Together AI Snags $800M as Valuation Rockets to $8.3B

Who gets paid if companies keep shifting AI workloads from closed frontier APIs to open source models running on rented GPUs?

Why does "Together AI raises $800M" matter beyond one startup’s cap table?

Together AI raises $800M at an $8.3 billion valuation, a sharp jump from the $3.3 billion valuation attached to its last round about 16 months ago, according to TechCrunch.

The company, founded in 2022, is an AI neocloud provider. In plain terms, it rents access to Nvidia GPU clusters and other AI-specific infrastructure to developers and companies running models, with a strong focus on open source deployment and inference.

The new Series C was led by Aramco Ventures. Participants included Vista Equity Partners, General Catalyst, Emergence Capital, Nvidia, March Capital, Pegatron, SentinelOne’s S Ventures, and others.

That investor list matters. Together AI isn’t being funded as a narrow developer tool. It’s being priced as a core AI infrastructure company, one that could sit between model builders, enterprise users, chip suppliers, and the growing universe of companies trying to avoid paying premium rates for every token sent through closed systems.

Together AI says annual bookings topped $1.15 billion as of its last quarter.

The speed of the repricing is the headline. Together AI last raised a $305 million Series B at a $3.3 billion valuation in early 2025. Before that, it raised a $102.5 million Series A in 2023, led by Kleiner Perkins, with participation from Nvidia and Emergence Capital.

There were signs this deal was coming. TechCrunch notes that The Information reported in March that Together AI was seeking $1 billion at a $7.5 billion valuation. If those figures were accurate, Together AI took less capital than floated in that report, but landed a higher valuation.

That’s a better headline for the company. It also raises the bar.

Why are investors still paying up for open source AI hosting?

Together AI sits in the part of AI infrastructure that has become easy to understand and hard to execute: give developers reliable access to GPUs, host the models they actually want to run, and reduce the friction around inference.

The pitch is different from the broad cloud playbook. Neoclouds are not trying to be everything to every enterprise workload. They target AI builders that need compute, model serving, fine-tuning, and performance tuning without buying their own clusters or routing all usage through a closed model provider.

TechCrunch reports that companies are increasingly using open source models through neocloud providers like Together AI rather than paying premiums on tokens for closed frontier models across all AI usage. Together AI also says open source model usage across the industry has tripled in the past year, pointing to research from OpenRouter, an AI gateway company.

The business opportunity is clear from the customer names Together AI chose to disclose. It says it has thousands of paying customers, including Cursor, Cognition, and Decagon. Those are exactly the kinds of AI-native customers investors want infrastructure companies to land: high usage, technical teams, and strong incentives to optimize model cost and latency.

Together AI’s earlier Series B announcement also showed how deeply the company has tied its strategy to open source models and Nvidia hardware. The company said then that its AI Acceleration Cloud supported over 200 open source models across chat, image, audio, vision, code, and embeddings, and that it had secured 200 MW of power capacity while deploying NVIDIA Blackwell GPUs across multiple North American data centers.

That older announcement, published by Together AI, also said the company planned a partnership with Hypertec to co-build a cluster of 36,000 NVIDIA GB200 NVL72 GPUs.

The strategic pattern is tight:

• Compute: secure GPU capacity before customers need it.
• Models: make open source models easier to run in production.
• Inference: compete on speed, cost, and deployment experience.
• Customers: win AI-native teams before they standardize elsewhere.

That explains why Together AI raises $800M at a valuation that would have looked aggressive a year earlier. The market being financed here is not just cloud hosting. It’s the operating layer for open source AI workloads.

Can an $8.3 billion valuation survive the economics of GPU clouds?

The harder question is whether Together AI can turn heavy compute demand into durable, high-margin revenue.

AI cloud infrastructure is capital hungry. GPUs are expensive. Power capacity matters. Reliability is non-negotiable. Pricing pressure can arrive fast when customers compare inference providers, shift models, or bring workloads back to larger cloud contracts.

Together AI’s valuation implies that investors believe the company can avoid becoming a pass-through reseller of scarce hardware. The company has to prove that its software layer, model support, inference engine, developer experience, and customer relationships create value beyond raw GPU rental.

That’s where the open source angle becomes more than branding. If customers are choosing open source models because they want control, lower costs, or customization, Together AI can try to attach itself to the full workflow: inference, training, fine-tuning, model selection, code execution, and synthetic data generation.

Its Series B announcement described an AI Acceleration Cloud spanning inference, training and fine-tuning, agentic workflows with built-in code interpretation, and synthetic data generation. If that platform vision holds, the company has more ways to expand accounts than simply selling another block of GPU time.

But the pressure is real. The same TechCrunch report names recent neocloud financings, including Upscale AI, which raised a Series A plus in an extension totaling $500 million at a $2 billion valuation last month, and TensorWave, which focuses on AMD GPU clusters and raised a $350 million Series B at a $1.55 billion valuation last month.

Together AI has more capital now, but rivals are also being funded to chase the same budgets.

Which question won’t be answered until the next spending cycle?

The next test is not whether companies want cheaper or more flexible AI infrastructure. The source material already shows demand: Together AI claims over $1.15 billion in annual bookings and thousands of paying customers.

The open question is whether those bookings mature into sticky platform revenue.

Watch for four signals from here: whether Together AI expands capacity without margin strain, whether it lands larger enterprise customers, whether it deepens ties around major open source models, and whether its developer tools make workloads harder to move.

Together AI raises $800M because investors see open source AI infrastructure as a high-value control point. The months ahead will show whether that control point belongs to Together AI, or whether the neocloud boom gets squeezed between hardware costs, rival specialists, and the largest cloud providers.

The Bottom Line

• Together AI’s $8.3 billion valuation signals strong investor demand for AI infrastructure beyond closed frontier model APIs.
• Its reported $1.15 billion in annual bookings suggests enterprise AI workloads are rapidly moving onto rented GPU infrastructure.
• Backers including Nvidia and Aramco Ventures position the company as a strategic player in the open source AI deployment stack.