Topics
Latest
AI
Amazon
Image Credits:Steve Jennings/Getty Images for TechCrunch
Apps
Biotech & Health
Climate
Image Credits:Steve Jennings/Getty Images for TechCrunch
Cloud Computing
Commerce Department
Crypto
Enterprise
EVs
Fintech
fund-raise
appliance
Gaming
Government & Policy
ironware
Layoffs
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
Security
societal
Space
Startups
TikTok
Transportation
speculation
More from TechCrunch
Events
Startup Battlefield
StrictlyVC
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
Andreessen Horowitz has a massive clump of Nvidia H100 GPUs to assist its portfolio of AI startups suffer their compute needs , the venture capital firm reassert for the first meter on Wednesday . The program , call Oxygen , allows their portfolio companies to train or work their AI fashion model without talk terms market rate .
a16z ’s Oxygen cluster gives startups some ventilation elbow room , so to speak , to compete against larger technical school troupe — such as Google , Meta , and Microsoft — in building large AI role model . For the past several year , those company have been engaged in a bidding war for GPUs , specially the AI diligence - standard H100s , often acquire deals by promising larger and longer contracts . That entrust a lot of AI startups out to dry , since they may not be able-bodied to take on the big contract necessary to batten GPUs , whereas the OpenAIs and Anthropics of the world can .
“ This get going from a realization that a telephone number of the AI founder we do every day had a common problem : We were in the midriff of a provision crunch , where Nvidia H100 capacity was in short supply , ” said a16z partner Anjney Midha , who help create the Oxygen program , in a podcast copy shared with TechCrunch . “ As startup , they were being deprioritized by the large cloud in favor of enceinte client , which was really tough for them . ”
The base is also a adept way for a16z to attract new startups . Companies building AI models often need unfettered access to massive compute clusters for myopic preparation runs , but then , do n’t involve them once the model are trained . Inferencing AI models typically postulate less compute than preparation , unless of course , usage of an AI model explode . a16z ’s Oxygen cluster offer startups tractability , giving admittance to GPUs when they require it , without a foresighted - terminus commitment to a cloud provider or immense immediate payment spending . or else , they give a16z a stake in their line , in commutation for ( among other things ) depressed cost to charter GPUs .
But a16z is n’t the only investor promising startup beginner GPUs , on top of fiscal capital and guidance , for stakes in their caller . Investing partners Nat Friedman and Daniel Gross offer their startupsaccess to a 4,000 GPU clusterdubbed the Andromeda Cluster , according to Forbes . Y Combinator also offers startup a bunch of GPUsfor breeding runs through partnership with various swarm supplier , most recently , Google Cloud .
An Andreessen Horowitz spokesperson refuse to notice on how large the Oxygen clustering is , however , a16z may have the prominent GPU war chest of any venture house . The Informationreported in July that Oxygen could incorporate more than 20,000 GPUs .
At one pointedness in the podcast , Midha says the Oxygen cluster can save imperativeness on AI inauguration to fundraise at overpriced valuation in ordering to pay their compute account . That ’s another way Oxygen benefits a16z . Instead of investing more in a startup so it can grease one’s palms GPUs from say , Microsoft , a16z can offer the GPUs themselves and invest at a lower monetary value .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
a16z views Oxygen as a core value proffer to its inauguration moving ahead , as long as the AI boom continues as projected . That ’s likely why the startup made the monumental investment need to secure these chip shot .
