Portuguese 
New architecture combines high-density servers, Nvidia GPUs, Cisco networking, and liquid cooling to tackle one of the largest data center expansions in history
Cisco has partnered with Supermicro to expand its artificial intelligence infrastructure based on Nvidia technologies, incorporating high-density servers and liquid-cooled systems. This new initiative will enable the handling of even larger AI workloads and is set to reach the market in October 2026, according to reports from UOL sourced from Estadão.
The move goes beyond simply adding new servers. Cisco aims to integrate computing, networking, security, monitoring, and cooling into a single architecture. This way, companies will be able to deploy structures capable of performing everything from AI inference to training models with trillions of parameters.
The expansion is occurring at a time when advancements in artificial intelligence are increasing demand for GPUs, electricity, cooling, and high-performance networks. According to Jeetu Patel, President and Chief Product Officer of Cisco, the industry is at the beginning of one of the largest data center expansions in history.
-
The Ground is Sinking and the Alert Grows: Soil May Give Way, Jam Doors, Crack Buildings, Damage Streets, and Worsen Flooding in Coastal Areas
-
Amazon to Spend Billions on New Mega Factory in the U.S. to Produce Robots Capable of Working Alongside Employees
-
China Submerges Hollow Sphere to 65 Meters, Turns Lake into Underwater Battery After 100+ Cycles in 10 Days
-
Nuclear Doomsday Submarine Becomes Reality: Russia Launches the Khabarovsk, Designed for Poseidon
Cisco Incorporates Supermicro Servers into Its AI Factory
The strategy revolves around the so-called Cisco Secure AI Factory with NVIDIA, an architecture designed to bring together the various components necessary for operating artificial intelligence at an enterprise scale.
Now, Supermicro will add an important piece to this structure: high-density servers with both air and liquid cooling.
With this, Cisco will be able to provide complete rack-scale systems. In practice, multiple servers, GPUs, and networking equipment will function as a large, integrated infrastructure rather than separate components that were purchased and configured individually.
This approach aims to simplify the deployment of large artificial intelligence clusters. Additionally, the company states that the infrastructure will be validated before it reaches customers, reducing technical risks during installation.
The strategy aligns with a much larger race for technological infrastructure. Major companies are investing in factories, data centers, energy, and physical equipment to support the next phase of artificial intelligence.
Liquid Cooling Takes Center Stage in the AI Race
One of the most significant aspects of the partnership is liquid cooling.
The greater the concentration of GPUs within a rack, the more challenging it becomes to remove the heat produced by the equipment. Therefore, cooling has become a strategic part of the infrastructure necessary for running advanced models.
The new architecture will allow for liquid-cooled deployments from the rack to the network. In this scenario, Cisco networking equipment will work alongside Supermicro’s liquid-cooled servers.
Additionally, the solution will have access to dense GPU systems based on Nvidia platforms.
Cisco itself emphasizes that energy and cooling are already influencing where new AI infrastructures can be installed. Thus, increasing computational power does not just mean adding more chips: it also requires addressing the physical bottlenecks created by this concentration of processing.
Systems Will Be Able to Work with Trillion-Parameter Models
The anticipated scale is noteworthy.
According to Cisco, the architecture will support applications ranging from edge processing to training models with trillions of parameters.
Among the platforms mentioned by the company are NVIDIA Vera Rubin NVL72 and NVIDIA HGX Rubin NVL8. The infrastructure also combines Cisco Silicon One switches and Nvidia Spectrum-X based equipment.
Thus, the company does not intend to sell only raw processing power.
The goal is to connect GPUs, servers, storage, networking, software, security, and observability tools into a unified architecture.
This type of integration becomes especially important as the scale of artificial intelligence projects continues to grow. The larger the cluster, the more complex it becomes to identify failures, manage performance, and coordinate thousands of components.
Nvidia provides the computing foundation for the new architecture
Nvidia plays a central role in the expansion.
Cisco had already been working with the chip manufacturer in its Secure AI Factory. Now, the integration of Supermicro’s systems broadens the computing options within the platform.
In addition to hardware, the architecture utilizes NVIDIA AI Enterprise, a suite of software aimed at deploying and operating artificial intelligence applications.
On the other hand, Cisco provides a significant portion of the network infrastructure.
The company claims to be the only technology partner of Nvidia using its own network switches and operating system within a solution compatible with the NVIDIA Cloud Partner architecture.
Consequently, the competition for AI infrastructure involves much more than just GPUs, which typically capture market attention.
Networks, storage, energy, cooling, software, and security also need to keep pace with the growing computational power.
Cisco aims to reduce the risk of deploying massive AI clusters
Another goal of the partnership is to lessen the complexity faced by companies.
Building a large-scale artificial intelligence infrastructure requires that servers, chips, networks, and software work together. However, an improper setup can limit performance even when extremely powerful GPUs are available.
For this reason, Cisco and its partners plan to provide pre-validated systems.
The company also announced new infrastructure validation services. These will check whether the equipment was installed according to the original design and adheres to the specified reference architectures.
Additionally, Cisco is investing in a dedicated large-scale AI laboratory to develop tools and test software associated with these services.
In this scenario, the goal is to accelerate the timeframe between the arrival of equipment and its actual use.
The company also aims to integrate metrics for processing, network cards, optical components, and network performance. This way, operators can observe different parts of the infrastructure within the same environment.
Partnership Targets Companies, Neoclouds, and Sovereign Clouds
The new architecture was not developed solely for traditional corporate data centers.
Cisco is also targeting the so-called neoclouds, companies specializing in providing intensive computing infrastructure for artificial intelligence.
Another strategic market involves sovereign clouds, structures designed to maintain data and processing under specific national or regional regulations.
Concerns over data control repeatedly surface in Cisco’s announcement. After all, organizations that train or run their own models may not want to transfer strategic information to external infrastructures without control over where it’s processed.
Therefore, the company combines performance with security and observability.
At the same time, the market is beginning to treat computing capacity as strategic infrastructure. Nvidia itself claims that so-called AI factories transform computation into intelligence and, subsequently, into economic value.
This expansion occurs while companies across various sectors are also allocating huge amounts to new technology and artificial intelligence operations.
New Systems Set to Launch in October 2026
There is a concrete date for the next phase.
Cisco announced that it will start offering Supermicro’s computing solutions within the Secure AI Factory with NVIDIA in October 2026.
This means that the announcement made in August still represents an expansion of the offering, rather than a confirmation of systems already deployed in clients through this new partnership.
Additionally, the company did not disclose a financial value for the agreement in the announcement. Therefore, there is no basis for attributing a specific billion-dollar figure to the partnership.
However, following the announcement, there was a positive reaction in the financial market. At 9:57 a.m. Brasilia time, Cisco stocks increased by 1.63% in pre-market trading in New York, while Supermicro shares rose by 2.56%, according to a report published by UOL.
The AI Race Begins to Transform Data Centers
During much of the current artificial intelligence race, GPUs have taken center stage. However, as models become more powerful, it becomes increasingly evident that another challenge arises: a massive infrastructure needs to be built around these chips.
Energy, liquid cooling, networks, security, and storage increasingly determine how much computational power can actually be brought online.
It is precisely at this point that Cisco, Supermicro, and Nvidia intend to collaborate.
Starting in October, Cisco will begin offering architecture that combines high-density servers, liquid cooling, and specialized networks to support immense AI workloads.
Thus, the next technological battle may not take place just within the chips. It will also be in the racks, cables, liquid cooling systems, and data centers required to keep thousands of GPUs operating simultaneously.
