Amazon is cutting jobs inside its artificial general intelligence organization, the unit that builds the Nova foundation models and houses the company's silicon and quantum computing teams. The company confirmed the layoffs on Wednesday but declined to say how many roles were eliminated or which sub-groups took the hit. The cuts land as Amazon forecasts $200 billion in 2026 capital expenditures, a more than 50% increase from 2025, and raises at least $25 billion in a bond sale to help finance the buildout.
The AGI group is the tip of Amazon's spear against OpenAI, Anthropic, and Google in frontier model development. Its 2024 release of the Nova family was meant to give Amazon Web Services a first-party model stack to sit alongside third-party options in Bedrock. Trimming the group even as the AI budget balloons signals a preference for concentrated bets over broad headcount.
Amazon has now laid off more than 30,000 staffers since October 2025, following the sweeping cuts of that month and a second wave in January 2026. The AGI reductions are part of a series of smaller, targeted rounds that have continued through the spring.
Key facts
- 01Amazon is cutting an undisclosed number of jobs inside its AGI unit, the group behind the Nova foundation models, silicon, and quantum work.
- 02The company has laid off more than 30,000 staffers since October 2025 across sweeping and targeted rounds.
- 03Amazon has forecast $200 billion in 2026 capital expenditures, a more than 50% jump from 2025, largely to fund AI infrastructure.
- 04Amazon is raising at least $25 billion in a bond sale to help fund the AI buildout, with Q2 results due next week.
- 05David Luan, who ran Amazon's AGI lab after joining via the 2024 Adept acquihire, left in February 2026.
The unit has churned at the top as well. Longtime cloud executive Peter DeSantis took over the AGI group in December 2025, replacing Rohit Prasad. Two months later, in February 2026, David Luan left his post running the AGI lab. Luan had joined Amazon in 2024 through the acquihire of his startup Adept, and his exit removed one of the more prominent research names in the org.
DeSantis has been candid about where Nova sits in the competitive stack. In an interview last month, he acknowledged that Amazon's models are not at the frontier for the largest, most demanding workloads, and said the company is working to close that gap and ship one of the most capable models on the market.
The spokesperson framed the layoffs as a sharpening of priorities rather than a retreat, adding that large-model work remains one of the most important things Amazon is working on. CEO Andy Jassy said at a February 26, 2025 event that Amazon is building more than 100 generative AI features across the company, most of which lean on the AGI group's model output either directly or through downstream fine-tunes.
The financial picture makes the cuts more striking, not less. A $200 billion 2026 capex figure puts Amazon at the top of the hyperscaler spending league, and the at-least-$25 billion bond sale is a rare move for a company that has historically funded infrastructure from operating cash flow. The bond issuance signals both the scale of the AI commitment and the pressure on free cash flow that comes with it.
Second-quarter results are due next week, and investors will look for signs of whether Nova and the broader AGI investment are pulling meaningful AWS revenue. Bedrock consumption, Trainium adoption, and the mix between first-party Nova usage and third-party models on the platform are the numbers that matter.
The counterweight is real. Amazon has spent years and billions on in-house models and has still watched Anthropic, in which it is a major investor, become the more frequently cited frontier lab on AWS. DeSantis's own admission that Nova is not at the top end raises the question of whether the AGI group can catch up on a timeline that justifies the capex, or whether Amazon's long-term AI value ends up flowing mostly through partner models running on its silicon.
For Amazon, the layoffs read as a bet that a smaller, more focused AGI team paired with $200 billion of infrastructure spend produces a better return than a sprawling research org. That is a defensible thesis if the company narrows Nova's scope to the workloads AWS customers actually buy, and lets Anthropic carry the frontier-bragging-rights fight. It is a much harder thesis to defend if the next Nova release lands another half-step behind Claude and GPT, because at $200 billion a year the cost of being a permanent number-three model provider is not a rounding error.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
