IBM isn’t the only one. “We’ve been doing liquid cooling since 2012 on our supercomputers,” says Scott Tease, vice president and general manager of AI and high-performance computing at Lenovo’s infrastructure solutions group. “And we’ve been improving it ever since—we’re now on the sixth generation of that technology.” And the liquid Lenovo uses in its Neptune liquid cooling solution is warm water. Or, more precisely, hot water: 45 degrees Celsius. And when the water leaves the servers, it’s even hotter, Tease says. “I don’t have to chill that water, even if I’m in a hot climate,” he says. Even at high temperatures, the water still provides enough cooling to the chips that it has real value. “Generally, a data center will use evaporation to chill water down,” Tease adds. “Since we don’t have to chill the water, we don’t have to use evaporation. That’s huge amounts of savings on the water. For us, it’s almost like a perfect solution. It delivers the highest performance possible, the highest density possible, the lowest power consumption. So, it’s the most sustainable solution possible.” So, how is the water cooled down? It gets piped up to the roof, Tease says, where there are giant radiators with massive amounts of surface area. The heat radiates away, and then all the water flows right back to the servers again. Though not always. The hot water can also be used to, say, heat campus or community swimming pools. “We have data centers in the Nordics who are giving the heat to the local communities’ water systems,” Tease says.

2) “Units of compute”: OneCore and SmartRun On the earnings call, Albertazzi highlighted Vertiv OneCore, an end-to-end data center solution designed to accelerate “time to token,” scaling in 12.5 MW building blocks; and Vertiv SmartRun, a prefabricated white space infrastructure solution aimed at rapidly accelerating fit-out and readiness. He pointed to collaborations (including Hut 8 and Compass Data Centers) as proof points of adoption, emphasizing that SmartRun can stand alone or plug into OneCore. 3) Cooling evolution: hybrid thermal chains and the “trim cooler” Asked how cooling architectures may change (amid industry chatter about warmer-temperature operations and shifting mixes of chillers, CDUs, and other components) Albertazzi leaned into complexity as a feature, not a bug. He argued heat rejection doesn’t disappear, even if some GPU loads can run at higher temperatures. Instead, the future looks hybrid, with mixed loads and resiliency requirements forcing more nuanced thermal chains. Vertiv’s strategic product anchor here is its “trim cooler” concept: a chiller optimized for higher-temperature operation while retaining flexibility for lower-temperature requirements in the same facility, maximizing free cooling where climate and design allow. And importantly, Albertazzi dismissed the idea that CDUs are going away: “We are pretty sure that CDUs in various shapes and forms are a long-term element of the thermal chain.” 4) Edge densification: CoolPhase Ceiling + CoolPhase Row (Feb. 3) Vertiv also expanded its thermal portfolio for edge and small IT environments with the: Vertiv CoolPhase Ceiling (launching Q2 2026): ceiling-mounted, 3.5 kW to 28 kW, designed to preserve floor space. Vertiv CoolPhase Row (available now in North America) for row-based cooling up to 30 kW (300 mm width) or 40 kW (600 mm width). Vertiv Director of Edge Thermal Michal Podmaka tied the products directly to AI-driven edge densification and management consistency, saying the new systems “integrate seamlessly

DCF founder Rich Miller has spent much of his career explaining how the data center industry works. Now, with his latest venture, Data Center Richness, he’s also examining how the industry learns. That thread provided the opening for the latest episode of The DCF Show Podcast, where Miller joined present Data Center Frontier Editor in Chief Matt Vincent and Senior Editor David Chernicoff for a wide-ranging discussion that ultimately landed on a simple conclusion: after two years of unprecedented AI-driven announcements, 2026 will be the year reality asserts itself. Projects will either get built, or they won’t. Power will either materialize, or it won’t. Communities will either accept data center expansion – or they’ll stop it. In other words, the industry is entering its execution phase. Why Data Center Richness Matters Now Miller launched Data Center Richness as both a podcast and a Substack publication, an effort to experiment with formats and better understand how professionals now consume industry information. Podcasts have become a primary way many practitioners follow the business, while YouTube’s discovery advantages increasingly make video versions essential. At the same time, Miller remains committed to written analysis, using Substack as a venue for deeper dives and format experimentation. One example is his weekly newsletter distilling key industry developments into just a handful of essential links rather than overwhelming readers with volume. The approach reflects a broader recognition: the pace of change has accelerated so much that clarity matters more than quantity. The topic of how people learn about data centers isn’t separate from the industry’s trajectory; it’s becoming part of it. Public perception, regulatory scrutiny, and investor expectations are now shaped by how stories are told as much as by how facilities are built. That context sets the stage for the conversation’s core theme. Execution Defines 2026 After

