Key Points:
- If there’s one seemingly safe bet about how the cloud will impact data centers in 2025, it’s that growing deployments of AI workloads in the cloud will lead to more pressure on data center operators to improve energy efficiency.
- The reason why is simple: AI (especially generative AI) is energy-hungry – so much so that some analysts believe utility providers won’t be able to supply enough electricity to power AI without major infrastructure upgrades.
- Overall, I’m bearish about how close we are to having quantum computers that are stable enough for real-world use. But steady advancements in quantum technology – including Google’s recent unveiling of a new type of quantum chip – are driving increased interest in quantum possibilities.
- The bar for what counts as “good” levels of performance and reliability from cloud-based workloads gets higher and higher all the time. It doesn’t matter how fast your servers or applications are if they can’t move data over the network quickly enough to meet user expectations.
- Cloud providers have begun talking about Water Usage Effectiveness (WUE) in recent years, and I expect that that conversation will intensify, due especially to ongoing water shortages in many of the places that are home to data centers.
- So, keep reading for my take on what’s ahead in the realm of the cloud, and what it means for data centers everywhere.
Welcome to Data Center Knowledge Discoveries!
You already know that every day at Data Center Knowledge brings advice, trends and strategies for data center professionals on how to design, build, and manage world-class data centers.
That means original reporting from our team of journalists and unique commentary you won’t see anywhere else! But in case you missed them, here are some of our other must-read favorites from this week:
Future Proofing Data Centers
Key Points:
- In 2025, five key forces will fundamentally reshape data center cooling: the mounting “heat crisis,” supply chain reengineering, power availability, sustainability as a core business imperative, and the rise of specialized cooling solutions.
- Analysts predict that AI workloads alone could push data center energy consumption up by 12% by 2030. This shift is due largely to the intense processing power required for AI applications, which generates substantial heat.
- In today’s competitive landscape, the surge in AI demand has intensified the race for critical components, with supply chain vulnerabilities now shaped by factors like port strikes and heightened global competition.
- Data center cooling has reached a tipping point. Embracing resilient and energy-efficient solutions will enable operators to thrive in the future. As resource constraints intensify, forward-thinking cooling strategies are no longer just an option – they’re essential to the data center industry’s survival and success.
This Hot Topic Has Cool Trends for 2025
Key Points:
- Perhaps the most important data center cooling trend that will impact the sector in 2025 is increased demand on cooling systems. This is due especially to ongoing deployment of AI workloads, which tend to generate more heat than traditional applications.
- One way to dissipate heat more effectively is to replace traditional air coolers with liquid cooling systems.
- These may include immersion cooling, which covers equipment in non-conductive liquid, an extremely efficient – but costly – cooling method, and direct-to-chip cooling, which transfers heat directly from chips.
- Traditionally, data center operators have aimed to keep server room temperatures in the low-70 degrees Fahrenheit or below.
- But some data center companies, such as
Equinix
, have begun experimenting with somewhat higher temperatures in their server rooms, and they’re reporting success. By increasing target temperatures to the higher 70s, they can reduce the load on cooling systems without experiencing overheating events for IT equipment.
Dissecting Google's Willow Chip
Key Points:
- Google’s most significant achievement is a major advancement in quantum error correction – a critical challenge in quantum computing due to the fragile nature of qubits. Qubits exist in a “superposition” state, making them highly susceptible to errors from environmental interference.
- Traditionally, increasing the number of connected qubits in a “surface code lattice” has led to higher error rates – the opposite of what is needed to create logical qubits from physical ones.
- Google scaled from a 3×3 to a 7×7 physical qubit lattice while reducing the error rate by a factor of 2.14, effectively doubling the lifespan of logical qubits compared to its earlier Sycamore chip.
- While Google’s Willow chip doesn’t alter our 10–15-year timeline for large-scale, fault-tolerant quantum computing, it sends a clear signal:?Start preparing now, because it could be sooner.
Major Moves Inside the Industry
The Data Center Knowledge News Roundup brings you the latest news and developments across the data center industry – from investments and mergers to security threats and industry trends.
Key Points:
- Finally, this wouldn’t be a Data Center Knowledge news roundup without a rundown of the most important data center developments of the past week. Even during the holiday season, new data centers were constructed and announced.
- On December 23,
Equinix
announced plans to acquire
英国电信集团
's Irish data center business, including data centers in CityWest and Ballycoolin, Dublin, for €59 million. Equinix currently operates six data centers in the Dublin area.
- On December 27,
Bit Digital | NASDAQ: BTBT
acquired a site for a 5 MW Tier-3 data center expansion project in Montreal, Canada. The New York-based HPC infrastructure company paid CAD $33.5 million for the 160,000 sq ft MTL2 site and expects to spend an additional CAD $27.6 million to complete a retrofit by May 2025.
Chip Watch: Commentary of the Week
Key Points:
- New research from Omdia?highlights the rapid growth in demand for
谷歌
’s Tensor Processing Unit (TPU) AI chips, a trend that may be strong enough to start chipping away at
英伟达
’s market dominance in GPUs.
-
博通
’s CEO
Hock Tan
has repeatedly revised up his target for AI semiconductor revenue, aiming for $12 billion for this year.
- Based on this, it is estimated that Google’s TPUs could account for somewhere between $6 billion – close to Omdia’s existing estimates – and $9 billion, depending on the breakdown between compute and networking devices.
-
Alexander Harrowell
, principal analyst at Omdia?noted: “Even though there is some uncertainty in the exact ratio between compute and networking devices, shipments of TPUs, even at the lower $6bn estimate, are growing at a pace fast enough to take share from NVIDIA for the first time.
Latest Major Tech Layoff Announcements
Original Story by Jessica C. Davis, Updated by Brandon Taylor
Key Points:
- As COVID drove everyone online, tech companies hired like crazy. Now we are hitting the COVID tech bust as tech giants shed jobs by the thousands.
- Updated January 3, 2025 with layoff announcements from Bench Accounting, Lilium, BionicHIVE and Refinery29.
- Check back regularly for updates to InformationWeek's IT job layoffs tracker.
This is just a taste of what’s going on. If you want the whole scoop, then register for one of our email newsletters,?but only if you’re going to read it.?We want to improve the sustainability of editorial operations, so we don’t want to send you newsletters that are just going to sit there unopened. If you're a subscriber already, please make sure Mimecast and other inbox bouncers know that we’re cool and they should let us through.
Our bi-weekly LinkedIn newsletters arrive on Friday afternoons, so keep your eyes peeled for the top stories you may have missed between now and then.
Sandeep
1 个月Netweb Technologies India Ltd. Black Box ????
Lead Brand Ambassador/Head of Resume Screening/Business Development Representative at BRUNS-PAK Data Center Solutions
1 个月The future paradigm will be focused on plasma energy, or any new source of energy that is transportable. Next level cooling systems beyond water maybe transdermal chemicals or new types of refrigeration systems that circulate air more efficiently. Data Center trends need to think beyond edge computing, quantum chips server performance to a next level AI integration to be discovered. Engineering the future is exciting!
Retired
1 个月I totally agree about the huge energy demands for data centers using next gen AI chips and server systems processing as well as serious power sourcing, delivery, control and moitoring upgrade requirements needed for the grid. Liquid cooling using natural water sources might be something to look at except for the possibilities of real “global warming”. I heard IBM is working on developing the largest quantum computer so far. Should be interesting to see the how the synergistic development of these two technologies converge ahead.