouds carries its own risk: a wave of overlapping GPU-cluster bets could create overcapacity if AI demand growth slows even modestly. Together’s investor-funded 500MW structure is smart hedging, but the sector as a whole is now making a very large, correlated wager on inference demand rising indefinitely.
Related
- The AI Giant That Just Leapfrogged OpenAI
- Spotify’s Founder Just Raised $700M to Scan Your Body
- Why Investors Bet $180M on Governed AI Communications
- The Coding Model That Uses 4x Fewer Tokens
- The Invisible Bottleneck Holding Back Every AI Agent
Sources
- TechCrunch — Neocloud Together AI raises $800M, leaps to $8.3B valuation
- Together AI — Announcing our $800M Series C to accelerate the shift to open-source AI
- Business Wire — Together AI Raises $800 Million at $8.3 Billion Valuation
AI Biz Insider · AI Business EN · aibizinsider.com
댓글 남기기