IBM and Together AI Forge a $240M Powerhouse Partnership to Revolutionize Open-Source AI Inference
In the rapidly evolving landscape of artificial intelligence, the bridge between experimental models and production-grade reality is often built on the strength of raw infrastructure. Taking a massive leap forward, IBM and Together AI have officially inked a multi-year agreement valued at $240 million. This partnership isn't just about capital; it’s about deploying the first large-scale inference cluster on IBM Cloud powered by NVIDIA’s cutting-edge HGX B300 systems. Designed to meet the skyrocketing demands of enterprises, this collaboration aims to make high-performance open-source AI inference more accessible, efficient, and scalable than ever before.
The $240 Million Blueprint for Scalable AI
At the heart of this agreement is a strategic deployment expected to go live in the first quarter of 2027. IBM is positioning itself as a critical player in the AI infrastructure space by hosting a massive cluster of NVIDIA HGX B300 systems. Together AI will leverage this high-octane hardware to provide open-source model inference at a scale that was previously difficult to achieve. This marks the first dedicated, large-scale cluster of its kind on IBM Cloud, utilizing NVIDIA Spectrum-X Ethernet networking to ensure that data moves as fast as the processors can think.
According to NVIDIA, this specific configuration is built to deliver up to 30 times more "AI factory" output compared to previous generations. For enterprises, this means the ability to run massive workloads with significantly lower latency and higher throughput, effectively turning complex AI computations into a streamlined utility.
Why Together AI is Betting Big on IBM Cloud
Together AI has been on a meteoric rise, recently closing an $800 million Series C funding round that brought its valuation to a staggering $8.3 billion. The company’s philosophy is rooted in the belief that open-source models are the backbone of the future. By building on modular, open stacks, developers can escape the constraints of proprietary "closed" models. Currently, Together AI is already a heavyweight in the space, serving an incredible 400 trillion tokens monthly across its inference, training, and fine-tuning platforms.
Selecting IBM and NVIDIA was a calculated move. Together AI needed a partner capable of delivering GPU capacity at an enterprise pace while maintaining the lowest possible cost per token. IBM’s long-standing expertise in enterprise-grade cloud environments provides the reliability and security that large corporations demand when moving their AI projects from the lab to the real world.
Leadership Insights: Moving Toward Agentic AI
Vipul Ved Prakash, the CEO of Together AI, highlighted that enterprises are hungry for the performance of top-tier frontier models without the heavy price tags associated with closed systems. He noted that the partnership with IBM and NVIDIA provides the reliable foundation needed to bring production-grade inference to more companies, making open-source the obvious choice for the modern enterprise.
Less busywork, more real work.
We build robust internal tools and scalable SaaS platforms so your team can stop drowning in spreadsheets and start focusing on growth.
Alan Peacock, General Manager of IBM Cloud, echoed this sentiment, pointing out that businesses are currently in a race to adopt "agentic AI"—AI systems capable of independent action to drive business outcomes. He believes that providing scalable and economical infrastructure is the only way to help innovators like Together AI accelerate the next generation of digital intelligence.
The Concept of the 'AI Factory'
NVIDIA’s Dion Harris described the current shift in technology as the rise of the "AI factory." Just as electricity and telecommunications became the essential infrastructure of the 20th century, accelerated computing is becoming the essential utility of the 21st. By integrating HGX B300 systems with Spectrum-X networking, IBM and Together AI are essentially building a power plant for intelligence, turning raw data into actionable insights in real-time.
This collaboration is part of a much larger, ongoing synergy between IBM and NVIDIA. Beyond just hardware, the two giants are working on GPU-native data analytics, unstructured data extraction, and consulting services designed to help organizations operationalize AI. For developers and global enterprises, this partnership signals a future where the power of open-source AI is no longer limited by hardware constraints, but rather empowered by them.