S&P Global Offerings
Featured Topics
Featured Products
Events
S&P Global Offerings
Featured Topics
Featured Products
Events
S&P Global Offerings
Featured Topics
Featured Products
Events
S&P Global Offerings
Featured Topics
Featured Products
Events
Corporations
Financial Institutions
Banking & Capital Markets
Economy & Finance
Energy Transition & Sustainability
Technology & Innovation
Podcasts & Newsletters
Corporations
Financial Institutions
Banking & Capital Markets
Economy & Finance
Energy Transition & Sustainability
Technology & Innovation
Podcasts & Newsletters
Research — 19 Jan, 2022
By Jean Atelsek
Introduction
With a few minor exceptions, our benchmark public cloud prices remained unchanged in the third quarter of 2021, but still waters run deep. During the quarter, the Cloud Price Index logged over 300,000 SKU changes — SKUs added, SKUs removed, price increases and price decreases — at Amazon Web Services Inc., Microsoft Azure and Google Cloud. Notable price cuts occurred in the categories of developer tools and machine learning, while interesting new services debuted in containers and healthcare.
In cloud computing, disaggregation is driving not only an abundance of complexity but also a great deal of efficiency and innovation. It is happening across all IT domains. First, there was the notion of separating compute from memory and storage — unbundling the "t-shirt sizes" of virtual machines, or VMs. Although predefined combinations of compute, RAM and storage are easy to understand and buy and sell, they lead to waste as buyers provision for peak usage on one dimension and then end up with unused resources on the others. To compensate for fluctuations in demand, the major hyperscaler cloud providers have developed "burst" capacity in compute and storage, whereby performance credits are accrued during slow periods and can then be throttled up to accommodate usage spikes. We are now seeing similar moves in managed services, with providers lowering the buy-in threshold for platforms (in this case, a developer platform — more below). The result is more granularity in metering and billing, higher utilization, lower cost, and ultimately a basis for machine-driven operations.
Alibaba adds new VMs, cuts object storage pricing
For over five years, the 451 Cloud Price Index has tracked the shifting cost of a standard basket of cloud services, collecting pricing and substituting new-generation services and instances to reveal what the market is paying for compute, storage, bandwidth, databases, load balancing and a host of other cloud-based offerings around the world. Benchmarks represent a market-weighted average of rates from the major hyperscalers: Amazon.com Inc.'s AWS, Azure, Google, International Business Machines Corp. and Alibaba Group Holding Ltd.
With few exceptions, benchmark pricing has held steady since 2020's steep (8% to 15%) drop following the rollout of CPUs with better price/performance. Since then, pricing of basic cloud services has remained largely unchanged. In the third quarter of 2021, however, updated Alibaba pricing rippled through to our benchmarks in the Asia-Pacific especially. The company has introduced a new series of VMs in China, Singapore and Indonesia with better price-performance: the g7 (general-purpose), c7 (compute-optimized) and r7 (memory-optimized) instances use Intel Xeon Ice Lake processors; versions based on Advanced Micro Devices Inc.'s EPYC Milan chips are in private preview.
If history is any guide, these options will eventually be available in all of Alibaba's cloud regions. Second, Alibaba reduced the cost of its pay-as-you-go Object Storage Service, or OSS, by 10% to 20% and also introduced "bulk storage" plans that offer a single price for bulk storage purchases in any region. Customers can purchase capacity in one-month, six-month or one-year increments with tiers starting at 40GB and ranging up to 3 petabytes.
We are increasingly seeing the phenomenon of standard pricing across regions. IBM Cloud, for example, now charges a single rate for load balancer usage hours and data processed regardless of data center. Effective September 1, AWS reduced the price of some of its Amazon Textract — a machine-learning service that extracts text, handwriting and data from scanned documents — APIs in eight regions (Mumbai, Seoul, Singapore, Sydney, Canada Central, Frankfurt, London and Paris) to be on par with pricing in its U.S. East (Northern Virginia) pricing, resulting in price drops ranging up to 32%. Oracle Cloud Infrastructure, although not included in our hyperscaler universe, touts its consistent pricing across geographies as a differentiator.
Service changes: Notable price cuts, service additions
Among the big three hyperscalers, all of the greater-than-10% price decreases in the third quarter of 2021 were logged by AWS and Azure. Besides its 50% reduction in ExpressRoute Global Reach data transfer pricing, Azure posted broad-based price cuts in large instances for SAP HANA on Azure (up to 15%), SQL Database (up to 63%), managed hardware security module pools for Key Vault (34%) and the Azure Spring Cloud developer platform. This last change, effective Aug. 1, effectively lowers the hourly buy-in rate for the Azure Spring Cloud service and allows a better fit between resource usage and the amount billed. This was presented as a pricing model change that lowered the base price by 50% to 60% but also reduced the amount of vCPUs and memory included in that price. For the basic (dev/test) edition of Azure Spring Cloud, overage vCPU and memory usage are considerably cheaper than the base price; for the more capable standard (production) edition, overages are priced on par with the base price bundle.
Besides the Textract price reduction mentioned above — and also in the machine learning category — AWS put through a 55% reduction, from $2.50 to $1.125 per hour, in on-demand pricing for the developer edition of Amazon Kendra, the company's machine learning-powered enterprise search service.
Categorywise, new frontiers in cloud services can be discerned by examining net SKU additions by category and then seeing if the momentum holds up when measured as a portion of the total service changes in that category. Especially noteworthy are areas where services are being added globally or in multiple regions by more than one provider, indicating cross-cloud competition. In the third quarter of 2021, a couple of categories stood out:
* Containers: In September, AWS made Amazon EKS Anywhere generally available, following general availability of Amazon ECS Anywhere in May. Introduced at the company's virtual re: Invent conference in 2020, these open source services enable customers to run AWS containers in the environment of their choice, including on-premises and in other clouds. Amazon EKS Anywhere joins similar services from fellow hyperscalers Google with Anthos, and Azure with Arc — all of which require customers to install Kubernetes-based middleware. Meanwhile, Google in September added new CPU allocation controls to its Cloud Run serverless container service. Previously, the only choice was to fire up CPUs only when requests were being processed — a new "always-on" CPU option should enable use cases for executing background tasks or integrating with apps (e.g., OpenTelemetry) that rely on background functionality.
* Healthcare: A noteworthy addition by AWS in the third quarter of 2021 was Amazon HealthLake, which reached general availability in July. This service aims to give healthcare and life sciences companies a managed repository for various types of health data from diverse sources that can be analyzed or used in machine learning models. This is part of a trend toward building out "industry clouds" — platforms that are secure and compliant with regulations in verticals such as financial services, healthcare, retail and manufacturing. We anticipate the move into vertical-specific services becoming a big driver as businesses in these sectors seek to take advantage of cloud capabilities without having to deal with the nitty-gritty complexity of cloud operations — IT-wise, it also effectively lowers the barrier for entry into these markets.
This article was published by S&P Global Market Intelligence and not by S&P Global Ratings, which is a separately managed division of S&P Global.