How To Overcome Rising Enterprise Hardware Lead Times

Enterprise hardware procurement has become significantly more difficult. Standard server orders that once took 6-10 weeks now require 20-52 weeks, with the gap widening. 

Key Drivers Include:

• AI infrastructure buildout: Hyperscalers have locked in multi-year manufacturing capacity with Tier 1 ODMs, consuming an estimated 40–60% of high-end server production slots in 2024 and pushing enterprise buyers to the back of queues. This constrains GPUs, HBM, cooling, and power components.
• Geopolitical diversification: Efforts to reduce reliance on Taiwan require lengthy new fab and supplier qualifications, temporarily reducing available capacity.
• Custom configurations: Specialized orders (GPU-dense servers, NVMe-heavy storage, high-radix switches) have increased sharply, reducing availability of standard bill-of-materials inventory.

The GPU Allocation Issue

For AI and ML workloads, GPU-optimized servers (Dell PowerEdge, HPE ProLiant, Lenovo ThinkSystem) now quote 32–52 weeks for H100/H200/Blackwell configurations due to NVIDIA GPU allocation priorities for hyperscalers. Standard x86 CPU servers are at 12-20 weeks (vs. 6-8 weeks pre-2022), driven by advanced packaging constraints for the Intel Sapphire Rapids and AMD Genoa/Bergamo, plus persistent demand from refresh cycles that were delayed during the pandemic period.

Custom Configurations are Taking Longer

Custom configurations diverge sharply: a standard 2-socket server might take 14 weeks, while one with non-standard PCIe, NVMe, or power options can take 28 weeks. The latter requires sourcing components outside the standard production run. This makes “design-for-availability” a key constraint alongside performance.

Enterprise networking hardware core switches, edge routers, campus access layers, data center spine-and-leaf fabrics are experiencing lead time pressure from a different source than servers: custom silicon. Networking faces ASIC bottlenecks from Broadcom, Marvell, and Intel, constrained by TSMC/Samsung capacity limits and AI-driven demand for high-radix switches. Core data center switches (400/800GbE) quote 18-28 weeks; campus equipment is 10-16 weeks, but still above historical norms. A single ASIC shortage can create sudden multi-quarter constraints across a vendor’s product line.

NAND flash supply has tightened after 2022-2023 oversupply, with manufacturers prioritizing premium AI configurations. Major manufacturers including Samsung, SK Hynix, and Micron have reduced wafer starts and shifted capacity toward premium NAND configurations used in AI training storage: high-density, high-bandwidth modules which command better margins.

All-flash and NVMe-dense arrays now take 14-22 weeks for standard configurations, compounded by software certification requirements. Purchasing a net-new all-flash array from a vendor like Pure Storage, NetApp, or Dell PowerStore involves not just hardware availability but also software version compatibility with the buyer’s existing infrastructure.

The HDD market went through a severe correction in 2021-2022 when cloud providers over-ordered, then cancelled. Seagate and Western Digital adjusted manufacturing accordingly. Now, high-capacity HDDs (20-32TB HAMR/MAMR) have also tightened due to growing AI data volumes, though less dramatically.

IT and procurement leaders cannot eliminate supply chain risk, but they can manage it. The most effective strategies share a common theme: move planning horizons earlier and create optionality.

• Extend forecasting to 18-24 months and align finance/infrastructure teams on multi-year planning. The old 90-day procurement cycle is incompatible with 32-week lead times. Data center teams that want hardware available for a Q4 project need to be placing orders in Q1 or Q2.
• Standardize configurations to stay on volume production lines; engage OEMs early for any necessary customs. Every deviation from an OEM’s standard bill of materials adds lead time risk. Where possible, select configurations from an OEM’s volume production lines.
• Partner with certified resellers holding in-stock inventory (e.g., IT Creations) for fulfillment in days instead of months. Working with a trusted enterprise hardware reseller that holds a broad, immediately available inventory is one of the fastest ways to bypass OEM queue times entirely. Unlike ordering directly from an OEM, a reseller with physical stock on hand can fulfill orders in days rather than months critical when a project timeline cannot accommodate a 30-week wait.

When OEM lead times are unacceptable and a project cannot wait, the market for alternatives to new OEM equipment has grown significantly.

• Certified pre-owned (CPO): Refurbished enterprise servers, switches, and storage arrays from reputable resellers offer immediate or near-term availability at 30-70% below new OEM pricing. For non-critical workloads and short refresh cycles, CPO is increasingly the first call rather than the last resort.
• Secondary market: Platforms and specialist resellers carry large inventories of off-lease and end-of-support equipment, a legitimate sourcing channel for development, test, QA, and production workloads where OEM support is not required.