Comparing Bandwidth Costs of Amazon, Google and Microsoft Cloud Computing

EC2, Azure, GCP Bandwidth Costs

In this article I compare the costs of network bandwidth transferred out of Amazon EC2, Google Cloud Platform, Microsoft Azure and Amazon Lightsail.

Bandwidth costs are one of the most ridiculously expensive components of cloud computing, and there are some serious inconsistencies in the industry, especially with Amazon.

First let’s take a look at two alternative bandwidth sources:

Alternative 1 – Colocation in a Data Center

100Mbps (32.85 TB / Month) @ $950 = $ 28.91 per TB or $0.028 per GB / month

About $0.03 per GB transfer is a typical price you’ll pay at a colocation facility for 100Mbps commitment per month without any negotiation, and it’s very easy to get.

Alternative 2 – Google Fiber for Business

1000Mps (328.5 TB / Month) @ $250 = $0.76 per TB or $0.00076 per GB / month

This is a glimpse into the future of bandwidth pricing and it’s available in about 27 regions in the U.S. – great if you can get it but not everyone can.

Microsoft Azure

1TB / month is $88.65 or $0.09 per GB

This is where things get ugly, that’s 3x higher cost than a colocation facility and 116x higher than Google Fiber.

Google Cloud Platform

1TB / month is $122.88 or $0.12 per GB

This is a real WTF moment, I can’t even comprehend how it’s possible for Google to price their bandwidth this high.  It’s 4x higher cost than colocation, and 162x higher than its own Google Fiber service.

Something is really amiss here given Google’s global reach, existing infrastructure and buying power, Azure’s pricing is ugly but Google’s is much worse.

Amazon EC2

1TB / month is $90 or $0.09 per GB

Again 3x higher cost than colocation, and 116x higher than Google Fiber, and if you have data coming in they charge you another $10 per month, making it $0.10 per GB.

Amazon Lightsail

2TB / month is $10 or $0.005 per GB

At first glance this looks great but it’s the one that really annoys me the most because it’s a giant “screw you” to Amazon’s EC2 customers that are paying 18x more for their bandwidth pricing on the first 2TB.

Amazon Lightsail is giving 2TB away free to their $10 / month plan Lightsail customers, yet they only give 1GB away free per month to their loyal EC2 customers.

Anyone that is running up to 5 t2.micro servers in EC2 and uses pretty much any amount of bandwidth should immediately switch to Lightsail unless they have some compelling reason not to, and I can’t imagine what that compelling reason might be because Lightsail instances can connect with other AWS services.

In fact, it doesn’t really matter what instance size you have.  You should just use Lightsail instances as Internet facing routers and put Cloudflare in front of them, you can have up to 5 Lightsail instances per account.

For $50 / month you can get 10TB of outbound traffic @ $0.005 per GB.


Amazon EC2, Microsoft Azure and Google Gloud Platform are all seriously screwing their customers over when it comes to bandwidth charges.

Every one of the big three has massive buying power yet between them their average bandwidth price is 3.4x higher than colocation facilities.

If you move a significant amount of data you should think twice before moving to the cloud, these bandwidth prices are truly ridiculous and there’s no way they can be justified when compared to colocation facilities.

New Article:
Comparing Performance of GP2 SSD vs NVMe SSD Instance Store Volumes in EC2

Want to disrupt the cloud computing industry?  Give bandwidth away at cost.

Next: Learn About Our Marketing AI called Maxine


    • At the Hurricane Electric data center in Fremont, California you can get 1 Gbps flat rate (no usage fees) and a full cabinet with power for $400/month

        • Yes. It works. They have tons of bandwidth so this isn’t that big of an issue for them. Biggest cost in HE’s Fremont facility is power costs, not space or network capacity. HE has a decent network, but being in a single-homed provider-owned datacenter can put you in certain situations not ideal.

  1. I agree with you that bandwidth pricing is quite frustrating and abusive at times, but I think perhaps you’ve misunderstood the way it’s priced. Consumer and business “line to the premise” bandwidth is priced using statistical usage averages, whereas cloud bandwidth is directly metered. If “line to the premise” users as a whole were in fact utilizing their full link capacities 24x7x365 then you would see the two different cost models converge to provider cost + desired margin – competitive pressure.

  2. You might want to give a try / take a look at 🙂

    No BW charge on dedicated server or Cloud (Public or Private) (except for the APAC zone were BW cost are so expensive, but they only charge 123 USD for a 100Mbps unmetered link, by server. Or you get 3/5/10TB included in your server price).

  3. You have to consider the following. Even though you are only charged for the bandwidth you use, cloud providers need to charge you enough to be able to pay for *all* the bandwidth they have available. And they have a lot of bandwidth available. Enough to cover some very high usage spikes, on several redundant lines, from different providers. Which your collocation setup doesn’t even come close to.

    • OVH has a 11 Tbps network and free Anti-DDoS included, by default, on all their services.. And they don’t charge you a ton of moneys for bandwidth. Even when considering that, nothing truly explain why bandwidth are so expensive on big players like AWS, Google, etc.

  4. Have you considered something like sia..

    Decentralized redundant storage solution.. you should compare these alongside sia.. I think sia will really revolutionize this industry..

  5. You’re leaving out that the big three clouds are highly fault tolerant and redundant on the network layer and have no cap at all, you can scale from 0 to millions of TPS without doing anything for your connectivity.

    You are forgetting the costs of multiple circuits to gain that fault tolerance. You are leaving out the massive costs of support contracts on all your core networking gear. You are leaving out the costs of having personnel to configure and maintain that networking gear.

    With the lens of a few vhosts applied it may seem expensive, but you are getting the highest level of carrier grade service with the big three, not a single pipe.

    Also you are leaving out how easy it is to just flood out a single provider’s feed with a rudimentary DDOS attack, where the feeds from the big three all have immense protections in place up-stream that you don’t even need to think about.

    This was a comparison of apples to space stations.


Please enter your comment!
Please enter your name here