Tech • IA • Crypto
NVIDIA DSX Powers Gigawatt‑Scale AI Factories at Maximum Efficiency
TL;DR
Nvidia is promoting DSX, an end-to-end blueprint for AI factories designed to maximize efficiency, revenue, and grid integration as global AI infrastructure rapidly expands.
KEY POINTS
AI factories as a new industrial race
The global push to build AI infrastructure is accelerating, with projections of 100 gigawatts of AI factory capacity coming online before the end of the decade. These facilities are highly complex systems where chips, racks, networking, power, cooling, and grid connectivity must be co-designed. Efficiency directly translates into revenue, making optimization critical.
DSX as a full-stack blueprint
Nvidia DSX is positioned as a reference architecture for building and operating AI factories. It spans the full lifecycle, from design and simulation to deployment and ongoing operations, aiming to standardize how large-scale AI compute facilities are constructed and managed.
Digital twin simulation with DSX Sim
The process begins with DSX Sim, built on the Omniverse platform, which allows partners to design and validate facilities before physical deployment. Operators can simulate layouts, power consumption, cooling systems, and networking, testing integrations and changes within a digital twin environment prior to installation.
Operational control through DSX OS
Once deployed, DSX OS provisions and manages infrastructure, handling monitoring and remediation. It converts installed hardware into multi-tenant, resilient, AI-ready capacity, enabling continuous operation at scale.
Reducing overprovisioned power
Current AI facilities often overprovision power by up to 40%. DSX Max LPS addresses this inefficiency by enabling more GPUs to operate within the same power envelope, potentially unlocking billions in additional annual revenue.
Advanced cooling innovations
A key feature is hot liquid cooling at 45°C, which reduces both water and energy consumption. This approach shifts more available power toward revenue-generating compute rather than facility overhead.
Dynamic power optimization
DSX introduces dynamic power allocation, redistributing energy across racks based on workload demand. It also includes in-rack power smoothing, which mitigates spikes and stabilizes electrical loads, improving reliability and utilization.
AI-driven facility management
Teams of AI agents coordinate power and cooling in real time, working alongside DSX systems to continuously balance workloads and infrastructure constraints, enhancing efficiency across the facility.
Integration with energy grids
Through DSX Flex, AI factories can respond to real-time grid conditions, adjusting power usage when needed. This positions AI infrastructure as a flexible energy asset that can support grid stability rather than strain it.
CONCLUSION
Nvidia’s DSX framework reflects a shift toward tightly integrated, energy-aware AI infrastructure, aiming to make large-scale compute both more profitable and more sustainable as global demand surges.
Full transcript
The world is racing to build AI factories, the largest infrastructure build out in human history. AI factories are incredibly complex, every layer, chip, rack, network, power, cooling, grid must be designed together from end to end because compute is revenues. Nvidia DSX is the blueprint, a reference design for building and operating AI factories at maximum efficiency and profitability. [music] It starts with DSX Sim. With the DSX Sim Omniverse blueprint, [music] partners design and validate an Nvidia Vera Rubin AI factory before a single rack lands. They plan the layout. Simulate the power and cooling. Design the network. Validate every integration. Test every change in the digital twin. The factory powers on. DSX OS takes over and provisions, operates, monitors, and remediates the infrastructure. Turning the installed systems into trusted, multi-tenant, resilient, AI-ready capacity. Today's AI [music] factories over provision power by up to 40%. DSX Max LPS lets operators safely [music] deploy more GPUs inside the same power budget. Adding billions in annual revenue. Breakthrough [music] hot liquid cooling at 45°C uses less water and energy. More power going to [music] revenue-generating compute. Incredible. Dynamic power allocation steers power from [music] rack to rack, recovering stranded watts, sending them where work is happening. In-rack [music] power smoothing flattens peak current spikes and power surges. Throughout the factory, [music] teams of AI agents work with DSX Max LPS, continuously coordinating to [music] balance cooling and power to meet workload demand. DSX AI factories are flexible energy assets that operate cooperatively [music] with the grid. DSX Flex reads real-time grid signals and dynamically [music] adjusts factory power when the grid needs relief. 100 gigawatts of AI factories [music] will come online before the end of the decade. Nvidia DSX AI factories run at highest [music] efficiency, produce the lowest cost tokens, and make the grid stronger.