Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/nvidia-closes-deal-to-acquire-startup-lepton-ai-report.22518/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Nvidia closes deal to acquire startup Lepton AI: report

Daniel Nenni

Admin
Staff member
NVIDIA CEO Says Restricted Supply Has Frustrated Its Clients, Increasing Business Tensions 1

NVIDIA has officially acquired the startup Lepton AI, as Team Green is now looking towards a complete "vertical integration" of the supply chain to maintain dominance.

Team Green's Deal With Lepton AI Will Now Allow Them To Offer Its AI Clusters To Clients Through Rental Services Directly

Well, NVIDIA has indeed made a surprise acquisition, given that the deal with Lepton AI hasn't been confirmed officially yet, but The Information reports that the firm has sealed the deal. For those unaware, Lepton AI is known for its "server-rental" services, which recently have been in tremendous demand due to their cost-effectiveness. It seems like the primary factor behind this acquisition is NVIDIA's attempt to have a grip on the supply chain entirely instead of just being responsible for the manufacturing part.

Lepton AI does the renting business by leasing AI clusters from CSPs and then offering them to smaller clients, running an "AI-as-a-Service" (AIaaS) model. The firm doesn't only provide computing resources; with its tech stack, users can benefit from features such as auto-scaling, error handling during model deployment, and much more. Clients looking to run LLMs can rent out hardware from Lepton AI, which allows them to avoid worrying about the financial costs associated with creating their hardware portfolio.

NVIDIA Blackwell AI Servers Exposed To Component Shortage, Limited Supply Expected In Q4 2024 1


With the acquisition, Lepton AI's founder Yangqing Jia, also Alibaba's former VP of AI and Data Analytics, will join NVIDIA, although his role isn't defined for now. Now, coming to the significance of the deal, it seems like Team Green wants the "whole share" of the AI markets, and this acquisition will indeed allow them to control more of the supply chain—from chip design to server rental. Essentially, Lepton AI can now access hardware directly from NVIDIA, avoiding all the hassle of leasing hardware from third-party sources.

Team Green could also ensure that it doesn't experience "inventory oversupply" with the deal, given that it manages to stack up AI clusters. Since renting is a far cheaper option, the demand would certainly be there. This also positions NVIDIA to compete with ASICs from the likes of Amazon, Google, and Microsoft. Team Green has yet to announce the deal officially, but it seems like the firm won't stop with its AI ventures.

 
Great acquisition! Yangqing Jia is a PhD from Berkely, Google, Facebook, Alibaba, hopefully he is taking a big role at Nvidia. He is an open-source guru.

Lepton AI Cloud

1744163566722.png
 
It seems like the primary factor behind this acquisition is NVIDIA's attempt to have a grip on the supply chain entirely instead of just being responsible for the manufacturing part.
Think the bigger motivation is that data centers are looking more and more like giant chips composed of rack-sized chiplets. That's why Jensen says that Dynamo “the operating system of the AI factory”. This is just a natural way of offering the most cost and performance-oriented solution.

 
The first general business article that seems to get why NVIDIA is designing, building and operating entire racks, and data centers for inference (plus explains why AMD/ZT and Intel are getting pulled into doing the same).

“Take the architecture of data centers. In recent decades, the idea was to link together huge numbers of cheap, commodity servers. But Nvidia believes in scaling up before scaling out—making a rack of computers as powerful as possible before they are yoked together in a vast infrastructure.”

“And more than any single technological innovation, Nvidia’s success turns on its ability to create an interconnected system in which customers and partners contribute to its infrastructure or build on top of it.”

 
Back
Top