Together AI (together.ai), a provider of the AI Native Cloud platform, has raised $800 million in Series C funding. The company delivers full-stack infrastructure for training, fine-tuning, and running inference on open-source models at scale, with research-driven optimizations that achieve 2x faster inference and up to 60% lower costs. The capital will accelerate the shift to open-source AI while securing commitments for over 500 MW of compute capacity.
Open-Source Inference Demand Surges
The timing aligns with a broader market pivot from proprietary models to open-source alternatives. Fireworks AI raised $250M Series C in October 2025, while Baseten is reportedly raising $1.5 billion. Together AI's research-backed approach — including FlashAttention-4 and custom kernels — directly targets the cost and control gaps that closed-model providers leave unaddressed.
Closed Model Costs Create Enterprise Pain
Enterprises face runaway inference expenses and vendor lock-in when relying on proprietary APIs. Inference now dominates lifetime AI system costs, with organizations reporting unpredictable pricing and limited ownership over their deployments. Current solutions force trade-offs between performance, cost, and data sovereignty.
Research Pipeline Drives Differentiation
Together AI built a full-stack platform spanning serverless inference, batch processing, dedicated clusters, fine-tuning, and evaluations. Its research team, including co-creators of FlashAttention, translates academic advances directly into production gains such as 31% higher throughput than competing open-source engines. Decagon achieved 6x cost reduction by migrating workloads to the platform.
"Proprietary inference has quietly created a tokenomics trap: unpredictable costs, black-box pricing, and a level of vendor dependency that would be unacceptable in any other layer of enterprise infrastructure."
Sovereign and Strategic Capital Validates Thesis
Aramco Ventures led the round alongside NVIDIA, Vista Equity Partners, General Catalyst, and Emergence Capital. The investor mix signals conviction in open-source infrastructure at global scale: Aramco brings energy and capacity expertise, NVIDIA reinforces hardware alignment, and Vista validates enterprise readiness with over $1.15 billion in bookings.
AI Infrastructure Market Expands Rapidly
The AI infrastructure market is projected to grow from $42.5 billion in 2025 to $532.8 billion by 2032 at 39.7% CAGR. Competitors like CoreWeave completed an IPO after raising over $14 billion, and Lambda Labs closed a $1.5 billion Series E. The structural shift toward inference workloads and open models is pulling capital into platforms that optimize both cost and control.
Research Pedigree Sets Apart Team
Co-founders include Stanford researchers Chris Ré, Percy Liang, and Tri Dao, inventor of FlashAttention. The team has published foundational work on Mamba architectures, RedPajama datasets, and kernel optimizations that power the company's performance claims.
Infrastructure Expansion Underway
With the new capital, Together AI is executing on 50x infrastructure scaling through partnerships including Hypertec for 36,000 NVIDIA GB200 GPUs and VC2 for 50+ metro inference data centers. The company has already served 400 trillion tokens in production.