Blog Post

3 areas where Amazon Web Services might be vulnerable (if the competition can exploit them)

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!

Amazon(s amzn) Web Services has been the biggest and best cloud computing platform for years, and many would argue it still is. But competition is picking up — led primarily by Google(s goog) and Microsoft(s msft) — leading some to question whether AWS’s old habits will end up helping or hindering the business. One of the critics is Brad Feld, Foundry Group managing partner and TechStars co-founder, who recently explained the situation via metaphor: It’s in a scorpion’s nature to sting even when that’s not in its best interests.

Feld came on the Structure Show this week to flesh out that metaphor and explain the infrastructure trends he’s seen among his firm’s portfolio companies. It’s definitely worth listening to the whole thing for more insights on the economics of computing in the cloud and the challenges startups face when dealing with large companies as customers or partners. But here are some highlights, explaining three areas where Feld thinks AWS and all cloud providers could stand to improve if they want to continue being the fan-favorite choices for IT infrastructure.

Download This Episode

Subscribe in iTunes

The Structure Show RSS Feed

When scale happens, every penny counts

“[A]s the scale continues to increase for a traditional service … your operational excellence as a web service, or whatever you’re doing as a business around that infrastructure, starts to become an important characteristic of your business,” Feld explained. “And depending on on Amazon, or AWS,  for that, you don’t get the cost savings that you think you’re getting because you still have to have the people to manage that. … The argument that you don’t need the people goes away at some point.”

Historically, he said, some companies tired of their AWS bills and experience used to test out SoftLayer(s ibm), Rackspace(s rax) or other options where they could still be in the cloud but also have more control over their resources. Increasingly, though, they’re just installing their own gear in colo data centers and seeing impressive results:

“When you get to the other side of it, it’s just awesome, because all of the things that you’re struggling with in terms of direct control is gone, and on just a raw economics perspective … we’ve seen as much as 20 percentage points margin uplift … which probably translates into 33 percent uplift in gross margin. Almost all of it falls straight to the bottom line.”

However, Feld noted, pricing is one area where one element of Amazon’s nature — to be the lowest-cost provider — is a benefit. That is, if AWS can keep Google from disrupting AWS’s economics too much, too quickly, like it did to Microsoft not so long ago. “It’s a very powerful position to be in to be able to say, ‘We’re just gonna change the pricing game,'” he said. “If you roll back 15 years, Microsoft effectively went to subscription pricing with Office … and then all of a sudden today, it’s like ‘Why am I paying for this stuff?’

Foundry Group's Brad Feld
Brad Feld

Finding the middle ground between self-service and expensive service

Feld acknowledged that AWS support has gotten better over the years, but that the improvement has come at a price, in terms of a percentage of the customer’  total AWS bill, that many startups think is too high. If growing companies think they’re paying top dollar for support, they’d better feel they’re getting top-level service:

[blockquote person=”” attribution=””]”The challenge is when it fails and … AWS might be just fine, but as a customer, something’s wrong with my instantiation on it somewhere, and it’s segmented from me, so it’s a black box, and the service people are working on it, but they get to it and there’s latency. That’s really bad … When you have 140,000 paying customers and you have a problem like that that is really abstracted away from you too far, and you’re paying what you think is premium for support, and there’s latency on it, that’s tough. That doesn’t have to happen very often for a company to start to say, ‘This isn’t working for me.’”[/blockquote]

Startup executives really get talking about exploring alternatives when they have to pay for credits or some sort of recompense after an outage happens. Feld said that has happened “in less than 10, greater than 1 of the 70 companies we’re invested in.”

Managing reputation in a market where reputation matters

As these sorts of complaints add up, Feld said, a company’s reputation starts to suffer. All of a sudden, the company that blazed the trail for so many others, and actually generated emotional responses from users, becomes just another IT vendor users have to deal with:

[blockquote person=”” attribution=””]”Interestingly, some of those companies when they get to that scale, are now pushing harder on Amazon. They’re paying the premium support fees, they’re saying ‘We really, really, really don’t want to do this.’ There’s a sort of inflection point for some of those companies where they justify what’s going on and hang in there, but they’re not delighted customers anymore, they’re not the ones screaming from the top of their lungs the praises of AWS. They’re kind of the customers that you kind of don’t want. They’re the ones that are stuck with you, but they’re not that happy.”[/blockquote]

Larger companies and their small-company partners can also fall prey to the trap of cannibalizing partner ecosystems — something of which AWS has been accused and which Feld, who was involved as an early Microsoft partner in the early ’90s, knows all too well. Maybe it’s because of customer demand or maybe it’s just greed, but large companies undercutting their smaller partners with cheaper, inferior products is not a good look.

“[O]ver time, the startups start saying to each other, ‘Hey, you should be careful, because we had that corp dev conversation with Amazon, and here’s how it went,'” Feld began. “‘And two months later they came out with a competitive product that was inferior and stuff like that, but it didn’t feel very good to us to go spend a whole bunch of time with them and tell them everything we’re doing under the guise of maybe a deeper partnership, and then watch them come out with something that competes with us.'”

4 Responses to “3 areas where Amazon Web Services might be vulnerable (if the competition can exploit them)”

  1. John B.

    The problem with Amazon’s prices is that they don’t always tell the whole story. Their prices don’t include persistent storage, high performance I/O, and network bandwidth. You need to pay extra for things like data transfer, local storage, and support. With a lot of competitors like the company I work for,, these are offered. We like to say that Amazon’s price list is the equivalent of “quoting the price of a car without the wheels included.” If you want the entire car, it can get real pricey really quickly.

