Hacker News new | past | comments | ask | show | jobs | submit login

I was working at Google when AWS launched in 2006. The internal message was, we don't want to build a competitor. We have the technology to compete in this area, but it is fundamentally a low-margin business, whereas reinvesting in our core business is high-margin, so we should keep our infrastructure to ourselves.

I wouldn't say the Google strategy was wrong per se - their stock has about 10x'd since then. But it's interesting to see how much things change over time.

I feel they entered the business because they AWS showed the industry that it actually has pretty good margins.

However, I think the sentiment is probably true that selling VPS is low-margin over the long-term, but i think the margins come from the other bits: global economies of scale, good infrastructure management & practices, and providing their proven internal technologies (ML/AI, managed distributed services). Long-term, it seems like Google could come out the winner, but only if the other aspects of business are also done well (like sales). It seems like Google Cloud is picking up market share, but still far away from the 3 As (AWS, Azure, Alibaba)

Market share is measured in some very, very funny ways. Like Office 365 being part of their cloud division, including when people download and install office on their computer. If Google was to switch things around, and have Ad-words, Gmail, maps, etc, all pay a small, internal fee to google cloud to host their servers own internal servers, I think the market-share conversation would be very, very different.

Most of internal Google services don't run on Google cloud. Was true as of a couple of years ago and I haven't heard differently from my Google friends recently.

Just reclassify Borg as Google Cloud Internal Edition and boom, everything runs on Google Cloud. This sounds stupid but it's no worse than what MS and IBM have done. Heck, when you are your own customer maybe you could even double-count some revenue...

FWIW... most of Amazon's internal infrastructure was NOT run on AWS as of a couple of years ago. The idea that AWS was built to support Amazon is a myth. In fact, most of their own IaaS was a very traditional data center built on converged infrastructure. This is why Amazon was up and running when AWS would have a large outage.

>> Like Office 365 being part of their cloud division, including when people download and install office on their computer.

It's even sillier than that: Microsoft counts (a significant portion of) boxed copies of Office that you buy in a store as Azure cloud revenue. The justification being that buying a copy Office in a box gets you the right to the cloud-based version rather than the offline one you actually bought. Ultimately it's about flexibility.

And in this case, the flexibility to make your Cloud business whatever percentage of your total company revenue it might take in order to claim the market share numbers you want.

I think you hit the nail on the head with managed services. In 2006 Google probably wasn't thinking much further than EC2 and S3

But I thought Google's first offerings was App Engine? If anything I think they were overambitious

(I work for Google, opinions are my own)

From early convos, I'd say they saw quite a bit further than that...

> I feel they entered the business because they AWS showed the industry that it actually has pretty good margins.

And they stayed because of Azure. Azure was making huge profit when google decided to increase number of data centers. This wouldn't be coincidence, IMO.

My understanding was that market share is: AWS > Azure > GCP > Alibaba

The ad business is almost completely margin, almost any other business is going to be lower margin.

Yep, it reminds me of Dutch Disease [1] in economics where countries that happen upon an enormously profitable natural resource see a decline or an inability to compete internationally in other sectors of their economy.

[1] - https://en.wikipedia.org/wiki/Dutch_disease

Lucking across a local maximum seems to warp peoples' perspective on what to do next.

Of course little else is going to look like a better idea! Every direction is downhill!

But every other local maximum can only be reached by first crossing a valley.

Whoever concluded that the margin is low is probably plain clueless about the enterprise IT, or they were looking at developer-friendly tech gigs or colo-providers, etc, not the IBM Oracle EMC Vmware NetApp etc.

When I joined Amazon in 2012, the message is clearly that AWS competes with IBM Oracle EMV Vmware NetApp. I do not have proof if the original plan was that in the early days. But given my impression on Amazon and AWS' executive team, I'd say very likely enterprise IT has always been Amazon's end game.

I think AWS made a change in strategy between 2006 and 2012 because their early architecture and marketing were totally anti-enterprise.

Naturally you have to start with developer friendly marketing as a new comer, there is no direct to enterprise marketing workable for aws (which will be too expensive and an uphill battle)

Obviously I have zero information about how they came to that conclusion, but it sounds to me like they looked too far into the future.

