Confirm condition, inspection evidence and commercial terms in the quote.
🖥️

NVIDIA

Photo Coming Soon

GPUs & Accelerators

NVIDIA

NVIDIA L40 — Used & Refurbished (NVIDIA)

Part No. NVIDIA L40

Price on request

Ask for a quote tied to the required configuration and condition.

Contact for availability. Stock and condition require confirmation for a specific unit.

Place a Bid or Ask

Talk to a specialist · parts@caladansemi.com

Get a quote for this partPricing and availability to be confirmed

✓Request unit-specific inspection and test records
✓Confirm included documentation and accessories
✓Agree delivery, payment and return terms in the quote

How Much Does NVIDIA L40 Cost in 2026?

In 2026, used NVIDIA L40 GPUs typically range from $3,000 to $25,000, depending on condition. Used units with unknown maintenance history start at $3,000–$5,000. Refurbished models (with replaced VRM, capacitors, and thermal paste) cost $8,000–$15,000. Vendor-certified (e.g., NVIDIA-Certified or third-party with warranty) units top the range at $18,000–$25,000. Compared to the 2026 OEM list price (~$30,000), buyers save 50–90%, but only if the card passes PCIe 5.0 compatibility and power delivery checks. Caveat: 30% of 2026 listings lack proper firmware updates (e.g., NVSwitch 5.2.3), which can void savings if re-flashing is required.

NVIDIA L40 Specifications and Compatible Systems

The NVIDIA L40 features 24GB GDDR6 VRAM (384-bit bus), 10,752 CUDA cores, 600 GB/s memory bandwidth, and a TDP of 300W. It requires a PCIe 5.0 x16 slot and at least a 750W 80+ Gold PSU. Compatible systems include Dell PowerEdge R760 (2022), HPE ProLiant DL380 Gen11 (2021), and Lenovo SR670 (2023). Avoid pre-2020 servers with PCIe 3.0 x8 links; bandwidth drops to 4GB/s, degrading AI training by 30%. Warning: BIOS versions prior to 2.3.4 (2023) may fail to initialize dual-slot L40s in multi-GPU setups.

What to Check Before Buying a Used NVIDIA L40 in 2026

  1. Firmware version: Use nvidia-smi -q to confirm firmware ≥ 32.41.5. Older versions lack ECC memory support.
  2. Thermal health: Measure GPU temp under load (e.g., stress-ng --cuda 1). Fail if >75°C at 50% utilization.
  3. VRM stability: Run i2cget on I2C bus 8 to check VRM temperatures. >85°C at idle = bad capacitors.
  4. Memory errors: Run MemTestG80 for 24hr.

Quick Info

ConditionConfirm in quote
CategoryGPUs & Accelerators
Secondary Market

Place a Bid or Ask

Submit your buying or selling requirements for NVIDIA L40 — Used & Refurbished (NVIDIA).

Want to Buy
Want to Sell

Confirm export requirements and payment terms before any transaction · parts@caladansemi.com