# Nvidia Gatekeeper Model Nvidia controls who can install their high-end GPU systems. For B200, B300, and GB300 platforms, you cannot simply buy the hardware and rack it yourself. Installation must be done by an Nvidia-authorized partner. **Authorized partners include:** Asus, Super Micro, Dell. The list is short and controlled. **The process:** - Authorized partner sends engineers to your site - Installation requires ~6 months of Nvidia-specific training - If you're not on the approved list, Nvidia will blanket refuse — no exceptions, no workarounds - This applies to the physical installation, not just procurement **What this means in practice:** Your choice of GPU dictates your choice of installer, your timeline, and a significant portion of your cost structure. You can't hire your own team to do it. You can't train your own people (unless you go through Nvidia's program). This is structural lock-in, not just commercial preference. **Strategic implication:** This gatekeeper model is hardening, not softening. As Nvidia's GPU platforms grow more complex (liquid cooling, high-power density, custom interconnects), the barrier to independent installation increases. Plan your deployment timeline around partner availability, not your own readiness. **The alternative:** Consider silicon that doesn't impose gatekeeper restrictions. Tenstorrent, for example, can be installed by any qualified team with no Nvidia-style certification requirement. --- See also: [[Tenstorrent — Alternative Silicon Path]] | [[Modular Data Centers MoC]]