Speed Is Becoming a Design Requirement<\/strong><\/p>\n\n\n\nRyan identified five forces shaping the market: compute densification, speed, scale, complexity, and changing load profiles.<\/p>\n\n\n\n
Each is putting pressure on conventional data center delivery models.<\/p>\n\n\n\n
Higher density means more power and heat. Larger projects require greater manufacturing and service capacity. Faster schedules leave less time for extensive onsite assembly and testing.<\/p>\n\n\n\n
Vertiv is responding by changing how systems are designed and delivered.<\/p>\n\n\n\n
Ryan highlighted solutions such as Woodcourt and SmartRun, which are intended to reduce onsite complexity and accelerate commissioning.<\/p>\n\n\n\n
The broader approach involves completing more assembly and testing before equipment reaches the customer site.<\/p>\n\n\n\n
Tokens Are Changing How Efficiency Is Measured<\/strong><\/p>\n\n\n\nThe discussion also moved beyond conventional measures of data center performance.<\/p>\n\n\n\n
Ryan identified four metrics that Vertiv considers important for an AI factory: tokens per second, tokens per watt, time to first token, and tokens per dollar.<\/p>\n\n\n\n
Tokens per second reflects output speed. Tokens per watt addresses energy efficiency. Time to first token focuses on responsiveness, while tokens per dollar connects performance with cost.<\/p>\n\n\n\n
Together, they show why power and cooling are becoming closely tied to AI economics.<\/p>\n\n\n\n
Ryan said the objective is to design power and thermal systems that operate efficiently as a whole, helping increase the amount of useful computing output generated from available energy.<\/p>\n\n\n\n
Taking More Work Off the Construction Site<\/strong><\/p>\n\n\n\nOne of the practical changes Ryan described is greater use of factory assembled and factory tested systems.<\/p>\n\n\n\n
Instead of bringing individual products to a data center and integrating them onsite, components can be assembled and tested before shipment.<\/p>\n\n\n\n
For a power system, this could involve switchgear, busway, UPS systems, and other equipment being brought together as a tested unit.<\/p>\n\n\n\n
The potential benefit is a simpler onsite process.<\/p>\n\n\n\n
Teams have fewer individual components to connect and test. Coordination between different suppliers can also be reduced.<\/p>\n\n\n\n
For large AI projects operating against compressed timelines, moving more work into a controlled manufacturing environment can help reduce onsite delays and commissioning requirements.<\/p>\n\n\n\n
SmartRun Shows the Approach<\/strong><\/p>\n\n\n\nSmartRun was presented as an example of this model in white space infrastructure.<\/p>\n\n\n\n
Ryan described the system as an alternative to integrating multiple individual components at the customer site.<\/p>\n\n\n\n
A traditional approach can require several organizations and teams to coordinate installation, integration, testing, and commissioning.<\/p>\n\n\n\n
Factory integration changes where much of that work takes place.<\/p>\n\n\n\n
The system can be tested before reaching the facility, potentially reducing onsite complexity, risk, and cycle time.<\/p>\n\n\n\n
For high density AI deployments, those gains can ultimately contribute to faster operational readiness.<\/p>\n\n\n\n
Cooling Is Moving Into a New Phase<\/strong><\/p>\n\n\n\nPower is only one side of the challenge.<\/p>\n\n\n\n
Higher compute density also produces more heat, increasing the need for advanced thermal management.<\/p>\n\n\n\n
Ryan noted that direct liquid cooling has moved from a relatively new technology two years ago to a standard solution for high density environments.<\/p>\n\n\n\n
That change is also creating new requirements for fluid management, maintenance, monitoring, and specialist services.<\/p>\n\n\n\n
The transition to 800 volt DC power is another development on the horizon.<\/p>\n\n\n\n
However, Ryan said it will happen progressively rather than through a single industry wide switch. Adoption will vary across regions and over time.<\/p>\n\n\n\n
Services Become Part of the Operating Model<\/strong><\/p>\n\n\n\nAs facilities become more complex, the role of services is also expanding.<\/p>\n\n\n\n
Ryan outlined activities ranging from consulting and design optimization to deployment, commissioning, maintenance, and ongoing performance optimization.<\/p>\n\n\n\n
He also highlighted Vertiv NextPredict, which uses operational data to support predictive maintenance and optimization.<\/p>\n\n\n\n
According to Ryan, intelligent tuning can deliver up to 15 percent energy savings in certain circumstances.<\/p>\n\n\n\n
For GPU intensive environments, the ability to continuously monitor and optimize power and thermal performance can have a direct impact on operating efficiency.<\/p>\n\n\n\n
Scaling Also Means Scaling Skills<\/strong><\/p>\n\n\n\nThe physical expansion of data centers is creating another requirement: more skilled people.<\/p>\n\n\n\n
Ryan said Vertiv has approximately 5,300 service engineers, more than 31 training academies across six continents, and approximately 320 service centers.<\/p>\n\n\n\n
He also discussed apprenticeship programs that allow employees in some countries to work while pursuing formal qualifications.<\/p>\n\n\n\n
Depending on the location, these programs can extend to national degree and master’s degree levels.<\/p>\n\n\n\n
The company’s academies also train employees on how equipment is designed and manufactured before they take that knowledge into customer environments.<\/p>\n\n\n\n
As deployment timelines become shorter, having trained teams available across the lifecycle becomes increasingly important.<\/p>\n\n\n\n
A Larger Technology Ecosystem<\/strong><\/p>\n\n\n\nThe changes are also affecting relationships across the technology supply chain.<\/p>\n\n\n\n
Ryan pointed to closer coordination between infrastructure providers, accelerator companies such as Nvidia and AMD, and component suppliers involved in areas such as pumps and heat exchangers.<\/p>\n\n\n\n
He said validated architectures and system level testing are becoming increasingly important.<\/p>\n\n\n\n
The reason is that performance increasingly depends on interactions between computing, power, thermal management, software, and physical systems rather than on any single component.<\/p>\n\n\n\n
What the Shift Means<\/strong><\/p>\n\n\n\nThe discussion at Vertiv Week points to a data center model built around speed, density, efficiency, and repeatability.<\/p>\n\n\n\n
The challenge is no longer simply adding more computing capacity.<\/p>\n\n\n\n
Operators need to deploy larger facilities faster, manage higher power and thermal loads, reduce onsite complexity, and maintain performance after systems go live.<\/p>\n\n\n\n
Technologies such as liquid cooling and 800 volt DC power will continue to develop at different speeds across markets.<\/p>\n\n\n\n
Meanwhile, factory based assembly, system testing, digital optimization, and workforce development are becoming increasingly important parts of the delivery model.<\/p>\n\n\n\n
As these changes accelerate, AI infrastructure is becoming less about individual pieces of equipment and more about how efficiently an entire data center can turn power and computing capacity into useful AI output.<\/p>\n\n\n\n
<\/p>\n","protected":false},"excerpt":{"rendered":"
AI infrastructure is being pushed to its limits by a […]<\/p>\n","protected":false},"author":58,"featured_media":107514,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"inline_featured_image":false,"_monsterinsights_skip_tracking":false,"footnotes":""},"categories":[9619],"tags":[7151],"contributor":[9732],"class_list":["post-107515","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-top-stories","tag-vertiv","contributor-news-desk"],"featured_image_src":"https:\/\/techxmedia.com\/en\/wp-content\/uploads\/2026\/09\/How-Is-Vertiv.jpg","author_info":{"display_name":"Lubna","author_link":"https:\/\/techxmedia.com\/en\/author\/lubna\/"},"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t