With basically every major tech company (Amazon, Google, Microsoft, IBM) now having some cloud offering and so many companies switching their on-site infrastructure for the cloud, it seems like it's only a matter of time until compute time becomes a commodity. Once that happens pricing becomes a race to the bottom and margins become much smaller.

> compute time becomes a commodity

The cloud is enormously expensive if people are buying it for "compute time".

But that's the miscalculation: thinking people buy cloud for "compute time". By and large they do not. They buy cloud for flexible provision, robust management, easier deployment, and better monitoring.

And those are differentiated qualities.

I see a very different list of core benefits: more of ops, compliance, procurement is someone else's problem; capital conservation; and it's what the kids use to go fast, familiarity.

Compute being a commodity doesn't matter when storage has significant lock in, with cheaper prices for using the localized storage service of a particular cloud provider.

Even if moving a containerized application across cloud services was as simple as clicking a few buttons, if you're doing any sort of logging, and especially later any computation over those logs, you will be locked in one way or another.

Why does this argument get made so much? Why is it assumed that the market will behave this way?

Probably because "product life-cycle" is MBA 101.

Not sure if this is the original from HBR but it is from 1965: https://hbr.org/1965/11/exploit-the-product-life-cycle

From "Maturity Stage":

> The market maturity stage typically calls for a new kind of emphasis on competing more effectively. The originator is increasingly forced to appeal to the consumer on the basis of price, marginal product differences, or both.

Because we have 100 years of free market capitalist based economic history that it would be unwise to ignore?

Enterprise computing was already pretty mature in the pre-cloud era yet prices were (and still are) very high. So when exactly does this commoditization kick in? 50 years? 100?

Enterprise computing? In what sense? Colocated hosting is indeed a race to the bottom, as are renting of bare metal servers and VPS’s

That's because that was 2006. They were wrong. There is nothing wrong with saying that. Corporations change strategies every 90 days, there is no reason for the infallible cult leader admiration.

Google AppEngine launched in 2008.

Google Cloud Storage launched in 2010

Google has been chasing Amazon since AWS started, and has been accelerating investment.

I thought margin on the cloud business is pretty high...

Margins are high because it's expensive. Cloud is in a big growth curve because it operationalizes capital spend and time to market (key for startups) and they have amazing software stacks which let you dump legacy people and costs.

But once you are there and make the cloud/devops process part of your DNA, you may find that it isn't the optimal place to live from a cost perspective. I moved a workload from a big cloud provider to a private datacenter recently and reduced the costs on an annual basis by like 60%. We did need to make a big capital investment.

There's more to being in a high-margin business than just putting your prices up and watching the money roll in. It depends (among other things) upon what you're selling. My sense is that what Google overlooked was the opportunity to add value.

It is today, but it won't last. As automation tooling gets better the barriers to entry drop and competition kicks in and lowers prices. You can see it already: Aws are constantly lowering their prices. It's not just to be nice, they're competing hard.

Core infrastructure may or may not be low margin (traffic rates are nearly pure margin designed to create a moat and may or may not eventually be challenged by regulators as anti-competitive; it remains to be seen). But the lion's share of cloud margin comes from value-added managed services e.g. AWS Lambda and GCP PubSub, which are exceedingly more expensive than their underlying infrastructure but not nearly so expensive as to justify small companies hiring FTE to manage open-source equivalents. And there aren't nearly enough cloud players to justify competition building out competing implementations for, say, the Lambda API, with the sole exception coming to mind being competing object storage providers implementing the S3 API.

Multi-cloud isn't a myth, but it's pretty close. It requires lots of discipline that most customers won't have. The smart money is on cloud providers improving their margins, not shrinking them.

I promise you this isn't true, maybe over the long term the value add services will be higher margin, but for most of the big players right now the margins are in the core infrastructure aspects like standard storage and compute.

I don't think your statements about (a) where cloud margins come from and (b) Lambda vs VM cost are correct.

> the barriers to entry drop

The barriers for entry on the Cloud will never drop. The Cloud grows more and complex and regional data centers across the globe are impossible for any company without gobs of money to compete with. The margins will drop, but only because the competition between the big 3 has barely even started; even that I believe is a decade away.

This is just history repeating itself.

