Google should have never been in the cloud business to start with. It is a race to the bottom, and revenue depends on your ability to scale up staff - a model Google doesn't have in any of it's other businesses.

You should tell AWS about that race to the bottom while they continue to produce a majority of Amazon's profit.

Someone can still win a race to the bottom. Look at Walmart a generation ago in retail.

Advertising is way more profitable than AWS. https://www.forbes.com/sites/jasongoldberg/2022/02/04/amazon...

> While Amazon did not disclose the operating income generated from their ad business, it is almost certainly bigger than the income generated by AWS

Not exactly making a strong case.

That is based on huge assumptions, and it is unclear how the internal billing works (i.e is the Ad business consuming AWS services "free" or is there internal chargebacks happening). Is the AWS Revenue just from external customers using the services.

> is the Ad business consuming AWS services "free" or is there internal chargebacks happening

AWS bills Amazon an internal rate. It’s a closely held secret that few employees know but supposedly it’s better than at-cost.

It is not secret. There is an internal spreadsheet that documents all the prices.

Everyone has access! Everyone I worked with at Amazon knew the prices they were spending for the different services they were using or thinking of using.

How could that work? Don't managers have budgets? Do they get drugged and blindfolded while their departments' AWS costs are deducted every month?

heres a sankey diagram of q3 FY 2022 for Amazon...https://imgur.com/a/OupE8ES. (from https://www.appeconomyinsights.com/p/amazon-day-1)

Ads + Twitch = 9.5B, AWS = 20.5B

Anybody who's not Amazon didn't get into the cloud business for profit: they did it for survival.

Post-AWS launch, writing was on that wall that amortizing your (already large) infrastructure costs across paying customers would be the only way to afford a certain scale and geographic distribution of hardware that was a prerequisite to taxing future value.

AWS was built from the ground up on new infrastructure. There was no reuse of existing Amazon infrastructure except tooling.

The whole "selling excess Q4 capacity" was a myth.

https://www.acquired.fm/episodes/Amazon-web-services goes into it in great detail.

Miscapitalized url, should be [0]. However, the page is mostly metadata pointing at a 2 hour youtube video [1].

[0] https://www.acquired.fm/episodes/amazon-web-services

[1] https://www.youtube.com/watch?v=APvj15_YCqk

Thanks for fixing! Mobile autocorrect fail.

It's a podcast first and foremost so the URL is still good for the direct links in various podcast players.

That doesn’t make sense in the context of google. Their own workloads are still way larger than their public cloud. Arguably their first offering, app engine, was the perfect fit because they shoved it into the cracks in between their other borg jobs. The fit of GCE in their overall platform is much worse.

It does if Google can double their footprint. They announced reduced profitability recently in part due to capex for data centers.

Last year google started warning some internal teams that they had to limit usage. It’s possible they’re running out of compute.

Google has been internally chopping budgets since decades. Is how they keep utilization up. Look up “steamroller” or “ucp” if you don’t know those.

Scale up staff? That’s not the cloud business at all. It’s scale up infrastructure and services independently of staff, 100%. Source: I was there

Nope. Major accounts require tons of dedicated technical resources and hand holding. You might not have been directly involved, but the people side of Cloud is scaling out of control right now - even during "hiring freeze" and layoffs.

Those are all marginally profitable staff billed through post sales. Consulting business sides of tech companies shouldn’t be considered part of the core footprint as they’re profit centers based on staff scale. The core product itself is designed to scale independently of people, because the marginal person doesn’t increase profits while the marginal unit of infrastructure. Sources, I was directly involved.

It’s a bit of both.

Culturally it didn’t get enterprise, e.g. “ditch oracle and move to spanner”.

Technology wise it was not setup for enterprise.

Staff wise it was paying way too high of a total-comp for non revenue roles. TAMs making 400k/year on accounts spending < 5M/year.

However it could have addressed all those. There were people inside advocating for those changes. Eng and Product leadership refused to listen. And the CEO refused to reward business success and instead rewarded product launches od multiple un-successful messaging apps.

What does a company with huge datacenters do to reclaim OpEx from committed compute capacity sitting idle at non-peak periods, other than building an IaaS with idle-load-consuming services (e.g. spot instances, cloud functions, other short-lived workloads)? Do you expect them to mine Bitcoin?

This is a completely false take. AWS was built from the ground up as separate infrastructure.

What do you think is going to happen at peak of Amazon needs it’s own excess capacity at the same time customers need access capacity?

Even today, much of Amazon is not on AWS.

> Even today, much of Amazon is not on AWS.

They’re the same CPUs they just use an internal API to provision and not AWS APIs. I’m pretty sure the internal APIs also just wrap ec2 et al at this point (for backwards compatibility)

I guess I should disclose that I work at AWS. But it’s publicly well documented how the infrastructure of AWS came to be.

