TD Synnex, a major GPU distributor, secured over 1,000 Nvidia GPUs for resale through Nebius Cloud, demonstrating capacity scaling.
TD Synnex, the Fremont, Calif.-based distribution giant that recently won Nvidia's Distribution Partner of the Year award for the Americas, has reserved over 1,000 high-end Nvidia GPUs in the cloud to help solution providers overcome supply constraints. The company announced a partnership with neocloud provider Nebius on Tuesday at its High-Growth Conference in Puerto Rico, committing to AI computing clusters based on Nvidia's B300 GPU. TD Synnex is accepting purchase orders for B300-based instances through Nebius now, with availability expected to begin in July. The agreement lasts for a year and gives TD Synnex the option for renewal.
According to TD Synnex, this engagement "marks the first time a global IT distributor has reserved a dedicated Nvidia AI factory-grade cluster from an AI-native cloud provider." Francisco Criado, senior vice president of cloud, security and AI at TD Synnex, explained that the model differs from typical distribution practice: "It's not your typical distribution go-to-market where you're buying and reselling something. We're actually committing and we've purchased this reserved instance within the Nebius data center ... and we will now work with our partners and their customers to [drive consumption]."
The partnership addresses a critical market challenge: high-end GPUs have become a scarce resource due to high demand from AI-native companies. Criado stated plainly, "There's literally no capacity available right now." Jack Donato, senior director AI and neocloud vendor solutions at TD Synnex, emphasized how the deal prioritizes channel partners who might otherwise struggle to secure capacity: "Our ability to give them that prioritization up there with the general hyperscalers and neocloud providers just gives them access, in general. Could they get the same capacity direct? Maybe, but there's no availability." Donato noted that the Nebius partnership helps channel partners gain access to what has become a scarce resource, ensuring strategic partners receive prioritization they would not otherwise obtain.
Nvidia Americas channel chief Craig Weinstein validated the importance of the partnership, telling CRN: "It means that we have an [Nvidia Cloud Partner] that's going to offer cloud capability to the ecosystem to ensure [that] enterprises can iterate and solution providers have a platform to leverage." Based on Nvidia's Blackwell Ultra architecture, the B300 debuted last year as the company's flagship GPU for x86-based servers, connecting eight accelerator chips through the vendor's NVLink interconnect. TD Synnex said Nebius' B300 instances are meant for a wide range of AI use cases, ranging from training to inference and fine-tuning.
The initial commitment has already generated substantial interest among channel partners. Criado indicated that the initial batch of instances "will probably be purchased pretty quickly," with the company already planning expansion. "We're committing to this very early for July, and as soon as this capacity is sold, we're going to be looking at [bringing more capacity through] the next data center buildout," he said.