AnswersPublished23 sources
Why Amazon Thinks Bedrock Could Become as Big as EC2
Jassy’s case for Bedrock is that businesses want one service for choosing among AI models, not a commitment to one provider.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- Jassy’s case for Bedrock is that businesses want one service for choosing among AI models, not a commitment to one provider.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What did Amazon CEO Andy Jassy say on the October 29 third-quarter earnings call about why Bedrock, AWS’s AI-model access service, could bec. Article summary: Jassy’s argument was that Bedrock could become as large as EC2 because enterprises want access to multiple AI models without being locked into one provider—making the platform that delivers those models a potentially eno. Topic tags: general, general web, user generated. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fa
Amazon CEO Andy Jassy says Bedrock could eventually become as large a business for AWS as EC2, its core cloud-computing service. The thesis is that companies will use several AI models—and value a managed service that lets them choose among them—rather than depend on a single model provider. Bedrock’s reported usage growth is striking, but it is evidence of momentum, not proof of future revenue or EC2-scale success. 6
Why Jassy sees a large opportunity
Bedrock gives customers access to foundation models from multiple providers through AWS. Jassy’s argument is that, if businesses use different models for different tasks, the service connecting them to those models could become valuable in its own right. He said he expects six to eight models to be roughly comparable—or capable enough for at least 95% of workloads—within two to three years. That is a forecast, not an established outcome. 6
33
The growth figures he cited show that customers are increasing their use of Bedrock. Fortune 500 spending on the service rose nearly sevenfold year over year, while total token volume grew nearly 17-fold. Jassy also described Bedrock as the fastest-growing multibillion-dollar service in AWS history. 35
Those measures should not be confused with revenue growth: spending by Fortune 500 companies is not total Bedrock revenue, and token volume measures usage rather than dollars earned. They indicate adoption, but don’t tell us whether Bedrock is already profitable at scale or how quickly it could approach EC2’s business size.
Model choice—and the work of deploying AI
The multi-model pitch is central to Bedrock’s case. If customers want to compare providers or use more than one model within an application, a shared access point can help them avoid relying on a single model vendor. Jassy’s prediction that several models will be capable enough for most workloads supports that logic, though the prediction itself does not guarantee that Bedrock will be the service customers choose. 6
33
Amazon also points to security and operational features as part of the enterprise offering. Jassy said Bedrock inputs and outputs are not shared with model providers or used to train their base models. AgentCore, AWS’s infrastructure for production AI agents, is designed to help teams connect agents to systems, secure them, debug them and scale them. 17
The earnings-call transcript reports that tens of thousands of customers use AgentCore and that twice as many agents were deployed with it in Q3 as in the previous three quarters combined. That is another sign of activity around AWS’s AI tools, but agent deployment is not itself evidence of Bedrock revenue at EC2 scale. 46
The adoption constraint: inference costs
Inference is the computing a model uses to respond to a prompt or perform a task. The available earnings-call transcript says inference already accounts for more than half of AWS’s AI revenue, showing that it is a substantial business for AWS today. 17
But cost remains a potential limit on broader use. In a separate October 23 report, Jassy was quoted saying inference was still roughly ten times too expensive for widespread adoption in business software. He argued that Amazon’s own chips could help lower costs. That comment was made before the earnings call, and it describes a challenge to wider adoption—not proof that costs will fall enough to unlock the growth Bedrock would need. 18
Bedrock’s growth sits within a much larger AWS investment
AWS revenue rose 39.3% to about $46 billion in the quarter, its fastest growth in nearly five years. The $184 billion annualized run rate is a calculation based on the quarter’s pace, not revenue AWS had already earned over a full year. Jassy also projected AWS growth of more than 40% in both 2027 and 2028. 4
7
13
Amazon’s overall results were less impressive against analyst expectations: revenue was $201.94 billion versus $202.31 billion expected, and earnings were $1.89 per share versus $1.96 expected. The company reported negative $42.4 billion in trailing-12-month free cash flow, kept its 2026 capital-spending plan at about $220 billion, and said it expected to spend “meaningfully more” in 2027. 1
3
The numbers make the opportunity and the risk clear. Bedrock’s growth supports Jassy’s view that enterprise demand for AI services is increasing, while AWS’s broader growth and spending plans show the scale of Amazon’s investment. But token growth, customer spending and optimistic forecasts are not the same as proof that Bedrock will become as large as EC2. 1
6
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "Why Amazon Thinks Bedrock Could Become as Big as EC2"?
Jassy’s case for Bedrock is that businesses want one service for choosing among AI models, not a commitment to one provider.