3 ms·

Following what your competitor is doing has not historically been a winning strategy in tech, I hope it works out for Google. Samsung with it's bespoke ARM chi

Following what your competitor is doing has not historically been a winning strategy in tech, I hope it works out for Google. Samsung with it's bespoke ARM chips and Apple with its chips make better, more capable, devices than Google can possibly make with off the shelf parts. That combined with having to pay the third party margin on the chips and you don't achieve leadership. But the challenge is that "organically" developing a core competency in semiconductor design is unlikely to succeed. Samsung was a chip company to begin with and Apple bought PA Semi for the expertise. Not that buying a company would necessarily help, after all their track record there is spotty at best.

Rd6n6 5y ago
> Following what your competitor is doing has not historically been a winning strategy in tech There is a long history of tech companies copying competitors and taking over markets by executing better
- ChuckMcM 5y ago
  You had me until you said "executing better" :-) Somewhat more seriously, I'm wondering if there has been a single adjacent market that Google has tried to enter and executed better than any existing player, much less the leader in the space.
  - jsnell 5y ago
    But other than browsers, mobile phone operating systems, ads, email, search, collaboration software, machine translation systems, mapping, ML hardware, internet video, identity providing, web analytics, autonomous driving, and virtual assistants, what products have the Romans ever executed well on?
jsnell 5y ago
This isn't coming out of nowhere, they have been doing custom ASICs for a while already. The article mentions TPUs as an example. A more recent example are YouTube's in-house transcoding chips: https://news.ycombinator.com/item?id=27034627 https://news.ycombinator.com/item?id=27034627
petra 5y ago
If I remember correctly, Google did some work on machine learning for optimizing chip layouts. Maybe they think we're at the stage where ML can significantly help with cpu design?

Rd6n6 5y ago

> Following what your competitor is doing has not historically been a winning strategy in tech There is a long history of tech companies copying competitors and taking over markets by executing better

ChuckMcM 5y ago
You had me until you said "executing better" :-) Somewhat more seriously, I'm wondering if there has been a single adjacent market that Google has tried to enter and executed better than any existing player, much less the leader in the space.
- jsnell 5y ago
  But other than browsers, mobile phone operating systems, ads, email, search, collaboration software, machine translation systems, mapping, ML hardware, internet video, identity providing, web analytics, autonomous driving, and virtual assistants, what products have the Romans ever executed well on?

jsnell 5y ago

This isn't coming out of nowhere, they have been doing custom ASICs for a while already. The article mentions TPUs as an example. A more recent example are YouTube's in-house transcoding chips: https://news.ycombinator.com/item?id=27034627 https://news.ycombinator.com/item?id=27034627

petra 5y ago

If I remember correctly, Google did some work on machine learning for optimizing chip layouts. Maybe they think we're at the stage where ML can significantly help with cpu design?