3 ms·
> look no further than NVIDIA's 100% ownership of the DL/ML processor space. [Disclaimer: I work for Google, opinions are my own] Google has their TPU that th
by kyrra 9y ago
> look no further than NVIDIA's 100% ownership of the DL/ML processor space.
[Disclaimer: I work for Google, opinions are my own]
Google has their TPU that they are letting others use via Cloud services[0]. It may not be the same as buying hardware, but they are giving Nvidia a run for their money.
https://cloud.google.com/tpu/ https://cloud.google.com/tpu/
- oneshot908 9y agoI would love it if you licensed this technology to the other cloud providers. IMO it would be an awesome embrace and extend play both in terms of HW and in making TensorFlow faster than its competing frameworks. Otherwise, you guys seem locked in an expensive Red Queen's Race with NVIDIA's 1000+ engineer HW team.