Together AI Completes $800M Series C Round, Valuation Rises 1.5x to $8.3B

AI cloud platform Together AI has announced the completion of an $800 million Series C round, bringing its post-money valuation to $8.3 billion. The round was led by Aramco Ventures, a subsidiary of Saudi Aramco, with participation from NVIDIA and others. Notably, just 16 months ago, its valuation was $3.3 billion—representing a 1.5-fold increase in one year. The company leverages an open-source model ecosystem to provide enterprises with cost-effective computing power, with annual bookings surging from $300 million at the start of 2024 to $1.15 billion—a 38-fold increase. The founding team, centered around Chinese alumni from Peking University, Zhang Ce, and Stanford professor Percy Liang, forms a premier Chinese technical powerhouse. NVIDIA continues to invest heavily in such emerging cloud providers to build a broader distribution network for computing power and transition from a purely hardware sales model to a revenue-sharing approach.

Author and source: WeChat official account "Dongsi Shitiao Capital"

Valuation increased by 1.5 times over the year.

The wealth-creation myth of the "water seller" in AI computing power continues.

Recently, AI cloud platform Together AI announced the completion of an $800 million Series C round, pushing its post-money valuation to $8.3 billion (approximately RMB 58.1 billion). The round was led by Aramco Ventures, a subsidiary of Saudi Aramco, with participation from prominent investors including NVIDIA, Vista Equity Partners, and General Catalyst.

Amid intense competition in large models, why has this AI cloud platform provider, founded just four years ago, attracted major players to rush in? The answer may lie in two key terms: “open-source models” and “a Chinese tech powerhouse.”

Together AI was founded in 2022, nearly simultaneously with ChatGPT. From its inception, the company clearly positioned itself as an “AI-native cloud” company, with its core business focused on renting out GPU clusters powered by NVIDIA to enterprises, enabling developers to run open-source large models at extremely low costs.

In the early stages of development, closed-source models led by OpenAI once dominated the market. However, businesses soon encountered a harsh reality: the high token costs of closed-source frontier models were relentlessly eroding the profit margins of AI applications. With the strong rise of high-performance open-source models like DeepSeek, the industry landscape is now undergoing a transformation.

During its transition from closed-source to open-source, Together AI directly addresses this pain point by integrating leading open-source models such as DeepSeek, MiniMax, and Kimi, offering enterprise customers a range of services including serverless inference, dedicated infrastructure, and batch inference, making it the biggest beneficiary of this surge in the open-source model ecosystem.

Leveraging a robust open-source model ecosystem, Together AI has developed highly cost-effective open-source model alternatives. Data shows that after integrating Together AI’s computing services, enterprises can reduce inference costs by 6 to 60 times, while achieving equal or even superior performance.

Taking its AI customer service platform, Decagon, as an example, when all workloads were migrated to Together AI, inference costs were reduced by sixfold. This significant cost advantage directly translated into tangible business results: the latest quarterly annual bookings show that Together AI’s order volume surged to $1.15 billion, up from just $30 million at the beginning of 2024—a 38-fold increase.

Annual bookings increased 38-fold in two years, fully unleashing Together AI’s commercial potential in the open-source AI space. First, with the explosion of AI agents, enterprise AI no longer requires simple Q&A bots but rather digital employees capable of managing complex workflows, generating massive volumes of token requests. Second, Together AI itself has broken through the physical bottlenecks of compute costs through innovations in underlying technology.

By leveraging its proprietary ATLAS inference engine and cutting-edge technologies such as speculative decoding, it has increased the speed of certain inference workloads by up to 400%. This end-to-end technical optimization enables Together AI to deliver enterprise-grade inference services at a fraction of the cost of traditional cloud providers, establishing a strong technological moat.

Precisely for this reason, Together AI's valuation has continued to rise. In February 2025, during its Series B round led by General Catalyst, Together AI was valued at just $3.3 billion. However, 16 months later, its post-money valuation reached $8.3 billion—equivalent to a 1.5-fold increase within a year.

Together AI's rapid rise, in addition to aligning with the industry trend of open-source AI, is also driven by a core technology team of Chinese origin.

A review of Together AI's official website reveals that its founding team is堪称 "a team of top scholars," with a relatively high proportion of Chinese members.

As one of the core representatives of this technical powerhouse, co-founder and CTO Ce Zhang graduated from Peking University. His background demonstrates a strong, top-tier academic foundation: in 2008, he earned his bachelor’s degree in mathematics from Peking University, followed by a Ph.D. from the University of Wisconsin-Madison, and he previously served as an assistant professor in the Department of Computer Science at ETH Zurich.

In his research, he has consistently focused on making machine learning cheaper, more trustworthy, and more accessible to a broader audience, believing that AI should not be an exclusive tool for tech giants but rather a technology that, through innovations at the system, scheduling, and distributed architecture levels, significantly reduces the computational costs of training and inference—enabling individuals, startups, and traditional industries to leverage open-source large models at low cost.

