A desktop that costs as much as a luxury car is not a typical product launch, but the ASUS ExpertCenter Pro ET900N G3 is not a typical machine.
ASUS has positioned this tower workstation as a desktop-level AI supercomputer, and the hardware inside justifies that framing. The centerpiece is NVIDIA's GB300 Grace Blackwell Ultra platform, which combines a CPU with 72 Arm Neoverse V2 cores and a GPU built on the latest Blackwell Ultra architecture. The CPU side brings 496GB of LPDDR5X memory with a bandwidth of 396GB/s. The GPU adds 252GB of HBM3e video memory running at 7.1TB/s bandwidth. Combined, the total unified memory reaches 748GB, which is the number ASUS leads with for good reason. Running large-scale AI models locally requires memory headroom that most workstations simply cannot provide, and 748GB puts this machine in a different category entirely.
The chassis itself keeps a traditional tower form factor, sized close to a standard high-end workstation. It sits next to a desk rather than requiring a server rack, which matters for the professional and enterprise buyers ASUS is targeting. The entire build is designed around sustained full-load operation. A 1600W power supply feeds the system, and ASUS claims the cooling solution keeps both the processor and GPU running at full speed under prolonged heavy workloads without thermal throttling.
Expansion options are solid. Three PCIe 5.0 slots are available, with one capable of hosting an additional NVIDIA discrete graphics card. Up to four M.2 SSDs can be installed, giving developers room to work with large datasets without storage becoming a bottleneck. The machine ships with Ubuntu pre-installed, which signals clearly that the target user is running AI development, scientific computing, or enterprise inference workloads rather than consumer tasks.
The interface selection is more conservative than the rest of the spec sheet. Eight USB 3.2 ports, one USB 2.0 port, and two 10Gbps Ethernet ports cover the basics. There is no USB4 or Thunderbolt support, which may feel like an oversight at this price point but likely reflects the workstation-first priorities of the design.
On pricing, the ASUS ExpertCenter Pro ET900N G3 is listed at $99,999 on US channels, which works out to roughly 720,000 yuan (approximately ₹83,00,000 at current exchange rates, expect higher after import duties and taxes in India). That number will filter the audience down to AI researchers, enterprise teams, and labs running local model training at scale.
For anyone in that segment, the alternative is typically cloud compute billed by the hour or a rack-mounted server solution that requires dedicated infrastructure. A machine this capable sitting next to a desk, running on standard power, changes the calculus on what local AI development looks like in 2026. It is really fun to deploy our own I at local machine so you don't have to sent data to any company and you can share whatever you want.
