Together AI is valued at more than $8 billion following a recent funding round, according to company reports and industry filings. The startup provides a cloud platform that enables developers to train, fine-tune, and deploy open-source large language models (LLMs) with higher efficiency and lower costs than proprietary alternatives.
How did Together AI reach an $8 billion valuation?
Together AI’s valuation surged after it secured $224 million in Series C funding led by Nvidia, according to reports from TechCrunch. This latest round follows a Series B that valued the company at roughly $1.25 billion, marking a nearly seven-fold increase in perceived market value within a short period. Investors are betting on the company’s ability to decouple AI development from the closed ecosystems of giants like OpenAI and Google.
The valuation reflects a broader market shift toward “open-weights” models. As enterprises seek to avoid vendor lock-in and maintain control over their data, Together AI’s infrastructure allows them to run models like Meta’s Llama 3 or Mistral on dedicated GPU clusters without relying on a single proprietary API.
What services does Together AI provide?
Together AI operates a full-stack AI cloud that focuses on three primary pillars:
- Together Inference: A serverless API that hosts popular open-source models, optimized for speed and low latency.
- Together GPU Clusters: Dedicated clusters of H100 GPUs that allow companies to train massive models from scratch or perform deep fine-tuning.
- Custom Model Development: The company develops its own high-performance models, such as the RedPajama project, which aims to create a fully open-source alternative to the most powerful LLMs.
Unlike traditional cloud providers, Together AI uses a proprietary software stack to optimize how GPUs communicate, which they claim reduces the compute power needed to train large models, according to Together AI’s official documentation.
Who are the key investors backing the company?
The investor roster for Together AI includes some of the most influential players in the semiconductor and software industries. Nvidia’s lead role in the Series C is particularly significant, as it signals a strategic partnership between the hardware provider and a platform that maximizes that hardware’s utility. Other participants in funding rounds have included Salesforce Ventures and various venture capital firms specializing in deep tech.
How does Together AI compare to proprietary AI platforms?
The primary tension in the AI market is between “closed” models (like GPT-4) and “open” models (like Llama). Together AI positions itself as the infrastructure layer for the open movement.
| Feature | Proprietary (e.g., OpenAI) | Together AI (Open Source) |
|---|---|---|
| Data Control | Data sent to provider’s API | Option for private, dedicated clusters |
| Model Flexibility | Fixed versions provided by vendor | Ability to fine-tune and modify weights |
| Pricing | Token-based usage fees | Mix of serverless and reserved GPU pricing |
While proprietary models often lead in raw reasoning capabilities, Together AI argues that open-source models are catching up quickly. By providing the tools to optimize these models, they’ve made it feasible for a mid-sized company to run a high-performing model that’s tailored to its specific industry data.
What challenges does the company face?
Despite the valuation jump, Together AI operates in a crowded space. It competes directly with other GPU cloud providers like CoreWeave and Lambda Labs, as well as the “Hyperscalers” like Amazon Web Services (AWS) and Microsoft Azure. These larger entities have deeper pockets and existing enterprise contracts.
The company’s growth also depends heavily on the continued viability of open-source AI. If a proprietary model achieves a breakthrough that open-source alternatives cannot replicate, the demand for Together AI’s specialized infrastructure could diminish. However, the current trend suggests that enterprises prefer the transparency and ownership associated with open-weights models.