One of the great innovations that drove the success of the hyperscalers and subsequently changed the entire IT landscape was the radical improvement in the efficiency and profitability of service delivery around a few basic components, chiefly compute and storage capacity. Hosting providers have always been more efficient than their customers at delivering infrastructure, and made comfortable operating margins doing so, but the true impact of the hyperscalers has been in making a step change to that efficiency by removing entire components of operations that hadn't occurred to other providers. This included rethinking facilities operations, automating every repeatable datacenter task in software, and standardizing operations and hardware to levels far higher than normal – all driven by software and ongoing improvement to the platform. For users, the attraction was convenience, reliability and on-demand fees; for providers, the surprise was in the eye-popping amounts of money hyperscalers were able to charge relative to the cost of operating basic capacity services.
The result: gross operating margins that far outpaced any erstwhile competitors and, thanks to economies of scale, likely still beat most competitors – even those using congruent operating models and similar levels of efficiency. The hyperscalers continue to use those massive efficiencies to fund ongoing investment in software, expansion and customer outreach, expanding service portfolios well beyond just supplying basic compute and storage capacity. Hosting providers, telcos, SIs and managed service providers are now following suit, and are developing cloud services with these hallmarks of efficiency as they aim to include cloud in their broader portfolios. The Cloud Price Index (CPI) can show where these margins are coming from.
If there is a cloud price war going on, we are at the very beginning. Our conservative evaluation reveals that cloud providers are still making healthy profits on services, and aren't yet slashing gross margins to attract highly price-sensitive end users. However, end users do care about price in that they don't want to get ripped-off; we are still in a value-orientated market. The fact that different types of cloud providers charge differently, and that there is no correlation between market share and price, is testament to this. Enterprises are willing to pay more for security, performance, brand, stability, compliance and a raft of other things, but like in the consumer market, they just want to pay as little as they can for the value they require. The CPI was created not to make the claim that cheapest is best, but to give end users, service providers and others the ability to put services, and the value obtained or provided, in the context of the market.
A public cloud is essentially a private cloud where resources are rented to external users rather than internal users. To deliver a public cloud service, a service provider essentially needs the same colocation, hardware, software and labor resources that would enable a private cloud; what makes it a public cloud is that the service provider slices and dices it and sells it to external parties.
Since the Cloud Price Index provides benchmark prices for public and private cloud services, we are in a position to assess gross margins of a range of cloud services. Gross margins on a product represent the difference between the revenue generated per unit and the cost to produce it, also known as the cost of goods sold (COGS). To assess cloud provider product margins, we can take our revenue to be the average price per virtual machine that an end user would pay to rent the resource, and our COGS to be the average cost per virtual machine to deliver that resource on a private cloud.
What can be stipulated as a COGS expense depends on jurisdiction, but a good rule of thumb is that it is the cost of the production of a unit sold. In the cloud provider world, there are four main expenses related to delivering a unit, be it a GB of object storage or a virtual machine for an hour:
- Colocation – the cost of hosting the infrastructure in a datacenter.
- Hardware – the depreciated cost of the physical infrastructure hardware.
- Software – the depreciated cost (or ongoing cost) of the software used to deliver the resource.
- Labor – the cost of labor needed to deliver the resource with support and management, as defined in the service description.
The Cloud Price Index provides benchmark prices of private cloud services that include all of the above components. The benchmarks are good guides, but are not perfect – our colocation pricing is retail, and we don't include discounts in our averages (although the CPI does hold data on this). However, as a per-hour price, they give us a benchmark COGS. We currently hold data on the cost of commercial and OpenStack clouds, and managed and self-managed private clouds, delivering compute, object storage and block storage.
In terms of revenue, the Cloud Price Index provides benchmark prices of 12 cloud services across four global regions. Our averages are based on 90% of the global IaaS market, and we can further delve into averages based on company type: hyperscaler, telco, regional provider or global provider. These benchmarks provide a basis for determining what each target market might be willing to pay for cloud resources. With both revenue (from public cloud index) and COGS (from private cloud index), we are in a position to make an estimate of profit and gross margin.
For a public cloud provider to make a profit, in general, it must be able to manage its cloud more effectively or efficiently that end users could manage their own. The provider must use tools, automation and processes to manage it efficiently, which means an engineer will be able to manage more resources on average than an end user. The provider must also make sure it has a high utilization, so that as many resources as possible are being used and generating revenue. This can be accomplished with clever scheduling algorithms, advanced pricing models, just-in-time hardware provisioning, custom procurement contracts and even resource sharing partnerships. Cloud Price Index data shows that private cloud can be cheaper than public cloud, but only at scale.
In terms of revenue, the Cloud Price Index provides benchmark prices of 12 cloud services across four global regions. Our averages are based on 90% of the global IaaS market, and we can further delve into averages based on company type: hyperscaler, telco, regional provider or global provider. These benchmarks provide a basis for determining what each target market might be willing to pay for cloud resources. With both revenue (from public cloud index) and COGS (from private cloud index), we are in a position to make an estimate of profit and gross margin.
