AI Build Configurator v1.4 Adds AMD, Reprices GB300

This is a meaningful shift from a GPU price list toward a procurement decision tool, where rack architecture, memory, cooling, and software ecosystem matter alongside sticker price.

• –AMD coverage gives buyers a credible alternative to NVIDIA, including MI355X’s 288GB HBM3E and 8TB/s bandwidth. [AMD](https://www.amd.com/en/products/accelerators/instinct/mi350/mi355x.html)
• –Separating GB200 NVL72 from GB300 NVL72 avoids blending materially different rack generations and performance profiles. [NVIDIA GB200](https://developer.nvidia.com/blog/nvidia-gb200-nvl72-delivers-trillion-parameter-llm-training-and-real-time-inference/) [NVIDIA GB300](https://www.nvidia.com/en-us/data-center/gb300-nvl72/)
• –The 26% GB300 repricing highlights why AI infrastructure models need regularly refreshed assumptions rather than static hardware estimates.
• –Helios should remain a forward-looking scenario: AMD positions it as a 72-MI455X rack-scale platform, with initial deployments planned for the second half of 2026. [AMD](https://ir.amd.com/financial-information/sec-filings/content/0001193125-26-129106/0001193125-26-129106.pdf)
• –The real payoff is comparing total cost of ownership and cost per token across complete systems, not simply comparing accelerator purchase prices.