Together AI's $3.3B Bet: Why VCs Are Pouring Hundreds of Millions Into Open Source AI
When a startup raises $305 million at a $3.3 billion valuation, you pay attention. When that startup is building infrastructure for open-source AI—and NVIDIA is writing checks—you pay very close attention.
Together AI just closed their Series B, led by General Catalyst and co-led by Prosperity7. The round included Salesforce Ventures, NVIDIA, Kleiner Perkins, Lux Capital, and Emergence Capital. This isn't just a fundraise. It's a signal.
The Thesis: Open Source Wins
Together AI's bet is simple: the future of AI is open source.
While OpenAI and Anthropic race to build proprietary models, Together AI is building the cloud infrastructure that lets everyone else compete. They offer training, fine-tuning, and inference for open-source models—Llama, Mistral, Qwen, and hundreds more.
"We believe the future of AI is open source. We contribute leading open-source research, models, and datasets to advance the frontier of AI."
— Together AI
The Founders
Together AI was founded in June 2022 by four heavyweights who bridge academia and industry:
- Vipul Ved Prakash (CEO) — Former Sr. Director at Apple, serial entrepreneur
- Ce Zhang — Professor at ETH Zurich, ML systems researcher
- Chris Ré — Stanford Professor, MacArthur Fellow, creator of Snorkel
- Percy Liang — Stanford Professor, founder of the HELM benchmark
This is a team with deep credibility. They've published foundational research, built systems at scale, and now they're commercializing that expertise.
The Numbers
| Valuation | $3.3 billion |
| Series B | $305 million |
| Total Funding | $400M+ |
| Customers | 50,000+ companies |
| Founded | June 2022 |
Who's Using It
The customer list reads like a who's-who of tech:
- Cursor — The AI code editor everyone's talking about
- Zoom — Video conferencing giant
- The Washington Post — Major media
- Quora — Q&A platform
These aren't experiments. These are production deployments at scale.
The Product Stack
Together AI offers a full-stack platform:
- Inference — Run 200+ open-source models with industry-leading latency
- Fine-Tuning — Train custom models on your data
- Training — Full pre-training capabilities for frontier models
- GPU Clusters — Dedicated compute for enterprise workloads
The pitch to enterprises: 2x latency reduction and ~33% cost savings compared to alternatives.
Why NVIDIA Invested
NVIDIA's participation is strategic. As the GPU kingpin, they need the open-source ecosystem to thrive. More open-source AI development means more GPU demand. Together AI is building the infrastructure that makes that happen.
It's a bet that the AI market isn't winner-take-all. It's a bet that thousands of companies will train and deploy their own models—and they'll need infrastructure to do it.
The Bigger Picture
2024 was the year open-source AI caught up. Meta's Llama models closed the gap with GPT-4. Mistral emerged from France as a serious contender. Alibaba's Qwen models are being adopted across Silicon Valley.
The moat for closed-model providers is eroding. And Together AI is betting that when the dam breaks, they'll be the ones selling shovels.
What This Means
For developers: Open-source AI is no longer a compromise. The infrastructure exists to train, fine-tune, and deploy models that compete with the best proprietary options.
For enterprises: You don't have to be locked into OpenAI or Anthropic. The tools exist to own your AI stack.
For the industry: The $3.3B valuation is a market signal. VCs believe open source wins.
Together AI's platform is available at together.ai.