As the top 10 cloud wars race to seamlessly blend their cloud services with new GenAI capabilities, “tens of thousands” of companies are using Amazon Bedrock to build “high-quality, cost-effective, low-latency, production-level generative AI applications,” Amazon CEO Andy Jassy said.
Given the level of excitement and investment the GenAI revolution has generated among companies, I wouldn't have been surprised to hear Jassy say that Bedrock now has thousands of customers. “Tens of thousands.” It's a sign of incredible momentum.
This is also a big factor in Jassy's assertion that “AI is already seeing billion-dollar revenue growth, but it's still relatively early days.” That means Bedrock, an AWS managed service that helps developers build GenAI apps in conjunction with key underlying models, is likely playing a big role in driving AWS's rapid growth opportunity.
Before I get to Jassy's rather lengthy comments, which fully reflect Bedrock's momentum, here's a little background: Jassy made these comments a few months ago on Amazon's Q1 earnings call, and the big takeaway for me was that AWS revenue hit $25B in the quarter, raising AWS growth to 17.2%.
As part of researching an article I'll post later this week about Amazon's Q2 outlook, I was reviewing Jassy's comments from the Q1 earnings call, when I came across Jassy's point that Bedrock has “tens of thousands” of customers. I'm embarrassed to admit that I didn't realize this when I first reviewed Amazon's Q1 transcript in early May. (And I promise I've taken an appropriate amount of self-blame to make up for this oversight!)
So here are the relevant excerpts from Jassy's Q1 earnings call comments that place the Bedrock surprise in the richer context of AWS' overall AI efforts and accomplishments:
“We're seeing a lot of momentum with people looking at how to run generative AI on AWS. I've said that we're already seeing billion-dollar revenue rates in AI, but it's still relatively early days. And, broadly speaking, I think there are a few things that are driving the growth,” Jassy said.
“First of all, there are a lot of companies still building models. A lot of them are built on AWS, and we expect to see more being built on AWS over time, given the operational performance and security of our company and both the chips that we offer and the chips that NVIDIA offers,” Jassy said.
Contact AI Ecosystem Copilot about this analysis
“While a lot of companies are at the stage where they're spending money on training models, I think what people don't realize is that they spend a lot more on inference than they do on training, because training is only done periodically, whereas prediction and inference are running all the time,” Jassy said.
“There are quite a few companies building generative AI applications that do inference on AWS, and a lot of it has to do with services. The main example we're seeing there is a huge number of companies — tens of thousands of companies — already building on Amazon Bedrock, which has the largest selection of large-scale language models and a set of capabilities that make it much easier to build high-quality, cost-effective, low-latency, production-level generative AI applications,” said Jassy.
“So we think both training and inference will be big drivers for AWS. Additionally, for many enterprises, the models and these generative AI applications will contain some of their most sensitive assets and data. So how they secure those applications will be incredibly important to them.”
Jassy then launched into a direct attack on Microsoft and the security nightmare they are experiencing, which I have documented extensively before, most recently in “Microsoft Security Takes Another Blow as Google Cloud Exposes Microsoft Vulnerabilities” and “Can Satya Nadella Fix Microsoft's Badly Broken Security Culture?”
“Certainly, if you look at what's happened over the last year or two, you can see that not all providers have the same track record,” Jassy said.
“And because we have a big advantage on the AWS side, we're now at a stage where companies are seriously experimenting and actually putting these applications into production, so people want to run their generative AI on AWS.”
Final thoughts
Keep in mind that Jassy's comment about Bedrock having “tens of thousands” of customers was made two months ago, meaning the number is probably even larger by now, possibly by tens of thousands.
