GPUs are expensive and setting up the infrastructure to make GPUs work for you properly is complex, making experimentation on cutting-edge models challenging for researchers and ML practitioners. Providing high quality research tooling is one of the most effective ways to improve research productivity of the wider community and Tinker API is one step towards our mission there. Tinker API is built on top of our experimental results on fine-tuning with LoRA: Beta starts and you can join the waitlist today:

Remember when GPUs were for gaming? Then crypto miners took them. Then crypto “died,” and we thought gamers were finally free. Now AI took every GPU on earth and somehow my 2080 still feels like a luxury item. Jensen @nvidia … what happened? 😭

Everyone's watching GPUs. The bottleneck already moved. Joini us live on Monday, May 18th at 8pm ET to walk through where it went — and the three constraints reshaping the AI trade now. 🎙️Set a reminder ↓  Panelists: @Cash_D66 @MikeSchiemer @JohnProv1 @ArchitectIncome @DivStockpile  @DivDegenerates @markusCarter17 @Selling4Premium  @AJacobRant @PatNeville7

THE MAN WHO CONTROLS THE WORLD'S GPUS JUST JOINED X Jensen Huang never had a personal account on X. Today he made one - and used his first ever post to take a side in the biggest fight in tech right now. - His first tweet linked to a policy letter, not a product - 25 companies signed it, arguing America wins by keeping AI open - It follows him telling Axios that banning Chinese models would weaken the US - The post hit 10 million views within hours He who controls the chips just told you who should control the models.

Frontier world models now run on edge GPUs. ⚡ Now openly available, NVIDIA Cosmos 3 Edge brings advanced world-model capabilities to local devices. The 4-billion-parameter omnimodel can understand and generate text, image, video, ambient sound and action for physical AI across robotics, autonomous vehicles and smart infrastructure. 🔗 #SIGGRAPH2026#

General Purpose Robots are the GPUs of Physical Labor The economics of leasing a robot fleet could look similar to that of Neoclouds and data center operators At $50k/Robot/Yr, it would only take 20k robots to reach $1B ARR