Based on this core philosophy, Zhang Ce led Together AI’s innovations in underlying compute scheduling and inference engines, guiding the team to develop the proprietary ATLAS inference engine and deeply integrate advanced distributed inference technologies such as NVIDIA Dynamo. These innovations enabled Together AI to achieve context-aware intelligent routing and KV cache management, effectively eliminating redundant computations in large-scale compute deployments, and ultimately securing outstanding commercial success with annual bookings exceeding $1.15 billion.

Second, Percy Liang, the company’s co-founder besides Zhang Ce, is also a leading Chinese scholar deeply entrenched in the field of AI. As a professor in the Department of Computer Science at Stanford University, Percy Liang serves as the director of Stanford’s Center for Research on Foundation Models (CRFM). In the fields of natural language processing and machine learning, he focuses on developing open-source foundation models and building model evaluation frameworks.

In terms of team collaboration, if Zhang Ce built an efficient, low-cost computing “highway” for Together AI at the infrastructure level, Percy Liang has served as the company’s “navigation system,” leveraging his deep academic background and industry insights. Through rigorous evaluation benchmarks, he helps Together AI continuously identify the optimal solutions within the complex open-source model ecosystem, ensuring the reliability and cutting-edge nature of the underlying technology.

Led by Zhang Ce and Percy Liang, together with the company’s chief scientist Tri Dao, they have built the foundational competitive edge of the company, collectively driving an exponential reduction in Together AI’s compute costs and strengthening its core algorithmic and system capabilities across multiple dimensions, enabling the company to maintain a leading position in algorithm iteration, model selection, system architecture, and compute resource scheduling.

Looking at NVIDIA’s move, it also reflects the giant’s strong desire to establish a presence in the AI computing ecosystem. As a co-investor, NVIDIA is not entering Together AI for the first time.

As early as November 2023, during the Series A round, NVIDIA participated in its early-stage financing. At that time, NVIDIA also invested as a follow-on investor, and in this round, the lead investors are Kleiner Perkins and Prosperity7, a subsidiary of Saudi Aramco.

NVIDIA’s continued investment in this company has a clear purpose. To date, Together AI, which NVIDIA has identified as a key partner, is no longer just a simple AI cloud service provider—it has become an indispensable “ecosystem node” within NVIDIA’s vast computing empire. In simple terms, through this investment, NVIDIA has directly secured massive hardware orders.

Disclosures reveal that Together AI’s business model is built on NVIDIA’s H100, H200, and even the latest Blackwell GPUs. After achieving annual bookings exceeding $1.15 billion, Together AI will inevitably require even greater demand for massive GPU clusters. NVIDIA’s investment conveniently paves the way for sustained consumption of its own chips.

In addition, NVIDIA is also seeking to break the monopoly of traditional cloud giants by partnering with these “neocloud” service providers to build a broader distributed computing network. For NVIDIA, this is equivalent to bundling the medium- and long-tail AI development and enterprise demands through partners like Together AI, thereby transforming them into more stable and predictable procurement curves.

Of course, the deeper reason is that NVIDIA is transitioning from a role of simply selling hardware to becoming a revenue-sharing partner in AI infrastructure. On July 1, NVIDIA officially launched the AI Compute Partnership, clearly positioning itself as a collaborative partner in building the industry ecosystem. This means NVIDIA will no longer just sell hardware—it will also provide credit support and revenue-sharing mechanisms, enabling companies like Together AI to rapidly scale their computing capacity without bearing massive upfront capital expenditures.

Under this model and new positioning, NVIDIA has partnered with other AI cloud service providers besides Together AI, such as Sharon AI and Firmus, to jointly build "AI factories" based on the NVIDIA DGX AI Factory architecture.

The latest quarterly earnings report showed NVIDIA delivering an outstanding performance: quarterly revenue reached $81.615 billion, a year-over-year increase of 85%; net profit attributable to shareholders surged 211% year-over-year to $58.321 billion.

This data not only broke NVIDIA's quarterly revenue record but also propelled NVIDIA to a $5 trillion market capitalization, fully demonstrating the strong global demand for AI computing infrastructure. A closer look at this earnings report reveals that customer diversification was one of the key drivers behind NVIDIA's significant revenue growth.

Financial reports show that demand growth in NVIDIA’s data center business comes not only from hyperscale cloud providers but also widely from AI clouds, industrial, enterprise, and sovereign AI customers. This diversified customer base reduces NVIDIA’s reliance on any single major client and builds a more resilient growth foundation. Meanwhile, as global large model iterations accelerate and AI applications rapidly deploy, demand for expanding AI factories and data centers worldwide has intensified, driving strong procurement needs for computing power from downstream cloud providers, AI-native companies, and government and enterprise clients—further strengthening expectations for NVIDIA’s upward performance trajectory.

Under NVIDIA’s open strategy, Together AI has also announced plans to expand its public cloud capacity—i.e., computing power—by 50 times over the next five years. This means that when the cost of computing power is no longer an industry bottleneck, the era of open-source large models may truly arrive, and for AI developers, it will be a genuine “Together AI.”

This AI unicorn, led by Chinese technological expertise, will also continue to advance, riding the wave of global AI infrastructure development.