The same thing happened with the underlying server/architecture/processor providers. For a while it a vibrant field with multiple competitors and high margins. After a while, competition and differentiation drove down margins.

When the margins drop only the best players will be left. Those players will buy out the competition until there's just 2 or 3 big players who will then ratchet up the prices until they reach a stable equilibrium.

And then another layer gets added to this tasty lasagna:

- Processor

- Architecture

- OS

- VirtualizationService

- CloudProvider

And then we start the process of innovation to margin squeeze again...

> The margins will drop, but only because the competition between the big 3 has barely even started; even that I believe is a decade away.

Why? My guess is that the market for Cloud computing is constantly expanding and there isn't a plateau in sight?

I suppose it depends on the amount of lock-in they can achieve, if it's hard to move from one provider to another then that will keep margins higher.

> barriers to entry drop

Are more providers entering into the cloud space? I thought AWS and Azure had this on lock down, with Google a third fiddle.

Are new companies able to enter this space now? I assume the capital investment required for allocating and maintaining tons of data centers at edge nodes that are globally distributed would be too steep.

Arguably the CDNs (Akamai, CloudFlare and friends) have made the capital investment, and are now in the cloud business.

But I don't see anybody competing with AWS/Azure/GCP anytime soon.

Maaaaaybe. I see the CDNs bringing on compute & durablish storage at the edge. And id buy the argument theyre positioning themselves for the next round of “cloud” ala Lambda + DynamoDB.

But I strongly object to the capital characterization. Akamai, the gorilla in the CDN room, to looks to have capital expenditures of $200M or so for hardware/colo/etc. Cloudflare was spending $20-30M IIRC. Conversely Amazon “Cash capital expenditures were $6.7 billion, $10.1 billion, and $11.3 billion in 2016, 2017, and 2018, which primarily reflect additional capacity to support our fulfillment operations and additional investments in support of continued business growth in technology infrastructure (the majority of which is to support AWS).”

Money alone doesnt buy success but the CDN guys dont seem to showing up to the game yet. As youve noted I dont really see anyone else on the trajectory to general competition with the big three.

Disclaimer: Im a principal at AWS and have worked on CloudFront. All my comments are based on my personal reading of publicly available data like 10K & S1 filings.

"Your margin is my opportunity." [1]

[1] https://quoteinvestigator.com/2019/01/13/margin/

You forgot storage and the huge lock-in that storage has.

It would cost $850,000 to move 10PB of data out of a cloud. People aren't going to do it. They aren't going to be able to afford to do it.

AWS Snowball puts 10PB at $350k. They are already paying 200k/month for 10PB, paying 2 months worth to move it doesn't seem outrageous to me.

It is, though it requires significant capital.

And it was widely thought to not be high margin at the time. Cloud prices have not fallen as fast as hardware.

FWIW, many people still think it will become low margin.

It is, but that wasn't anticipated at the time and Amazon didn't reveal its margins until much later.

I believe most GCP services lose money. A few years ago when I left only a handful were profitable. It's probably profitable now overall, but not as wildly as one might think. Hence the recent curb-stomping of Diane Green.

I would say the Google strategy was wrong.

For every AWS API with significant adoption, Google should have implemented it. Make transitioning from AWS to GCE as simple as possible.

I'm really glad they took the opportunity to change the APIs and abstractions around. Some of the ones exposed by AWS are absolutely insane.

Not to mention some of the paradigms: regionalized dashboards/endpoints, the mess that is EC2Classic/VPC, lack of projects for compartmentalization/namespacing ... a lot of those APIs also leak internal implementation details of their products, and trying to match them would be silly.

And they did follow one API that everybody else has: S3. And that one is decent only because it's basically HTTP with some auth headers sprinkled in.

My overall experience with GCloud is much better thanks to these redesigns. AWS just seems like an underdesigned mess of loosely coupled components implemented by siloed teams.

> AWS just seems like an underdesigned mess of loosely coupled components implemented by siloed teams

That was my experience as well. It felt like they never included any UX person throughout any of the decisions. Their naming convention alone always seemed confusing to me, although it does seem to be improving.

I don't follow you. Your suggestion just serves to turn this into a commodity low-margin business. What benefit does google have for doing that?

