NVIDIA launches Lepton—an online marketplace for AI computing power
What's the story
NVIDIA has launched a new software platform called Lepton to create a marketplace for cloud-based artificial intelligence (AI) chips.
The company, which dominates the market with its graphics processors (GPUs) used for training AI models, wants to make the buying and selling of GPU capacity easier.
Lepton will be a centralized hub where cloud computing firms can sell their GPU capacity.
Platform expansion
Multiple firms to join NVIDIA's Lepton platform
Several firms have already signed up for NVIDIA's Lepton platform, including CoreWeave and Nebius Group.
Other participating companies include Crusoe, Firmus, Foxconn, GMI Cloud, Lambda, Nscale, SoftBank Corp., and Yotta Data Services.
Alexis Bjorlin from NVIDIA said that despite the growing demand for chips from both start-ups and big corporations, finding available chips has been a "very manual" process until now.
Enhanced accessibility
Lepton to simplify access to NVIDIA's ecosystem
Bjorlin stressed that Lepton is aimed at making the process of finding available chips more seamless.
"It's almost like everyone's calling everyone for what compute capacity is available," he said.
The idea is to give global clouds and new cloud providers easy access to NVIDIA's entire developer ecosystem.
The move comes amid a spike in demand for chips at start-ups and large companies.
Participation status
Major cloud providers yet to join Lepton
So far, leading cloud providers like Microsoft, Amazon Web Services (AWS), or Alphabet's Google have not signed up for NVIDIA's Lepton partner list.
However, Bjorlin said these companies would have the option to sell their capacity on the marketplace, if they want.
This indicates potential for future expansion and participation from other big players in the industry.
Future features
Lepton to offer location-based search for NVIDIA chips
Bjorlin revealed Lepton will eventually let developers search for NVIDIA chips in specific countries. This would help meet data storage requirements more efficiently.
Plus, companies that already own some NVIDIA chips would find it easier to search for more to rent through this platform.
However, NVIDIA has not disclosed its business model or if it would charge commissions/fees on either developers or clouds.