As it has in the past, Amazon.com (AMZN) used the main keynote event at its annual AWS re:Invent conference to unveil a very long list of new services, as well as talk up its competitive standing relative to public cloud rivals such as Microsoft (MSFT) and Alphabet/Google (GOOGL) .
Among other things, new cloud computing, storage, database, machine learning and security services were unveiled for Amazon Web Services (AWS), which remains unmatched among public cloud players not only in terms of its revenue, but also its end-to-end feature set and software and IT services ecosystems. Amazon also revealed a slew of new partnerships, including ones with Verizon (VZ) and international telcos to offer "edge computing" services that enable low-latency applications to be supported on 5G devices.
Here's a look at three main themes from the keynote, which was delivered on Tuesday morning by AWS CEO Andy Jassy and comes ahead of a more developer-focused Thursday keynote from AWS CTO Werner Vogels.
1. Amazon Wanted to Show How Much It's Steadily Improving Popular Services
During the keynote, Jassy noted that AWS now offers four times as many cloud computing instances as it did two years ago, including some that he asserts are unmatched in terms of the computing power they deliver.
He also observed that AWS has added more than 50 new features over the last year for its SageMaker managed service for building and deploying AI/machine learning models -- an integrated developer environment (IDE) known as SageMaker Studio was revealed during the keynote -- and has also been rolling out dozens of new features for its Aurora database (a thorn in Oracle's (ORCL) side) and Redshift data warehouse service.
Much like those old BASF commercials, Amazon's message was that AWS' R&D teams aren't just busy rolling out brand-new services, but are also busy making existing services better. And at a time when rivals are investing heavily to match some of cloud infrastructure (IaaS) and developer platform (PaaS) services that AWS already offers, that's an important message to send.
2. Amazon Played Up the Breadth of AWS's Offerings
Jassy made sure to spend some time extolling the sheer number of cloud computing instances and storage services AWS provides. He also noted the company's analytics services include offerings for interactive querying, big data processing, data warehouses, operational analytics, real-time and data visualization; that it now supports more than a half-dozen internally-developed databases to go with popular third-party and open-source databases; and (following the launch of an innovative new service known as EKS on Fargate) that it also supports four different services for deploying app containers.
You get the idea. Jassy wanted to stress to CIOs and other IT decision-makers that -- even if rivals have highly competitive services in one particular field or another -- AWS has a unique ability to cover every base (or close to it) when an enterprise is considering a cloud migration or project that requires the use of numerous different services.
3. AWS Continues Investing a Lot in Custom Silicon
In line with recent reports, Amazon unveiled a second-generation, ARM-architecture server CPU known as Graviton2 -- it arrives a year after the original Graviton CPU was unveiled -- and claimed that Graviton2-based computing instances will offer 40% better price/performance than existing instances relying on Intel (INTC) and AMD's (AMD) x86 server CPUs.
The company also announced that computing instances supporting its Inferentia ASIC (also unveiled a year ago) are now available, and claimed that the instances deliver a cost advantage of "up to 40%" relative to instances featuring Nvidia's (NVDA) popular Tesla T4 server GPU when performing AI inference (the running of trained deep learning models against real-world activity). In addition, the company showed off AQUA, a hardware-accelerated cache for Redshift data warehouses that leans on a custom ASIC and an FPGA to deliver (per Amazon) up to 10 times better performance when running queries.
The fine print does need to be read with some of these claims. ARM server software support remains a work in progress, and the Graviton2 price/performance claim might not account for computing instances based on AMD's second-gen Epyc server CPUs (Rome), which have been announced but not yet launched. Also, Nvidia, which has put a lot of work into optimizing the performance of its GPUs for popular inference workloads, might disagree with the cost advantage claimed for Inferentia.
Nonetheless, Amazon is clearly taking custom chip development seriously, and betting that it will be a major long-term differentiator. Following the Graviton2 reveal, tech analyst Patrick Moorhead observed that the CPU relies on IP that was only released by ARM in February, and that putting IP into a designed and manufactured CPU this quickly is "unheard of."
Certainly, no one has ever accused Amazon of moving slowly when it comes to businesses and corporate initiatives that it considers priorities.