Geoff Hinton, before he was Geoff Hinton, once asked NVIDIA for a free GPU for his students Alex Krizhevsky and Ilya Sutskever. Nvidia declined
Anecdote reveals NVIDIA once rejected Geoff Hinton's request for a free GPU for his students Alex Krizhevsky and Ilya Sutskever
Story Overview
Years before AlexNet made waves, Geoffrey Hinton emailed Nvidia after promoting their cards to roughly a thousand machine-learning researchers at a conference, asking for one free GPU to support students Alex Krizhevsky and Ilya Sutskever on early neural-net work at Toronto. The company turned the request down, leaving the group to source hardware elsewhere for experiments that later benefited from GTX 580 cards.
Compute shortages hit before the boom
This 2009-era anecdote captures the hardware friction facing neural-net researchers when CUDA was new and skepticism toward deep learning ran high.
Missing pieces stay missing
Exact email date, precise wording of Nvidia's reply, and whether the ask covered one card or more are not established in the reported accounts.
Positive users praise Nvidia for holding its ground against Geoff Hinton's free GPU requests while negative users criticize the approach taken when seeking corporate handouts.
Most Activity
I remember, I was there :-) Fun fact, it wasn’t NeurIPS, it was a room they rented in the same hotel, organizing an unrelated “workshop”, because NeurIPS wouldn’t accept them - Neural Networks weren’t cool. The room was packed, though.
Geoff Hinton, before he was Geoff Hinton, once asked NVIDIA for a free GPU for his students Alex Krizhevsky and Ilya Sutskever. Nvidia declined

(from Stephen Witt's wonderful book on Nvidia and Jensen, 'The Thinking Machine')

@ABiggerSpalash NVidia would send my PhD lab a free GPU every time a new one came out. This was back in 2008, right after CUDA was developed. NVidia was very keen on the lab adding GPU support for our popular chemical simulation package.

@ABiggerSpalash Everyone starts investing in you, once you become a brand big enough....big lesson in here...

@ABiggerSpalash In 2012, he who was 64 at that time, also worked as intern in Google. Interesting story.

@ABiggerSpalash Little known fact : Hinton pioneered the modern 'influencer-run CAC' model before side-questing modern Deep Learning 😏😏😏

@ABiggerSpalash This is not the way to ask if you want to get free stuff from a corporation.

@ABiggerSpalash The scary part is that somewhere today, another Alex or Ilya is probably being ignored for the exact same reason.

@ABiggerSpalash And now Jensen is hand delivering free B300 systems to Karpathy!

@realpaulmahler @ABiggerSpalash What is the way?

@dargason @ABiggerSpalash did the lab add said sim support?

@ABiggerSpalash i was saying the same thing in 2012 :)

@DanAdvantage @ABiggerSpalash No. Memory transfer rate was a killer.
The codebase was developed across multiple labs, pre-git. So it was dozens of semi-independent F77 modules and data structures. It would have required a dedicated rewrite. Ultimately others built MD engines from the ground-up for GPU.

@MParakhin Interesting. In 2004, as UofT student I asked NVIDIA for a free (RMA replacement) GeForce 2 and got it. It was to replace electrically damaged one. I still have the replacement 🫶 Jensen signed it last year.

@MParakhin packed room full of people who knew something nvidia didnt

@Don1Corleoni @ABiggerSpalash Everyone who starts there is an intern. They use the word differently at Google compared to the rest of the world.

@MParakhin that history is actually wild. love that they held their ground.

@MParakhin funny how nvidias "no" ended up shaping everything

@MParakhin ah the classic "workshop that was definitely not a workshop" route
guess packing the room was the ultimate flex

@MParakhin What made you believe back then that neural networks were promising?