Strategy: good customer support, huge scale, low margins.
Good support wins loyalty, helps debugging the system to eliminate systemic problems, and helps against Google (which is really bad at customer service). Huge scale drives down the cost and allows reusing knowledge from user to user, thus each user suffers less while Amazon learns. Low margins keep high-margin competitors away (Microsoft, IBM) and make users happy.
Low margins making this a good fit for Amazon was really a light bulb going off for me. Microsoft, IBM, and even Google hate low margins, but it is inherent to Amazon's other businesses and built into their DNA. Microsoft, specifically, knows that this space is important, but they can't get their heart into it because they know the low margins will eventually have an impact on their other high margin businesses.
I do think there's a ton of room for someone to destroy AWS. Amazon is pretty good. I'm sure they can beat Google and Microsoft. But they're not nearly as fast or smart as they could be with AWS.
I think a small team of sysadmins and programmers could get together with $40MM and wipe the floor with them.
I sincerely doubt that this is true, for a few reasons:
1. Amazon can afford to lose money on AWS for years, while a new company with just $40MM can't.
2. Many of the early adopters using AWS are "rogue developers" within large organizations. A key defense for a strategy like that if you get caught would be that you went with one of the largest, most reliable companies in the industry, as opposed to a small, unproven startup.
3. A small team of very smart, very motivated developers and sysadmins with $40MM would be better off going after a high-margin business where they don't have to compete with three or four of the biggest and most powerful technology companies in the world.
Not saying that there's not room for improvement, or that a startup in this space couldn't do well, but I highly doubt they'd "wipe the floor" with Amazon.
1. $40MM is probably less than Amazon invested initially in AWS. That's enough to buy hardware/bandwidth/datacenter space in quantity for significant savings. It's also enough to afford to lose some percentage of money every year for a while. Definitely enough to prove to investors that you're on to something. This is the strategy Amazon used themselves. Also, I truly doubt Amazon designed or operates their datacenters as cheaply as is possible (whether they think so or not). IT infrastructure is probably a relatively small cost for their retail business. They're also surely quite bloated by legacy choices.
2. Fair enough. They certainly have a great brand name, that helps them. They're going to have an easier time getting rogue developers. That still leaves startups and other small businesses that only care about what's actually better for them. I think it's safer to bet on what's actually better winning than what has a good brand name, otherwise Google would be winning already with App Engine.
3. I agree it'd be easier to play it safe by going after some high margin niche. That doesn't mean it's "better" though. Some people want to do really big things.
I think a great team with great financial backers could absolutely wipe the floor with Amazon. Amazon only looks untouchable because they haven't faced a really serious contender yet.
1. If IT infrastructure was a relatively small cost for their retail business, why on earth would Amazon have built a core competency building infrastructure and services that operate at a scale only a handful of other companies match? A lot of AWS is evolved from internal services that supported their retail business.
2. Given the above, what possible reason do you have for thinking Amazon doesn't strive for efficiency in its datacenters? They spent ~$80M on servers in 2008, and it seems very likely that they spent at least that much this year, so most of their equipment is fairly new. Also, given that power & cooling costs are a significant portion of both capital and operating costs in a data center, the economics for aggressively retiring older equipment is quite strong.
3. Bloated legacy choices: See the above two points. I think the fact that the dominant infrastructure as a service vendor grew out of a f'ing bookstore should tell you that Amazon is not a company that gives bloat and inefficiency a free pass. Furthermore, given the growth they've seen in the last few years, they've had both the opportunity and probably, the necessity, to revisit earlier choices. And by the way, they hired a VP earlier this year who clearly lives & breathes data-center efficiency.
4. Don't underestimate the advantages of scale in this business. Say saving 10% in costs is worth $50K to your startup. Consider how many dollars shaving just 1% in costs is going to be worth to Amazon. It could easily be enough to hire an engineer for a year. For a startup, it might be enough to buy lunch for the company, once a month.
It going to play out in other ways too. Amazon is surely transferring enough data that they can negotiate free peering, something a little startup is going to have a harder time doing. And having a broader customer base likely gives them an advantage in higher average utilization while still keeping a reasonable safety margin.
5. You act as if the Amazon name offers nothing to small business and startups. The relative certainty that Amazon is going to be around tomorrow, next week, next month, and next year is not some foolish superstition. How big is their Amazon bill going to be that it is worth risking data-loss or multiple days of downtime because they were dependent on a startup that ran out of funding?
In the end though,its really not clear to me how you think a startup is going to "absolutely wipe the floor" with Amazon. What's actually going to be better? Clearly, there is room for improvement, but you're talking about taking on the market leader in a growing, and all you talk about is efficiency.
Good support wins loyalty, helps debugging the system to eliminate systemic problems, and helps against Google (which is really bad at customer service). Huge scale drives down the cost and allows reusing knowledge from user to user, thus each user suffers less while Amazon learns. Low margins keep high-margin competitors away (Microsoft, IBM) and make users happy.