Yotta NM1 Navi Mumbai
- Uptime Institute Tier IV certified
- Dual utility feeds from separate substations
Ada Lovelace · PCIe
L40S servers, bought and hosted in India.
The L40S is a PCIe data-centre GPU that handles inference, rendering and graphics. It costs far less per card than SXM systems and fits standard servers.
- GPU memory
- 48 GB GDDR6 per GPU
- GPU power
- 350 W per GPU
- Server power
- 2.8 kW of GPU power for 8 cards, plus CPUs and fans
- Rack space · cooling
- Typically 4U · Air
Shipping Lead time, October 2026: Quoted per order · All GPU lead times
Good fit for
Who buys L40S.
- Inference for small and mid-size models
- Rendering and visualisation
- Mixed AI and graphics workloads
The hardware
What an L40S server looks like.
4U server with up to 8 PCIe cards. 48 GB GDDR6 per GPU, 350 W per GPU, typically 4u of rack space.
- Server power: 2.8 kW of GPU power for 8 cards, plus CPUs and fans
- Cooling: Air
Hosting
Where an L40S server runs.
L40S servers fit a standard rack and power budget, so they are the easiest GPU class to place quickly. Several can share one rack.
How managed hosting worksYotta D1 Greater Noida
- Dual 220 kV feeds with an on-site substation
- 48 hours of on-site backup power
L40S questions
Is L40S good for LLM inference?
For models that fit in 48 GB per card, or that split cleanly across a few cards, yes. For very large models, H100 or H200 memory bandwidth wins.
Can L40S be used for rendering?
Yes. It has RT cores and is supported by GPU render engines; check your renderer’s licensing for server use.
How many L40S cards fit in one server?
Common 4U platforms take up to eight double-width cards.
Other GPU servers
H200141 GB HBM3e per GPU · Up to ~10.2 kW per serverH10080 GB HBM3 per GPU · Up to ~10.2 kW per serverB200180 GB HBM3e per GPU · Up to ~14.3 kW per serverB300288 GB HBM3e per GPU · Up to ~14 kW per serverGB300 NVL7220 TB HBM3e across 72 GPUs · ~132–140 kW per rackRTX PRO 6000 Blackwell96 GB GDDR7 per GPU · Up to 4.8 kW of GPU power for 8 cards, plus CPUs and fansAMD MI300X192 GB HBM3 per GPU · 6 kW of GPU power for 8 GPUs, plus CPUs and fans
Owning one
Managed GPU hosting
Rack, power, cooling, network and remote hands for servers you own.
Learn more →Finance the purchase
Lender-ready quotes and asset records for your bank or NBFC.
Learn more →
Power and cooling guide
kW, amps and tons of cooling for every GPU class.
Learn more →
Sell or swap
Tested, valued and sold, or traded up for newer GPUs.
Learn more →