For a public cloud provider to make a profit, in general, it must be able to manage its cloud more effectively or efficiently that end users could manage their own. The provider must use tools, automation and processes to manage it efficiently, which means an engineer will be able to manage more resources on average than an end user. The provider must also make sure it has a high utilization, so that as many resources as possible are being used and generating revenue. This can be accomplished with clever scheduling algorithms, advanced pricing models, just-in-time hardware provisioning, custom procurement contracts and even resource sharing partnerships. Cloud Price Index data shows that private cloud can be cheaper than public cloud, but only at scale.
Analysis
In a custom study for Hewlett Packard Enterprise focused on its Flexible Capacity program, 451 Research surveyed 500 enterprise end users to assess current infrastructure management metrics of private cloud. Of these, 10 stated that they were managing their private clouds at a labor efficiency rate of over 1,000 VMs per engineer. The average hardware utilization rate across these 10 entities was 66%. These companies represent the best in terms of enterprise infrastructure management – they are using tools, automation, economies of scale and capacity management to optimize their infrastructure. We believe that service providers offering cloud services should at least be able to perform management as well as these enterprises; after all, it is critical for service providers that they optimize to make a profit while staying competitive in the market. For hyperscalers, we anecdotally hear that labor efficiency ratios are in the order of thousands of virtual machines per engineer.
The heat map below shows these benchmark prices, with the utilization threshold showing the point where private cloud beats the average public cloud cost. Darker blue shades show lower costs using public cloud, and darker red shades show lower costs using private cloud.
The heat map below shows these benchmark prices, with the utilization threshold showing the point where private cloud beats the average public cloud cost. Darker blue shades show lower costs using public cloud, and darker red shades show lower costs using private cloud.
As represented with the star in the map above, according to CPI data, at labor efficiency of 1,000 VMs per engineer and 66% utilization, these enterprises are poised to beat public cloud on price regardless of whether they use a commercial orchestration software package, an OpenStack distribution or the OpenStack source. More importantly to this analysis, if a service provider is operating at these levels of utilization and labor efficiency, and is buying technology at list pricing, it is still making a profit if it is selling the resources at the current market price. The CPI public cloud market price for a 1vCPU, 4GB memory, 150GB storage VM is $66.80 – in our gross margin calculation, this represents revenue per unit. If a provider is using a cloud built upon the OpenStack source to deliver the cloud service, the CPI price at these levels of utilization and labor efficiency is $48.51 – this represents our COGS. So our profit is $18.29 ($66.80-$48.51), or a 28% margin. And remember that margin is worst case – most providers are likely to beat our conservative estimates of labor efficiency and utilization, and can get substantial discounts for hardware and software.
Let's be a bit bolder in our analysis. Imagine a telco is offering an IaaS compute service, charged by the hour. The telco wants to make sure it is going to profit against its competitors on this specific service, namely virtual machines. The current CPI benchmark price for a telco-delivered virtual machine (1vCPU, 4GB memory, 150GB storage) is $74.75 per month. In our analysis, this represents the telco's revenue per unit.
The telco has built a cloud on the OpenStack source and is beating our conservative metrics by just 20%, bringing utilization to 80% and labor efficiency to 1,200 VMs per engineer (shown as an X in heat map above). The CPI benchmark price for a self-managed private cloud with these attributes is $35.71 – this represents the provider's COGS. A telco provider operating with these levels of labor efficiency and utilization would be making an estimated profit of $38.10 ($74.75-$35.71) per VM, a gross margin of 51% in the worst case.
Here we are analyzing just the gross margins of a single product. Of course, a provider will sell a range of products, each of which with different revenue and COGS. As such, it is not necessary for a product to have a positive margin; virtual machines, for example, could be loss-leaders, losing money to aid the upsell of other services, such as databases. But we are a ways away from this 'race to the bottom.'
One question specific to this scenario might be how telcos are charging more for virtual machines ($74.75) than hyperscalers ($46.30) and getting away with it? Better network connectivity, an existing installed base and unified support might make a telco premium worth paying. Perhaps the network connectivity benefit is pushing up our COGS relative to our assessment, but remember that network is typically paid as a separate line item in public cloud, so we can't attribute this cost to our virtual machine. However, if the managed service 'wrap' is more than would typically be offered by a hyperscaler, this COGS might be larger due to an engineer managing a smaller number of VMs to deliver this wrap. The CPI benchmarks provide a basis for assessing the value of different options relative to average market prices. The fact that different types of cloud providers sell at different average price points suggests that end users still appreciate value beyond just the cheapest service.
Let's be a bit bolder in our analysis. Imagine a telco is offering an IaaS compute service, charged by the hour. The telco wants to make sure it is going to profit against its competitors on this specific service, namely virtual machines. The current CPI benchmark price for a telco-delivered virtual machine (1vCPU, 4GB memory, 150GB storage) is $74.75 per month. In our analysis, this represents the telco's revenue per unit.