Much of the Amazon infrastructure predates AWS and was built on homegrown systems. A lot of wasn’t ported over.

If you want to see what it looks like to have an Amazon product made available to AWS and that wasn’t designed specifically for AWS, look no further than Amazon Connect. There were no publicly available APIS for years to automate anything. Every AWS native service is API first.

AWS was re-built from the ground up. Based on systems Amazon knew it already had and how to do better.

S3 was built from the old image hosting designs, but better.

StepFunctions and SWF before it are re-designed based on the 5 internal workflow engines that existed (and still do exist).

SQS is a re-do of AMP.

DynamoDB is a re-do of Sable.

Redshift is a re-do of their internal partitioning and querying of Oracle databases but on non-Oracle databases. Their new iteration of Redshift is designed to replicate Andes + Dryad/Cradle.

I am aware of the history.

The people responsible for writing and maintaining AWS Services are my coworkers…

> This is a completely false take. AWS was built from the ground up as separate infrastructure.

Your comment above doesn’t seem to reflect that history, so I just wanted to make sure it was stated.

AWS is also entirely built on Amazons builder tools like Apollo and Pipelines. So it’s really unfair to say “AWS was built from the ground up”

How we saying anything different? Of course AWS was built based on what Amazon learned from operating at scale.

But that doesn’t mean they just took a few extra computers they had to spare and start hosting external customers on the same metal used by Amazon retail.

The only product that I know of in recent history that came from Amazon to AWS without many changes (and without any API or automation story) is Amazon Connect.

built on vs built with.

You wouldn't say something is built "on" Github.

Your continuous deployment and continuous delivery toolset isn't really what's running a multi-tenant cloud environment.

That just does not describe google. If you look at their published systems papers, particularly the borg traces and the B4 SDN, their hardware runs essentially flat out 24x7. They don’t need to backfill diurnal or annual capacity

IMO it makes perfect sense for Google to be in cloud. Cloud infrastructure services are a race to the bottom—storage, compute, etc. You get people into your ecosystem and upsell them on SaaS offerings that integrate with your IaaS offerings. You have to be able to scale without growing staff linearly, which is already Google’s model (although Apple and Amazon are better at it).

Google’s main problem IMO is the lack of customer focus. The products are good, the experience is not as good.

Google doesn't have the ability to speak with a customer at ALL. If there is a problem, basically the only recourse you have is to flail about on HN, Reddit or Twitter and hope that you are noticed by a random Google employee who feels sympathy for your plight and decides to try and help you.

The ONLY official channel they have where you can talk to someone that I am aware of, is sales and restricted to sales only.

This was probably a conscious decision they made, derp-derp if we eliminate everything we build value for the shareholder, except for the fact they didn't think it through all the way in the sense that they have competitors. Microsoft is also bad, but you can actually eventually get to an MS employee if there is a problem and MS does have official employees on Reddit to keep an eye on things, etc. So, if I was a business, I'd look to Microsoft first, NOT GOOGLE, because Google can't be trusted because they are just an automated chatbot with absolutely zero support.

I do use Google services, but I'd be frightened to rely on them for anything serious.

Once I had an issue with a Google App Store (I paid for some app expensive subscription but did not get access) payment and got a pretty quick and efficient customer support. They were very polite, checked the issue, explained what happened, and offered some solution. Perharps they provide better support in my country due to customer protection laws or I just got lucky.

Google Ads has full fledged customer service. I’m tempted to mention everyone’s favorite saying here, but won’t :)

When Google released Anthos, we were in the market for something along those lines for Kubernetes on our own hardware (policy requirements). Talked to our reps, asked for some quotes … nothing. Tried again. Nothing. Tried again. Nothing. This left everyone wondering what it'd be like if we (as an existing GCP customer) needed support for any reason.

The only company I've ever dealt with which was worse at taking your money was IBM. In both cases, it was eerily similar: you could get a room full of people to do a demo but the salespeople seemed to assume their reputation meant success was inevitable and they didn't have to actually do anything like … eeew … work to close deals. In neither case were we the biggest deal in town but also far from the smallest, and in both cases the experience meant they missed out on larger sales later.

I think you may be misreading the situation. When Anthos was announced, it didn't exist yet. They didn't have anything to sell you and you weren't important enough to be trusted with that knowledge. In this regard, your comparison was probably apt (especially if you were inquiring about Watson).

Released, not announced - they actually contacted us about it and then ghosted when we asked for a quote. At the very least I’d have expected a simple “we’ll get back to you”.

IBM was similar: we were trying to buy Opteron servers, let our rep know, had a bunch of people show up, ended with them saying they’d send us a quote, no reply to multiple requests. Maybe they weren’t interested in mere 6 figure purchases but they could have said so.

Wrong Amit. You are thinking about the google search Amit. That’s a different google exec.