Google sees Amazon as its biggest and most threatening competitor. Starving your competitor of lots of high-margin revenue is a good thing to do.

If there's margin to be had, the first step is having customers.

AWS was the clear leader, and sure, you can try to attract different customers... but they should have also tried to make it as easy as possible for existing AWS customers to migrate.

Then, you offer your customers additional services that the other guy doesn't offer.

Your question is basically how does hosting a quarter of the internet make money, well Amazon benefits a mil+ an hour in profit off AWS margins... plenty to entice Google and Microsoft to compete.

If we're using stock price as a metric to judge strategy, GOOG's gains have been stomped by AMZN since 2006 by about 8x: https://yhoo.it/30KNk3A

By that chart, GOOG's only about 5x-ed since 2006.

If you compare most companies to one of the best-performing publicly traded company in the history of the world, you'll find they'll get "stomped". Who is the say they wouldn't have lost money and done worse?

If they would have invested in on-demand streaming video, maybe they wouldn't have been "stomped" by Netflix's 60x return since a similar date?

We're looking at the specific decision of whether or not to have invested in cloud services, which the GP poster said the Google execs had looked into and made a decision that was clearly wrong, which should have been obvious even when they made it. AWS helped Netflix build a competitor to Prime Video, but if Amazon didn't have AWS, Netflix would have just used another vendor.

> we should keep our infrastructure to ourselves.

Hilarious, now that we know amazon builds better infrastructure. How they say it? „don’t buy your own hype“!

(I work for Azure, so I don't really have a dog in this fight.)

Google's cloud seems pretty capable, they're just going through teething. They got to live migration extremely fast, and they seem to have nailed BigQuery and Spanner (albeit not the pricing model.) They will get more reliable.

Internally though, Google was (and maybe still is) a decade ahead. With Borg they deployed their entire infrastructure as containers, running at insanely high utilization via Heracles, and built homogeneous and highly-scalable networking. They invented synchronous georeplication with Spanner. From all accounts, it's a paradise.

It just turned out they were too greedy with keeping their innovations to themselves, too egotistical to believe the rest of the world would catch up, and it was too difficult to productize their infra.

For instance, their containers can only rely on API filters for security, so they couldn't bring Borg to the masses, other than cutting it down dramatically to make Kubernetes and requiring it to run on VMs. It looks like Spanner is also priced by the core, which implies it also can't be safely/compliantly commingled with their internal instances.

Still, for a company playing catch-up and needing to completely change their mindset (remember that initially they didn't even offer VMs), they're doing darn well. Especially now that they've pivoted to GKE/big data/ML and targeting multi-cloud or niche clients rather than enterprises.

Google deserves fame for its infra, it just didn't have good business sense when it came to cloud.

> They will get more reliable

The road to higher reliability is a slow one when we are talking about cloud-level 9s. I get the feeling they are still in the early stages of adopting a B2B culture and everything that comes with that.

For example, the June global network outage was caused by a software bug that had global implications. AWS moved away from building software that crossed region boundaries years ago, I would guess because they had an issue like this one. It feels like GCP is many mistakes behind AWS.

Cloud reliability isn't due to genius, it's just software+infrastructure that has been improved again and again in response to new failure modes. I think it will be several years before GCP has reliability that is comparable to what AWS has today.

Compared to their ad business the cloud is lower revenue. Selling people is the real money maker.

I'd say it was wrong. Amazon and Microsoft now make huge amounts of money from their clouds. It can't be that low margin!

But I feel like their real mistake was trying to do things right, rather than compatibly. I think Google Cloud Engine is undeniably a better system than ad-hoc Linux VMs. But nobody really wants to rewrite all their code to work with it.

Google Compute Engine offers ad-hoc VMs. Google (and Amazon) have many offerings.

Yes they do now because they eventually realised that they had to. But for 5 years (2008-2013) Google App Engine was their only offering.

> The internal message was, we don't want to build a competitor.

Which is a fundamentally wrong analysis. Managed compute, network, and storage was going to happen whether or not Google got into the business, helping build Google's competitors. The only difference is that Amazon gets to collect the rent.

Just like Microsoft, which made bad decisions about the Internet but grew anyway so continued to reward its execs, Google completely fumbled in cloud but grew in other places and continued to overcompensate its execs.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact