[jmull]

I think there's a lot of doubt whether, in the long run, doing your own hardware makes sense for a cloud provider.

If you work yourself into a performance dead-end, you can spend billions and end up with a chip that isn't particularly competitive. Whereas if your supplier does that, you can switch suppliers.

You could spend $X billion, end up with a chip that isn't particularly competitive and just have to eat those billions.

You bring up airlines... airlines don't build their own airplanes.

[ksec]

Copying a reply [1] I did on HN some time ago. Which should add some perspective on the point being discussed.

AWS are estimated to be ~50% of HyperScalers.

HyperScalers are estimated to be 50% of Server and Cloud Business.

HyperScalers are expanding at a rate faster than other market.

HyperScaler expanding trend are not projected to be slowing down anytime soon.

AWS intends to have all of their own workload and SaaS product running on Graviton / ARM. ( While still providing x86 services to those who needs it )

Google and Microsoft are already gearing up their own ARM offering. Partly confirmed by Marvell's exit of ARM Server.

>The problem is single core Arm performance outside of Apple chips isn’t there.

Cloud computing charges per vCPU. On all current x86 instances, that is one hyper-thread. On AWS Graviton, vCPU = Actual CPU Core. There are plenty of workloads, and large customers like Twitter and Pinterest has tested and shown AWS Graviton 2 vCPU perform better than x86. All while being 30% cheaper. At the end of the day, it is workload / dollars that matters on Cloud computing. And right now in lots of applications Graviton 2 are winning, and in some cases by large margin.

If AWS sell 50% of their services with ARM in 5 years time, that is 25% of Cloud Business Alone. Since it offer a huge competitive advantage Google and Microsoft has no other choice but to join the race. And then there will be enough of a market force for Qualcomm, or may be Marvell to Fab a commodity ARM Server part for the rest of the market. Which is why I was extremely worried about Intel. (Half of) The lucrative Server market is basically gone. ( And I haven't factored in AMD yet ) 5 years in Tech hardware is basically 1-2 cycles. And there is nothing on Intel's roadmap that shown they have the chance to compete apart from marketing and sales tactics. Which still goes a long way if I have to be honest, but not sustainable in long term. It is more of a delaying tactics. Along with a CEO that despite trying very hard, had no experience in market and product business. Luckily that is about to change. Evaluating ARM switch takes time, Software preparation takes time, and more importantly, getting wafer from TSMC takes time as demand from all market are exceeding expectations. But all of them are already in motion, and if these are the kind of response you get from Graviton 2, imagine Graviton 3.

[1] https://news.ycombinator.com/item?id=25808856

[tekno45]

All the cloud providers also USE their huge infrastructure for other purposes as well. So when they do hit, it has a big effect.

make a chip 2% more efficient in your public cloud AND your private cloud, that is also generating money? You've got BIG savings.

[veltas]

Although the difference with airlines is nobody is pooling the wealth of the world into airlines, investors think tech is going to take over everything. Your cloud providers are being valued like they are going to either run the world at some point or get some compensation for being dismantled by international policy. And when your investors think this way, and you make as much money as these companies do, you might as well make your own chips.

[mywittyname]

Plus you have to provide most, or maybe all, of the support for software. Not just making compilers and related dev tools, but also building and testing of pre-compiled binaries for popular OSS.

That's fine for a few major pieces of OSS, and for internal, cloud-only software (Redshift, BigQuery), but I don't think that concept scales well, or is particularly quick to adapt to market shifts.

[mritchie712]

Also, to pile on, for any company that runs a real sales process, Redshift and BigQuery are losing to Snowflake. So the argument for custom chips doesn't seem that compelling.

[zapita]

> You bring up airlines... airlines don't build their own airplanes.

Some of them refurbish and operate their own engines (by far the most complex part of the airplane) instead of going to GE or Rolls Royce.

Delta actually operates their own oil refinery.

So airlines are in fact a good analogy.

[unethical_ban]

It might be the megacorporate equivalent of "strategic assets" for nation-states. Keep a chip division going and spun up, if only to have a core competency as a backstop.