The telco has built a cloud on the OpenStack source and is beating our conservative metrics by just 20%, bringing utilization to 80% and labor efficiency to 1,200 VMs per engineer (shown as an X in heat map above). The CPI benchmark price for a self-managed private cloud with these attributes is $35.71 – this represents the provider's COGS. A telco provider operating with these levels of labor efficiency and utilization would be making an estimated profit of $38.10 ($74.75-$35.71) per VM, a gross margin of 51% in the worst case.
Here we are analyzing just the gross margins of a single product. Of course, a provider will sell a range of products, each of which with different revenue and COGS. As such, it is not necessary for a product to have a positive margin; virtual machines, for example, could be loss-leaders, losing money to aid the upsell of other services, such as databases. But we are a ways away from this 'race to the bottom.'
One question specific to this scenario might be how telcos are charging more for virtual machines ($74.75) than hyperscalers ($46.30) and getting away with it? Better network connectivity, an existing installed base and unified support might make a telco premium worth paying. Perhaps the network connectivity benefit is pushing up our COGS relative to our assessment, but remember that network is typically paid as a separate line item in public cloud, so we can't attribute this cost to our virtual machine. However, if the managed service 'wrap' is more than would typically be offered by a hyperscaler, this COGS might be larger due to an engineer managing a smaller number of VMs to deliver this wrap. The CPI benchmarks provide a basis for assessing the value of different options relative to average market prices. The fact that different types of cloud providers sell at different average price points suggests that end users still appreciate value beyond just the cheapest service.
So-called price wars
If we were truly in a cloud price war, we would expect margins to be at least in single digits. In a real price war, goods commoditize as a result of a lack of differentiation – the only way to win market share is price. But that is not what we're seeing; cloud prices are coming down, but not as fast as many think. The CPI has previously shown that there is no correlation between market share and price, suggesting price isn't the big driver in using a provider.
But price still does matter: In a 451 Research custom study commissioned by Microsoft earlier this year, the biggest reason to change primary provider was price, cited by 34% of respondents. Consumers don't necessarily want the cheapest cloud service, but they don't want to feel ripped off. If there is a cheaper option elsewhere, it appears end users will take it into consideration.
Announcements on price cuts gather attention, and are a great publicity and discussion tool for service providers. We think cloud prices will continue to come down through 2017, and may spread beyond virtual machines into object storage, and perhaps even databases – virtual machines came down 7% globally in 2015, but the cost of our small application only came down 2.4%. The fact that margins are still healthy suggests providers aren't sacrificing huge amounts of gross margin to give such cuts. If they are, it might be a few nickels and dimes here and there, but it's more likely that they are reducing costs through better procurement and management. If we are in a cloud price war, we've yet to see it really get off the ground.
If we were truly in a cloud price war, we would expect margins to be at least in single digits. In a real price war, goods commoditize as a result of a lack of differentiation – the only way to win market share is price. But that is not what we're seeing; cloud prices are coming down, but not as fast as many think. The CPI has previously shown that there is no correlation between market share and price, suggesting price isn't the big driver in using a provider.
But price still does matter: In a 451 Research custom study commissioned by Microsoft earlier this year, the biggest reason to change primary provider was price, cited by 34% of respondents. Consumers don't necessarily want the cheapest cloud service, but they don't want to feel ripped off. If there is a cheaper option elsewhere, it appears end users will take it into consideration.
Announcements on price cuts gather attention, and are a great publicity and discussion tool for service providers. We think cloud prices will continue to come down through 2017, and may spread beyond virtual machines into object storage, and perhaps even databases – virtual machines came down 7% globally in 2015, but the cost of our small application only came down 2.4%. The fact that margins are still healthy suggests providers aren't sacrificing huge amounts of gross margin to give such cuts. If they are, it might be a few nickels and dimes here and there, but it's more likely that they are reducing costs through better procurement and management. If we are in a cloud price war, we've yet to see it really get off the ground.
Owen Rogers
Research Director, Digital Economics
As Research Director, Owen Rogers leads the firm's Digital Economics Unit, which serves to help customers understand the economics behind digital and cloud technologies so they can make informed choices when costing and pricing their own products and services, as well as those from their vendors, suppliers and competitors.
Carl Brooks
Analyst, Service Providers
Carl Brooks is an Analyst for 451 Research's Service Providers Channel, covering cloud computing and the next generation of IT infrastructure.
Speaker Name
Speaker Title
Sed ac purus sit amet nisl tincidunt tincidunt vel at dolor. In ullamcorper nisi risus, quis fringilla nibh mattis ac. Mauris interdum interdum eros, eget tempus lectus aliquet at. Suspendisse convallis suscipit odio, ut varius enim lacinia in. Lorem ipsum dolor sit amet, consectetur adipiscing elit.