Amazon has cut jobs in part of the organization it calls artificial general intelligence (AGI). While maintaining its stance that developing large-scale AI models remains a key priority, the company says it is reallocating staff toward initiatives that matter most to customers. Amazon has not disclosed the number of people affected or which teams were targeted, and there is no evidence suggesting a retreat from its AI business. What emerges instead is a picture of Amazon pouring massive sums into computing infrastructure while carefully sorting through research themes and personnel.
Cuts confirmed in part of AGI organization, headcount undisclosed
The job cuts were first reported by Reuters on July 22, 2026, and confirmed by Amazon to The Register the following day, July 23. The company told The Register that it has been developing large-scale AI models for several years and that this remains one of its most important initiatives. At the same time, Amazon acknowledged that it had cut roles in "part of the AGI organization" in order to focus resources on plans that matter most to customers.
However, Amazon has not disclosed how many people were affected or which operations were scaled back. Reuters reported that employees under Adeeb Shanaa, Vice President of AGI Data Services, and Vishal Sharma, Vice President of AGI Information, said online that they were among those affected by the cuts. This is based on what employees themselves have observed and is not a team-by-team breakdown confirmed by Amazon.
| What's being verified | What is known as of July 23 |
|---|---|
| Cuts implemented | Confirmed by Amazon for part of the AGI organization |
| Number of cuts | Undisclosed |
| Affected operations | Undisclosed by Amazon; Reuters reported impact on staff under two managers based on employee posts |
| Impact on product plans | Undisclosed |
Based on this breakdown, what has been confirmed is limited to the fact that cuts took place and the company's own explanation. Extending the interpretation to specific teams or product impacts would require additional disclosure.
Without knowing the headcount, it's impossible to characterize these cuts as an overall retreat from AGI strategy. At the same time, it's clear that even the core organization responsible for model development is not exempt from staffing adjustments. Amazon's stated focus on "initiatives that matter most to customers" suggests that decisions are being made about where to keep staff—whether in basic research, model refinement, or productization.
Amazon's use of the term "AGI" is also the name the company has used internally for a specific team. A December 2025 official announcement referred to it as "the team we've been calling AGI," not as an announcement that artificial general intelligence had been achieved. These latest cuts should also be understood as personnel actions affecting the model development division that this organizational name refers to.
A reorganization that bundled Nova, custom silicon, and quantum computing
Ahead of these cuts, in December 2025, Amazon significantly restructured its AI development chain of command. CEO Andy Jassy consolidated large-scale AI models including Nova, the AGI team, custom silicon such as Graviton and Trainium, and quantum computing under Peter DeSantis. DeSantis now reports directly to the CEO, creating a structure in which models and computing infrastructure fall under the same management accountability.
At the same time, it was announced that Pieter Abbeel would lead the AGI frontier model research team. Rohit Prasad left the company at the end of 2025. The latest job cuts occurred after this new structure—which brings together models, custom silicon, and long-term research—was established. However, because the affected areas remain undisclosed, it's unclear from which part of this integrated structure the staff reductions came.
Amazon has explained that as models grow in scale and capability, they require ever-larger amounts of computing resources, and that the performance and cost of custom silicon will determine competitiveness. The aim is to feed future computing requirements identified through model research back into the silicon roadmap early, while also feeding advances in silicon back into model design. Because chip development takes years, if the capabilities models will need aren't shared early enough, there's a risk of misalignment between available computing resources and research priorities.
Prioritizing this kind of coordination could lead management to concentrate staff on plans that improve models and computing infrastructure together, rather than simply increasing headcount across the board. However, Amazon has not explained which plans it is prioritizing this time, or from which plans it is cutting staff. While the goals of the reorganization can be confirmed, they cannot be directly tied to specific staff reduction targets.
$200 billion in capital spending and 30,000 role cuts
Amazon plans approximately $200 billion in company-wide capital expenditures in 2026. Because this includes not only AI and semiconductors but also robotics and low-earth-orbit satellites, it cannot be treated entirely as investment in AGI. Still, it is clear that the company is prioritizing AI computing infrastructure as its top investment target. In its shareholder letter, the company explained that a significant portion of AWS's planned 2026 capital expenditures is already tied to customer usage contracts, with much of the monetization expected to occur from 2027 to 2028.
Actual spending is also increasing. According to filings with the U.S. Securities and Exchange Commission, cash-based capital expenditures in the first quarter of 2026 totaled $43.2 billion, up from $24.3 billion in the same period the previous year. The main investment targets are the technical infrastructure supporting AWS growth and additional capacity for the logistics network. Here too, AI-specific figures are not broken out separately, but the pattern of securing servers and networking capacity ahead of demand continues.
On the staffing side, the trend is moving in the opposite direction. Amazon announced cuts of approximately 14,000 roles in October 2025 and approximately 16,000 roles in January 2026, bringing the total to roughly 30,000. The number of AGI division cuts in July has not been disclosed separately from these figures.
Capital expenditures and personnel costs are both forms of spending, but they serve different roles. Data centers and custom silicon are assets used over multiple years, and there is no basis for concluding that the salaries cut this time were directly redirected toward capital investment. Amazon is locking in capital in anticipation of future computing demand, while reviewing staff allocation on a much shorter cycle.
Did AI cause the job cuts?
In a 2025 message to employees, Andy Jassy said that the adoption of generative AI and agents would reduce the number of people needed for some current jobs while increasing the number needed for others. He also indicated that efficiency gains from AI are expected to reduce overall corporate headcount over the next few years. Amazon itself has linked AI investment and changes in workforce composition over the long term.
There are also internal examples of using AI to accomplish work with fewer people. According to the 2025 shareholder letter, when overhauling the inference engine for Amazon Bedrock, a task normally estimated to require 40 people working for about a year was instead completed by six engineers using the AI coding service Kiro, in 76 days. This is an example Amazon chose to highlight as a success story, and it does not prove the reasoning behind the AGI division's job cuts.
The reason Amazon gave for the latest cuts was not replacement by AI but a focus on customer-facing initiatives. Peter DeSantis also said in June 2026 that for AI to bring about truly transformative change, its capabilities still need to improve by several more orders of magnitude, and that humans will remain at the center of complex technological innovation for the foreseeable future. Directly linking job cuts to automation would mean assuming a causal relationship the company has not actually stated.
Amazon will report its second-quarter 2026 earnings on July 30. That report should reveal how its $200 billion capital expenditure plan and AI demand are progressing. To properly assess the AGI organization's decisions, however, we still need specific information beyond the earnings report—namely the number of people affected, which operations were involved, and whether the development timelines for Nova and the frontier models will change as a result.