It’s that time of year… when luminaries across industries pull out the crystal ball and try to discern what’s in store for the next 12 months. Here at OCOLO, we appreciate when industry experts look both back and forward, grading themselves on their prior years calls and baking that take into the next year’s. So, let’s hear what Uptime Institute has to say about their 5 Data Center Predictions for 2024 and what they’re expecting for 2025.
2024 Report Card
Uptime Institute gives themselves high marks for every one of their 2024 predictions. We agree… except would argue with #2… While we take the point that hyperscalers and larger colocation data centers were by far the greatest beneficiaries of surging AI demand in 2024 and will continue to be going forward, we at OCOLO feel it’s a reach to say that demand for AI had a “limited impact on most operators” in 2024. Beyond AI, the proliferation of data for general use cases continued unabated as well, lifting all boats in the DC-related space.
5 Data Center Predictions for 2025
Here’s what Uptime Institute predicts is onboard for 2025. No surprises here: continued AI-driven growth; chipmaking juggernaut NVIDIA maintaining its dominant positioning; deep-pocketed hyperscalers leading the DC investment charge; ongoing capacity constraints and the need for ingenuity in developing both new technology and renewable energy sources to navigate all the growth and evolution.
What’s more, these predictions are underpinned by some massive numbers that aren’t just predictions – they’re actually already manifesting today in the industry. Consider these facts around increasing densification with each new generation of AI compute racks developed by NVIDIA:
- Today’s NVIDIA H-series products are typically built around 40kW per standard 19-inch rack – already ahead of the 30kW standard most organizations have as standard high-density racks
- NVIDIA is on track to begin volume shipments of its first rack-scale GPU systems in 2025 with high-end configurations of 72 GPUs per rack rated at 130kW, with half-sized versions in the 60kW to 70kW range.
- Subsequent chip upgrades demand event greater power in 2025, with the 2026 generation project to push thermal power ratings beyond 2kW per GPU.
- NVIDIA’s product roadmap calls for doubling the number of GPUs per rack — and then doubling again. These rack-scale systems, 300kW and above, are slated to start shipping as early as 2026.
It will be interesting to see if/how these 5 predictions unfold throughout the year in an industry as dynamic and unpredictable as digital infrastructure. The executive summary give a nice taste of how Uptime Institute is thinking about the road ahead, and if you’re interested, you can read the executive summary here.
