Price Pressures Drive up Hardware Costs

Posted by:

Carl Easterday

|

On:

|

IT procurement teams are still seeing price pressures drive up hardware costs in server builds not because every part is scarce, but because a few high-value components remain structurally tight—and they set the floor for the entire configuration. We’re seeing this across the board by all manufactures. We are recommending anyone who needs server hardware or client hardware to place the order immediately, as we don’t see any immediate relief and indeed you may see 20% increases in the next month.

GPUs/accelerators remain the clearest choke point. Even as overall supply capacity expands, demand for AI-capable systems is rising faster, keeping allocation and long-commit purchase behavior in place. NVIDIA has publicly signaled that next-generation demand is likely to outstrip supply for multiple quarters, which typically translates into longer queue times, less discounting, and more bundle-like deals (GPU + platform + networking) rather than clean line-item competition.

Memory is the second major driver—especially server DDR5 and “AI-adjacent” memory capacity needs. TrendForce expects DDR5 to rise from ~40% of server DRAM bit shipments in 2024 to ~60–65% in 2025, a transition that usually brings pricing volatility as buyers chase specific validated SKUs and capacities. In parallel, AI buildouts pull manufacturing focus toward high-end memory products, tightening availability and negotiating leverage for conventional server DIMMs. TrendForce also notes market dynamics (including shipment front-loading and inventory moves) that can amplify quarter-to-quarter pricing swings.

The knock-on costs matter as much as the headline parts. GPU-dense servers frequently require higher-watt power supplies, advanced cooling, and faster networking, turning a “GPU shortage” into a full-platform cost increase. For OEM quotes, that often shows up as shorter quote validity windows, constrained substitutions, and delayed delivery dates that force buyers into expedited freight, staging services, or interim capacity (cloud/colo) spend.

Finally, software is adding a parallel layer of budget pressure. Broadcom’s VMware packaging/licensing changes mean hardware refresh math increasingly needs a “stack TCO” view (compute + support + virtualization/subscription), not a chassis-only comparison—pushing procurement to negotiate multi-year bundles and exit options earlier in the cycle.

Sources / links

Posted by

Carl Easterday

in