Nvidia Strengthens Its “AI Alliance Front” From PCs to DC Infrastructure
IT DAILY ·
✦ AI Summary
Nvidia pushed ahead with joint hardware and software design with Microsoft to turn Windows PCs into a platform for running AI agents.
Nvidia announced 'Lenovo AI Express' with Lenovo based on pre-validated infrastructure, and the small configuration ships in as little as 15 days.
Vertiv's 2.3MW CoolChip cooling distribution unit (CDU) met the NVIDIA DSX Ready requirements and was officially recognized.
Nvidia is deepening collaboration across the board with global technology companies. Its partnerships now span everything from personal PC environments to enterprise AI factories and data center physical infrastructure, expanding the Nvidia-centered AI ecosystem into a structure that encompasses software platforms as well as power and cooling infrastructure.
Through its partnership with Microsoft, Nvidia is building a PC environment that runs OS-level AI agents. The photo shows Microsoft CEO Satya Nadella and Nvidia CEO Jensen Huang.
Through its partnership with Lenovo, Nvidia unveiled a turn-key solution based on pre-validated server infrastructure. The solution can be delivered in as little as 15 days.
Vertiv's 2.3MW cooling distribution unit (CDU) meets the requirements of Nvidia's next-generation AI factory specification. As a result, Nvidia is broadening the scope of its collaborations with global companies across PC software environments, enterprise server supply, and data center power and cooling equipment.
Nvidia and Microsoft are pushing ahead with joint hardware and software design aimed at turning the Windows PC ecosystem into a platform for running AI agents, and they jointly announced related plans at the 'Windows AI and Surface' event held on the 8th in San Francisco, California.
At the event, the two companies announced measures to ensure stable and secure operation of local AI agents on desktops and laptops. Microsoft said it would officially provide the 'Microsoft Execution Container (MXC)'.
The 'Microsoft Execution Container (MXC)' is a foundational technology that supports the secure, continuous background operation of agents under OS control. Nvidia unveiled the 'RTX Spark (RTX Spark)'.
The 'RTX Spark (RTX Spark)' combines a Blackwell RTX GPU and a Grace CPU, and comes with up to 128GB of unified memory and 1 petaflop of performance. The product will be installed in PCs from major manufacturers, including Microsoft Surface Laptop Ultra, and supports directly running large-scale models in a local environment without usage limits or cloud data transfer.
Nvidia unveiled the 'NVIDIA DGX Station for Windows' for enterprise developers. The device is based on the GB300 Grace Blackwell Ultra Desktop Superchip and delivers up to 20 petaflops of performance. It can also process models with up to 1 trillion parameters locally.
The unveiling was presented as an effort to ease the burden of having to move back and forth between Linux and Windows environments during high-performance AI development. It also laid the groundwork for building always-on AI agents while maintaining a normal work environment.
Lenovo and Nvidia then announced 'Lenovo AI Express,' which is aimed at solving the long lead times and complexity that have been cited as problems in enterprise AI infrastructure deployment. The announcement was made on the 8th.
'Lenovo AI Express' is offered as a combination of pre-validated infrastructure and operational services. It comes in three versions — small, medium and large — so companies can choose based on their workloads and budgets.
The small configuration includes two Nvidia RTX 6000 PRO Blackwell Server Edition GPUs. This setup supports dozens of users and models ranging from 7 billion to 70 billion parameters. Shipments begin in as little as 15 days.
Vertiv announced the performance and delivery schedule for each AI infrastructure configuration on the 8th. The medium configuration includes 8 GPUs of the same class and supports agentic AI for hundreds of users, with shipment possible in as little as 20 days. The large configuration is intended for models with up to 1 trillion parameters, is based on Nvidia HGX B300, and is delivered in as little as 25 days.
The two companies will also provide PoC support, ROI validation support, and GPU resource optimization support. In addition, they will support the entire process from adoption to operations.
The two companies have also strengthened their support services. 'Premier Support Plus for Servers' offers proactive support features and predictive support features.
These services are intended to reduce the risk of infrastructure failures. They are also aimed at significantly shortening the time needed to transition to actual operations.
Vertiv also has its own 2.3MW CoolChip cooling distribution unit (CDU) model. The model meets the 'NVIDIA DSX Ready' requirements.
The NVIDIA DSX AI factory platform is an architecture that integrates the design and operation of computing, networking, power, cooling, and facilities as a whole. As a result, individual pieces of equipment also need to be verified against these requirements.
Vertiv's 2.3MW CDU passed Nvidia's own qualification assessment tool and was officially recognized as an initial solution that meets DSX requirements. The equipment is a liquid-to-liquid portfolio that supports D2C (Direct-to-Chip) cooling and rear-door heat exchanger applications, effectively controlling heat generated in high-density environments.
As AI infrastructure scales up, cooling and power are emerging as key factors that determine the production capacity of entire facilities. With this certification, AI factory builders can adopt proven cooling products during the complex infrastructure integration process.
Source: IT DAILY · Kwon Young-seok
Original: https://www.itdaily.kr/news/articleView.html?idxno=242084
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.