Big Bets on A.I. Open a New Frontier for Chip Start-Ups | Hacker News Reader

Big Bets on A.I. Open a New Frontier for Chip Start-Ups

nytimes.com

54 pointsjoubert7 comments

We should expect that as Moore's law slows down specialized hardware will more and more often become a viable solution. As new process nodes take longer to come online the optimal number of different designs per node goes up. At some point computation will probably end up switching to a different substrate because we're not particularly close to the efficiencies allowed by physics. But I expect quite an interregnum after MOSFETS get as small as they can.

Sure, a few of these companies are going to win. But good luck picking the winners. Meanwhile, Nvidia has the deep pockets to build a custom chip of their own and a well-established ecosystem to drive it. And up until they decided to dictate where you can place their GPUs and to reserve the right to audit your usage of their GPUs and software, I believed they had already won the war before the first shots were fired.

there's a lot of engineering, technical, and cultural debt at Nvidia, while I wouldn't be surprised if they came out on top, I also wouldn't be surprised if they didn't.

Nvidia will keep focusing on GPUs to the detriment of anything else. That's how larger established corporations always work, they have a lot of institutional inertia.

So the thing is that as long as Nvidia tapes out a new architecture every single year, the best you can really do is beat them by a factor of two or so by matching them without the constraint of fixed function graphics transistors along for the ride.

And I think you need to beat them by a factor of 5 to 10 against currently shipping hardware (not last year's model or worse) to get traction. Also you have to deliver perf, not just perf/W. Those two constraints alone will eliminate 80% or so of these companies I suspect.

When I was reading rationale on Broadcom's hostile takeover of Qualcomm, the oft repeated mantra is - semiconductor growth has stagnated. So, now the industry is consolidating.

In which case, what exactly is the growth expected out of these companies?

can someone fix the typo in the headline?

Reply on news.ycombinator.com