Together AI raises $800M to grow its AI-optimized public cloud

2026-07-01 · Source: AI – SiliconANGLE · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure · Depth: Intermediate, quick

Summary

Together AI Inc., an operator of a cloud platform optimized for open-source artificial intelligence models, has secured \$800 million in Series C funding, led by Aramco Ventures, valuing the company at \$8.3 billion. The platform offers serverless inference services, claiming twice the performance of alternatives, alongside dedicated and cost-efficient Batch Inference options, which provide up to a 50% price reduction. Powering these services is the custom ATLAS engine, utilizing Nvidia chips and speculative decoding to accelerate workloads by up to 400% by adapting lightweight models. The platform also provides fine-tuning and training clusters with automatic technical issue detection. Together AI reported \$1.15 billion in annual bookings for Q2 and plans to expand its public cloud capacity by a factor of 50 over the next five years, enhancing both training and inference features.

Key takeaway

For Directors of AI/ML evaluating cloud infrastructure for open-source models, Together AI's \$800 million funding and performance claims warrant attention. Your teams could achieve up to 400% faster inference with their ATLAS engine and speculative decoding, potentially reducing operational costs by 50% using Batch Inference. Consider exploring their platform for scaling open-source AI deployments and enhancing training reliability, especially given their planned 50x capacity expansion.

Key insights

Together AI's platform uses advanced techniques like speculative decoding and ATLAS to optimize open-source AI model inference and training.

Principles

• Open-source AI models benefit from specialized cloud optimization.
• Adaptive lightweight models enhance speculative decoding accuracy.
• Automated issue detection improves AI training cluster reliability.

Method

Speculative decoding integrates a lightweight neural network to draft responses, which the main model then verifies and corrects, significantly speeding up output generation. ATLAS adapts the lightweight model.

In practice

• Use serverless inference for high-performance open-source AI.
• Employ Batch Inference for up to 50% cost reduction.
• Manage training clusters with Kubernetes or Slurm.

Topics

• AI Cloud Platform
• Open-source AI Models
• Serverless Inference
• Speculative Decoding
• ATLAS Engine
• AI Training Clusters
• Venture Capital Funding

Best for: CTO, VP of Engineering/Data, Entrepreneur, Tech Journalist, Investor, Director of AI/ML

Related on AIssential

• not much happened today — Rapid advancements in non-invasive brain-to-text, efficient AI inference, and sophisticated agent orchestration are reshaping AI development.
• How to Cut Inference Costs Without Slowing Down Your AI Stack* — Iteration-level scheduling can significantly reduce AI inference costs and improve throughput and reliability.
• Accelerating the next phase of AI — OpenAI's massive funding and infrastructure expansion aim to unify AI capabilities into a superapp, driving global adoption.
• AI inference startup Baseten reportedly raising $1.5B months after its last mega-round — Baseten's rapid funding highlights the intense investor interest in AI inference optimization and cost control.
• GV’s Dave Munichiello On Qualcomm’s Modular Purchase, The Firm’s 10x Return And The Shift In AI Software — AI software layers are gaining value, enabling efficient model execution across heterogeneous hardware amidst scarcity and the shift to disaggregated inference.
• How Superhuman and Databricks built a 200K QPS inference platform together — A collaborative engineering approach can achieve high-scale, low-latency AI inference by optimizing infrastructure and runtime.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI – SiliconANGLE.