NVIDIA CEO Jensen Huang Officially Launches DSX Product Line at GTC Taipei 2026
NVIDIA CEO Jensen Huang officially launched the DSX product line at GTC Taipei 2026, marking it as the third major product line following RTX and DGX, positioned as an end-to-end reference design and operational platform for AI factories.
DSX includes DSX SIM (Omniverse Digital Twin Planning), DSX OS (Automated Operations), DSX MAX LPS (Dynamic Power Optimization), and DSX Flex (Grid Interaction Regulation), utilizing 45°C high-temperature liquid cooling technology to significantly reduce energy consumption.
Huang stated that the cost of AI factories has risen to $50-60 billion, expected to reach $80-100 billion in the future, with 100GW of AI factories projected to be built globally by the end of this decade.
In market mechanisms, AI infrastructure construction is shifting from single device procurement to full-stack factory-level solutions, with funding accelerating towards end-to-end AI platforms. NVIDIA benefits from its transformation from a GPU supplier to an AI infrastructure company, while traditional data center builders face pressure due to gaps in system integration capabilities.
Source: Public Information
ABAB AI Insight
NVIDIA has previously accumulated system integration experience through the DGX and MGX series. The release of DSX continues its strategic transformation from a chip company to a complete AI factory solution provider. DSX SIM allows for full factory simulation before physical deployment, and DSX OS enables multi-tenant automated operations, focusing on addressing power over-provisioning and grid interaction issues.
In terms of capital pathways, NVIDIA is concentrating resources on a full-stack reference design and operational platform, motivated by customers no longer just purchasing GPUs but rather acquiring complete "AI Factory as a Service" capabilities. Dynamic power optimization and flexible grid regulation significantly enhance capital return rates, with emerging AI cloud companies like CoreWeave and Nebius rapidly deploying based on its full-stack solutions.
Similar cases include NVIDIA's early establishment of a developer ecosystem through CUDA and the current trend of cloud giants like Amazon and Microsoft building their own AI factories. NVIDIA is at a critical juncture in its comprehensive transformation from a hardware supplier to an AI infrastructure general contractor.
Essentially, this represents a restructuring of the industry chain: AI computing power is shifting from fragmented GPU procurement to integrated AI factory platforms. The mechanism is that the scale and cost of a single factory are exploding, enabling companies with end-to-end capabilities to gain higher pricing power and long-term service revenue through system optimization, thereby reshaping the construction model and capital allocation of global AI infrastructure.
ABAB News · Cognitive Law
Customers no longer buy equipment; they want to buy entire factories.
Every watt of performance equals revenue, every token is profitable.
Excellent companies sell AI factories, traditional companies sell GPUs.