IBM and Together AI have signed a $240 million multi-year agreement to scale open-source artificial intelligence inference using NVIDIA-powered infrastructure, Reuters reported on August 11, 2026. The deal represents one of the largest dedicated commitments to enterprise open-source AI inference to date and signals growing corporate demand for alternatives to closed, proprietary model ecosystems.
Under the agreement, IBM will provide the cloud and data-center infrastructure while Together AI contributes its open-source model optimization and serving platform. The partnership aims to make it cheaper and faster for enterprises to run large open-weight models in production. For readers following the rapid evolution of the AI infrastructure landscape, AI Buzz Wire offers continuous coverage of enterprise AI deals and model deployments.
Scaling Open-Source AI Inference for Enterprises
The arrangement centers on a dedicated NVIDIA-powered AI inference cluster, according to IBM's newsroom announcement. Together AI, a startup specializing in high-performance inference for open-source models, will use the compute capacity to serve enterprise customers that want the flexibility of open-weight models without managing their own GPU fleets.
IBM, meanwhile, is positioning itself as the infrastructure backbone. The company has been steadily expanding its hybrid cloud and AI offerings, and the Together AI partnership extends that strategy into the open-source inference tier. By combining IBM's global data-center footprint with Together AI's model-serving expertise, the two companies are betting that a meaningful share of enterprise AI workloads will eventually run on open-weight models rather than closed APIs.
That bet is not without support. Meta's release of its Muse Glimmer model, Alibaba's continued development of the Qwen family, and Nvidia's own trillion-parameter Nemotron 4 project have all reinforced the view that open-source AI is becoming a viable production-grade alternative to systems from OpenAI and Anthropic.
Why Open-Weight Inference Is Attracting Capital
The $240 million price tag reflects the economics of AI inference at scale. Running large language models in production requires sustained access to expensive GPUs, and enterprises increasingly want control over which models they deploy and how their data is handled. Open-weight models address both concerns: they can be hosted on infrastructure the customer controls, and they eliminate the per-token API fees charged by closed-model providers.
Together AI has built its business around this premise. The company optimizes open-source models for faster, cheaper inference and offers them through an API and a managed platform. The IBM deal gives Together AI a long-term home for the compute capacity it needs to meet enterprise demand, while IBM gains a differentiated AI offering for its cloud customers.
Industry analysts have noted that the inference market is where the bulk of AI spending ultimately lands. Training frontier models commands headlines and enormous capital expenditure, but the recurring revenue from serving those models to millions of users is what sustains the business. Securing dedicated NVIDIA infrastructure for open-source inference is a direct play for that recurring revenue stream.
NVIDIA Remains the Indispensable Layer
The deal also underscores NVIDIA's continued dominance of the AI compute stack. Although IBM and Together AI are partners on the infrastructure and software layers, the underlying accelerators are NVIDIA's. The agreement explicitly names NVIDIA AI infrastructure as the foundation, and BNN Bloomberg reported that the cluster will be powered by NVIDIA GPUs.
NVIDIA's position has only strengthened as open-source models proliferate. Every additional open-weight model deployed in production generates demand for NVIDIA inference hardware, regardless of which company built the model or which cloud provider hosts it. The IBM-Together AI deal is the latest evidence that competition among model developers and cloud providers ultimately funnels toward the same silicon supplier.
What the Deal Means for the Competitive Landscape
For IBM, the partnership is part of a broader effort to remain relevant in the generative AI era. While the company has developed its own Granite series of open-source models, it has also leaned heavily on alliances. The Together AI deal complements IBM's existing AI partnerships and gives enterprise customers a clear path to open-weight inference on IBM-managed infrastructure.
For Together AI, the agreement provides the scale and credibility that come with a Fortune 500 infrastructure partner. A $240 million multi-year commitment from IBM signals to the market that open-source AI inference is mature enough for mission-critical enterprise use, not just experimentation by developers and startups.
The deal arrives amid a wave of AI infrastructure investment. Amazon has hiked its 2026 capital expenditure guidance to $220 billion, citing an AWS surge driven by AI demand. Volta, an AI cloud startup, recently emerged from stealth with a $10 billion Anthropic deal. And Anthropic itself is seeking Google's financial backing for data-center leases ahead of a widely anticipated initial public offering.
A Defining Moment for Open-Source AI in the Enterprise
The IBM-Together AI agreement may prove significant beyond its dollar value. It demonstrates that open-source AI inference has reached the point where a major enterprise technology company is willing to commit hundreds of millions of dollars to it on a multi-year basis. If the economics hold, similar deals are likely to follow as cloud providers race to offer open-weight inference as a standard service.
For enterprises, the practical implication is more choice. A dedicated IBM-backed platform for open-source AI inference means organizations can weigh the trade-offs between closed APIs and open-weight deployments on a more level playing field, with a partner that can guarantee uptime, security, and compliance at scale.
Stay Ahead of AI
The AI infrastructure market is moving fast, with billion-dollar deals reshaping the competitive map every week. Bookmark AI Buzz Wire for daily reporting on the partnerships, funding rounds, and model releases driving the industry forward.
Read more AI news →