  2. Robert Jenkins

    Brad raises some great issues and challenges for cloud service providers in a way that is very easy to understand.

    He’s absolutely right that the vast majority of cloud service providers don’t have great models to deal with customers once they reach scale and support models that go around that. Let’s not forget that AWS (and Google) both have primary business models that are for consumers and are geared up in that way. Business to business isn’t their focus traditionally so it isn’t entirely unsurprising that dealing with customers scaling out and higher touch support models are areas they struggle to deliver on well.

    I believe fundamentally that public cloud can deliver on these shortcomings but it needs the right approach structurally from the service provider with a deep understanding of business customer requirements.

    Remaining a Competitive Option at Scale
    AWS and others don’t address this well in their pricing models for a couple of reasons. Firstly they bundle resources and offer a cookie cutter approach that doesn’t matter when you are small but adds up to a lot of waste and frustration when customers begin to scale out. This issue can be addressed head on through an unbundled resource purchasing model that allows customers to define their own computing. This extends beyond simple virtual machine sizing to more complicated networking set-ups, CPU and hypervisor settings. All of which have efficiency and performance benefits that are worth investing in once a customer hits a decent scale. None of this is possible in a cookie cutter cloud approach.

    Secondly pricing and billing models need to be aligned to support customers as they scale. This means both volume and time discounting models that are transparent to the end user so they can predict and model their costs as they grow. AWS and others lack real, transparent price tiering based on volume which means larger customers don’t see the volume discount benefits they should. Again this is solved by the cloud service provider properly applying cloud economics to their own pricing and billing models that means customers can scale to very large sizes and the public cloud solution remains competitive on a price and feature set basis with an in-house private solution.

    Having support tiers that actually deliver what the customer wants is not easy. This goes beyond simply having the right boxes ticked in terms of service level. In reality as Brad rightly points out this means having the right answers at the right times for customers so they can take appropriate actions to mitigate issues when they happen. This comes down to approach to communications and to organisational structure. Firstly a transparent proactive approach that alerts customers on a very granular basis to service issues is extremely important. This means investment in systems that can deliver this level of communication accurately and on a timely basis. Having global status page isn’t enough for customers.

    Secondly support needs access to real information and an ability to communicate back customer requirements and actions to those within the organisation that can address them. Where does the support team sit? How does support interact with the operations teams managing the deliver of cloud services? Often times a cloud service provider will have a very responsive support team that can frankly do nothing to help the customer and often doesn’t even have timely or accurate information to assist them with either. A cloud service provider needs structurally to be organised in a way that can support demanding customers that run business critical services within their cloud.

    Channel Conflict
    In my opinion a cloud service provider should stick to what they know and not compete with customers. Many leading cloud providers are folding more and more services at the PaaS and SaaS layer into what they offer. They see providers of these services growing rapidly within their clouds, they understand from this business intelligence that it is a valid opportunity and they direct software development resources to offer (an often inferior) version of the same type of service. Customers should be way of any cloud provider that tries to offer competing services to it’s customers. There is a clear conflict of interest there and hoping it won’t happen to you really isn’t a valid strategy. Simply by hosting with a cloud provider that takes such a strategy is communicating valuable information to a potential, very powerful future competitor.

    In my role at CloudSigma these last few years we have worked very hard on addressing these issues through taking a ‘virtual data center’ approach to allowing customers to build out highly unique cloud infrastructure within a public cloud setting, by applying a simple utility style billing model that meters cloud resources in aggregate and applies appropriate time/volume discounting, having a unified support/operations structure to ensure the right support for customers and finally focusing entirely on core IaaS whilst allowing customers to build out powerful PaaS/SaaS on top. It’s good to see more people within the industry and customers realising that these deficiencies from major providers are not trivial however I’d argue giving up the benefits of a public cloud infrastructure footprint isn’t a necessary consequence of their approach. These are vendor specific issues not fundamental issues with public clouds in general.

    Best wishes,


  3. Per NSA PRISM revelations, Microsoft Cloud services & Google Cloud services have Backdoor Access……….Once your Data is in the Cloud, the Vendor OWNS your Data…….

  4. aaronklein14

    “we’ve seen as much as 20 percentage points margin uplift …”

    With all due respect, I strongly suspect that the cost metric cited is not a fair comparison.

    At CloudCheckr, with a sample size of a few thousand deployments, we have found that – on average – simply using our recommendations and purchasing in a more efficient manner enables companies to use the same AWS resources and still reduce their spend by over 35%.

    This dwarfs the 20% he is finding as savings — and it would actually flip the bottom line result. Paying 50% more than you need to and then saying, “see, that is more expensive” is not particularly persuasive.

    Further, we have found that there is even greater saving to be realized when users address issues surrounding idle resources, over-sized resources, and under-utilized resources.

    Beyond that, assuming proper sizing and architecture to benefit from AWS scalability, a proper cost comparison would recognize that a co-lo or on-premise environment would need to be significantly larger — unless you are completely static and expect zero growth or change in your business. AWS allows you to raise capacity to meet needs as they arise. On your own, you need to have already bought the capacity before the need arises.

    Finally, the logic of the argument fails. AWS benefits from scale — they can purchase, house, manage, and replace hardware at an enormous scale and command market power in doing so. Their advantage (relative to a smaller company attempting to replicate processes) could be offset by AWS looking for profit margin, but as we have see with the parent company, Amazon does not run fat margins.

    Aaron Klein
    Take a Free Trial and start saving today!