Tinybox Drops Nvidia Options as Tiny Corp Raises AMD Build Price

What happened

Tiny Corp has removed every Nvidia-based option from its tinybox range of configurable deep learning workstations, and has raised the price of its new 192GB, six-GPU AMD build by $3,000, according to TechRadar. The tinybox pro has also been taken off sale.

In place of the Nvidia configurations, the company has added two builds based on Radeon AI Pro R9700 GPUs. The changes came less than two weeks after the workstation range launched with a headline price of $11,000.

Buyers who want Nvidia hardware in a tinybox can no longer order it, so the AMD builds are now the only route into the range for that class of machine.

Why it matters

For organisations buying on-premise AI compute, a workstation is a commitment to a specific hardware ecosystem. When a vendor withdraws an entire GPU family within weeks of launch and reprices its flagship configuration, the decision shifts from a simple specification choice to a question of supply stability and total cost. This is a reminder that AI infrastructure is still a fast-moving market, where product catalogues can change faster than internal procurement cycles.

The story also shows how quickly the accelerator landscape is diversifying. AMD's Radeon AI Pro R9700 is now the basis for two of the range's configurations, giving buyers an alternative to Nvidia for local deep learning work. Teams evaluating AI hardware will need to test software compatibility and performance on each platform rather than assume a single default.

By the numbers

• $11,000 headline price reported at the workstation's launch
• $3,000 price increase applied to the new 192GB, six-GPU AMD build
• Six GPUs in the 192GB build at the centre of the price change
• Two new Radeon AI Pro R9700 configurations added to the range
• Less than two weeks between the launch and the revised lineup

The Renascence take

Most coverage will treat this as a pricing story about a niche workstation. The more useful reading is that a configurator is a promise, and buyers plan around what it shows them. Removing an entire GPU family while the product is still new sends a signal that the choices on screen may not be the choices available tomorrow.

Customers configure a machine on the assumption that the options they pick will still be there when they check out. When a vendor changes that catalogue and its prices in the same fortnight, the experience breaks at the exact point where trust is built. A customer-obsessed operator in AI hardware should version its configurations, timestamp prices, and tell existing quote-holders what has changed and why. Treat the lineup as a service commitment, not a marketing page.