Back in September 2025, we examined an ambitious proposal from infrastructure developer Joule Capital Partners – often branding the effort as “Joule Power” – in partnership with Caterpillar. The concept is straightforward but consequential: acquire a vast rural tract in Millard County, Utah, and pair an AI-focused data center campus with large-scale, on-site “behind-the-meter” generation to bypass the interconnection queues, transmission constraints, and substation bottlenecks slowing projects nationwide. The appeal is clear: speed-to-power and greater control over delivery timelines. But that speed shifts the project’s risk profile. Instead of navigating traditional utility procurement, the development begins to resemble a distributed power plant subject to industrial permitting, fuel supply logistics, air emissions scrutiny, noise controls, and groundwater governance. These are issues communities typically associate with generation facilities, not hyperscale data centers. Our earlier coverage focused on the technical and strategic logic of pairing compute with on-site generation. Now the story has evolved. Community opposition is emerging as a material variable that could influence schedule and scope. Although groundbreaking was held in November 2025, final site plans and key conditional use permits remain pending at the time of publication. What Is Actually Being Proposed? Public records from Millard County show Joule pursuing a zone change for approximately 4,000 acres (about 6.25 square miles), converting agricultural land near 11000 N McCornick Road to Heavy Industrial use. At a July 2025 public meeting, residents raised familiar concerns that surface when a rural landscape is targeted for hyperscale development: labor influx and housing strain, water use, traffic, dust and wildfire risk, wildlife disruption, and the broader loss of farmland and local character. What has proven less clear is the precise scale and sequencing of the buildout. Local reporting describes an initial phase of six data center buildings, each supported by a substantial fleet of Caterpillar

The Production Readiness Gap AI teams continue to confront a familiar challenge: moving from experimentation to predictable production performance. Models that train successfully on small clusters or sandbox environments often behave very differently when deployed at scale. Performance characteristics shift. Data pipelines strain under sustained load. Cost assumptions unravel. Synthetic benchmarks and reduced test sets rarely capture the complex interactions between compute, storage, networking, and orchestration that define real-world AI systems. The result can be an expensive “Day One” surprise:  unexpected infrastructure costs, bottlenecks across distributed components, and delays that ripple across product timelines. CoreWeave’s view is that benchmarking and production launch can no longer be treated as separate phases. Instead, validation must occur in environments that replicate the architectural, operational, and economic realities of live deployment. ARENA is designed around that premise. The platform allows customers to run full workloads on CoreWeave’s production-grade GPU infrastructure, using standardized compute stacks, network configurations, data paths, and service integrations that mirror actual deployment environments. Rather than approximating production behavior, the goal is to observe it directly. Key capabilities include: Running real workloads on GPU clusters that match production configurations. Benchmarking both performance and cost under realistic operational conditions. Diagnosing bottlenecks and scaling behavior across compute, storage, and networking layers. Leveraging standardized observability tools and guided engineering support. CoreWeave positions ARENA as an alternative to traditional demo or sandbox environments; one informed by its own experience operating large-scale AI infrastructure. By validating workloads under production conditions early in the lifecycle, teams gain empirical insight into performance dynamics and cost curves before committing capital and operational resources. Why Production-Scale Validation Has Become Strategic The demand for environments like ARENA reflects how fundamentally AI workloads have changed. Several structural shifts are driving the need for production-scale validation: Continuous, Multi-Layered Workloads AI systems are no longer

Cisco Targets the AI Fabric Bottleneck Cisco introduced its Silicon One G300, a new switching ASIC delivering 102.4 Tbps of throughput and designed specifically for large-scale AI cluster deployments. The chip will power next-generation Cisco Nexus 9000 and 8000 systems aimed at hyperscalers, neocloud providers, sovereign cloud operators, and enterprises building AI infrastructure. The company is positioning the platform around a simple premise: at AI-factory scale, the network becomes part of the compute plane. According to Cisco, the G300 architecture enables: 33% higher network utilization 28% reduction in AI job completion time Support for emerging 1.6T Ethernet environments Integrated telemetry and path-based load balancing Martin Lund, EVP of Cisco’s Common Hardware Group, emphasized the growing centrality of data movement. “As AI training and inference continues to scale, data movement is the key to efficient AI compute; the network becomes part of the compute itself,” Lund said. The new systems also reflect another emerging trend in AI infrastructure: the spread of liquid cooling beyond servers and into the networking layer. Cisco says its fully liquid-cooled switch designs can deliver nearly 70% energy efficiency improvement compared with prior approaches, while new 800G linear pluggable optics aim to reduce optical power consumption by up to 50%. Ethernet’s Next Big Test Industry analysts increasingly view AI networking as one of the most consequential battlegrounds in the current infrastructure cycle. Alan Weckel, founder of 650 Group, noted that backend AI networks are rapidly moving toward 1.6T architectures, a shift that could push the Ethernet data center switch market above $100 billion annually. SemiAnalysis founder Dylan Patel was even more direct in framing the stakes. “Networking has been the fundamental constraint to scaling AI,” Patel said. “At this scale, networking directly determines how much AI compute can actually be utilized.” That reality is driving intense innovation