Story perspectives

Nvidia launches DSX AI factory to cut data-center power loss

By Drooid · Sep 15, 2026, 11:01 PM UTC · How we work

Summary

NVIDIA Launches DSX

NVIDIA launched its DSX AI factory platform to improve data-center power efficiency.
The DSX design adds an 800 VDC architecture that reduces conversion loss.
Lambda saw a 24% token-throughput rise using DSX MaxLPS while keeping the same power budget.
Emerald AI’s Conductor cut power from four to three megawatts during a demand-response event.
NVIDIA plans a 96-megawatt Vera Rubin AI factory in Manassas, Virginia, as its first DSX Flex deployment.

Summary

NVIDIA Secures HBM Deal

Nvidia secured a multiyear memory partnership with SK hynix for next-generation HBM, tying $279.0 billion supply obligations.
Nvidia’s data-center segment generated $89.02 billion in Q2 FY2027, a 117% year-over-year increase.
Nvidia forecast Q3 FY2027 revenue of $108.0 billion, ±2%, with Vera Rubin contributing roughly 20%.
Top-five hyperscalers plan nearly $800 billion AI capex in 2026, rising to $1.3 trillion in 2027.
OpenAI commitments amount to roughly 12 gigawatts of Nvidia compute capacity.

1 of 2

NVIDIA Launches DSX

NVIDIA launched its DSX AI factory platform to improve data-center power efficiency.
The DSX design adds an 800 VDC architecture that reduces conversion loss.
Lambda saw a 24% token-throughput rise using DSX MaxLPS while keeping the same power budget.
Emerald AI’s Conductor cut power from four to three megawatts during a demand-response event.
NVIDIA plans a 96-megawatt Vera Rubin AI factory in Manassas, Virginia, as its first DSX Flex deployment.

1 / 2