# Enterprise AI inference gets a dedicated operator: Rackspace to run Together AI on GB300 racks from 2027 - Published: 2026-11-11 - Author: [CTO Sync](https://ctosync.com/author/ctosync-com.md) - Topics: [News](https://ctosync.com/news.md) - About: [Rackspace](https://www.rackspace.com/), [Together AI](https://www.together.ai/) - Based on: [Rackspace and Together AI Announce Strategic Agreement to Support Enterprise AI Inference](https://www.globenewswire.com/news-release/2026/11/10/3399517/0/en/rackspace-and-together-ai-announce-strategic-agreement-to-support-enterprise-ai-inference.html), GlobeNewswire, 2026-11-10 - Based on: [NVIDIA GB300 NVL72](https://www.nvidia.com/en-us/data-center/gb300-nvl72/), NVIDIA - Based on: [The 2025 AI Index Report](https://hai.stanford.edu/ai-index/2025-ai-index-report), Stanford HAI - Based on: [Rackspace](https://www.rackspace.com/), Rackspace - Based on: [Together AI](https://www.together.ai/), Together AI - Source: https://ctosync.com/news/rackspace-together-ai-inference-dedicated-gb300-racks/ - Publication: CTO Sync - Publisher: Connectively (Terkel, Inc.) - Editorial policy: https://ctosync.com/editorial-policy/ - AI use: Connectively's automated system wrote this article with AI from a press release issued by Rackspace on November 10, 2026, linked under Sources. Statements about the company and its products are the company's. The illustration was generated with AI. The article was assembled and published automatically, without an editor's review. Report errors to corrections@connectively.us. Rackspace and Together AI signed a multiyear agreement on November 10, 2026, under which Rackspace plans to run dedicated NVIDIA GB300 NVL72 racks for Together AI inference from 2027. For enterprise teams, the Rackspace Together AI inference deal points to open-model inference served from hardware a named operator runs end to end. ## Key points - Rackspace (NASDAQ: RXT) says it will provide and operate dedicated NVIDIA GB300 NVL72 systems, with power, liquid cooling, networking and storage, to carry Together AI inference workloads. - Deployment is planned for 2027, subject to conditions in the definitive agreement, including Rackspace arranging financing for the GPU systems. - The two companies also signed a non-binding letter of intent to explore a joint go-to-market offer for enterprises that need more control, compliance and data sovereignty. - Effective November 6, 2026, Rackspace agreed a $66 million private placement of common stock with an investor group led by Voss Capital, LP, with proceeds earmarked in part for GPU-related initiatives. - Each GB300 NVL72 rack combines 72 Blackwell Ultra GPUs and 36 Grace CPUs in one liquid-cooled system, according to [NVIDIA's product page](https://www.nvidia.com/en-us/data-center/gb300-nvl72/). ## Why does dedicated hardware for AI inference matter now? Because inference has become cheap enough per query that running it in production, at volume, is now the cost and reliability problem enterprises have to plan for. The [Stanford HAI 2025 AI Index Report](https://hai.stanford.edu/ai-index/2025-ai-index-report) found that the inference cost for a system performing at the level of GPT-3.5 fell more than 280-fold between November 2022 and October 2024, while hardware costs declined by 30% a year and energy efficiency improved by 40% a year. Falling unit prices push more workloads into production, and production brings the questions a CTO answers before signing off: who runs the hardware, where the data sits and who is accountable when it goes down. Rackspace's approach speaks to that last one directly. ## What will Rackspace and Together AI each run? Rackspace runs the physical layer, and Together AI runs the cloud platform on top of it. Rackspace, an NVIDIA Cloud Partner, says it will house the GB300 NVL72 systems in a purpose-built, liquid-cooled facility connected with NVIDIA Spectrum-X Ethernet. Together AI says it will supply the managed cloud platform on that hardware, including access to open source models, inference optimizations and developer tools. The NVIDIA hardware is built for this kind of work. [NVIDIA describes the GB300 NVL72](https://www.nvidia.com/en-us/data-center/gb300-nvl72/) as a rack-scale system purpose-built for test-time scaling inference and AI reasoning, with 1.5x more dense FP4 Tensor Core FLOPS than Blackwell GPUs. Rackspace CEO Gajen Kandiah said the agreement applies the company's experience running enterprise environments to the dedicated hardware behind Together AI's inference platform. Rackspace says it wants to make this a repeatable model for AI platform companies, systems integrators and enterprises, by securing contractual commitments for dedicated capacity and matching its financing to expected customer demand and the useful life of the hardware. ## What could the joint enterprise AI offer include? The letter of intent targets enterprises that need more control, compliance and data sovereignty, with AI running closer to their own data. The companies say they will evaluate how Together AI's inference, post-training and reinforcement learning could combine with Rackspace's infrastructure, platforms and engineering services, and Rackspace says its work across private cloud, hyperscaler and edge environments could support configurations that bring Together AI's cloud nearer to enterprise workloads. Data residency, security and compliance features would be set by each offering and customer arrangement. Together AI founder and CEO Vipul Ved Prakash said scaling in the enterprise needs infrastructure that can grow with the company while delivering the performance and reliability its customers expect. ## What does the Rackspace Together AI inference deal mean for CTOs? Our analysis: it adds an option between renting shared GPU capacity from a hyperscaler and buying racks outright. A team that wants open models served on dedicated hardware, with one operator accountable for power, cooling and uptime and an inference platform on top, can watch how this pairing is packaged as the 2027 deployment and the joint offer take shape. The sovereignty angle is the part to track for regulated workloads. As the joint offer takes shape, the questions to put to both vendors are where the data is processed, which compliance controls ship as standard, and which come only with a custom arrangement. For other moves to put AI compute under an operator's control, see [Dorado Joins NVIDIA DSX AI Factory Ecosystem](https://ctosync.com/news/dorado-joins-nvidia-dsx-ai-factory-ecosystem/) and the sovereign AI platform [Sultan Meghji announced for institutions](https://ctosync.com/news/sultan-meghji-announces-kundi-3-a-self-improving-sovereign-ai-platform-that-institutions-o/). ## Frequently asked questions ### When will the GB300 NVL72 deployment go live? Rackspace says deployment is planned for 2027, subject to the conditions set out in the definitive agreement. ### What will Rackspace spend the private placement proceeds on? Rackspace says the net proceeds go to general corporate purposes, working capital, capital expenditures and growth capital, including GPU-related initiatives. It plans to finance the GB300 NVL72 systems themselves separately and says it is in active discussions with lenders. ### How large is Together AI's developer base? Together AI says it serves more than two million developers. It was founded in 2022. ### Who led Rackspace's $66 million private placement? An investor group led by Voss Capital, LP, under a share purchase agreement effective November 6, 2026. ## Sources - [Rackspace and Together AI Announce Strategic Agreement to Support Enterprise AI Inference](https://www.globenewswire.com/news-release/2026/11/10/3399517/0/en/rackspace-and-together-ai-announce-strategic-agreement-to-support-enterprise-ai-inference.html): press release from Rackspace and Together AI via GlobeNewswire, 10 November 2026. - [NVIDIA GB300 NVL72](https://www.nvidia.com/en-us/data-center/gb300-nvl72/): NVIDIA's product page and specifications. - [The 2025 AI Index Report](https://hai.stanford.edu/ai-index/2025-ai-index-report): Stanford Institute for Human-Centered Artificial Intelligence, 2025. - [Rackspace](https://www.rackspace.com/): the company's website. - [Together AI](https://www.together.ai/): the company's website.