Published 21 hours ago

Amazon Cuts Jobs in AGI Unit While Doubling Down on AI

Spending $200 billion on AI while simultaneously laying off the people building it sounds contradictory. At Amazon, it’s apparently just strategy.

Amazon confirmed this week that it has cut roles inside its AGI (artificial general intelligence) unit — the group responsible for foundation models, silicon development, and quantum computing initiatives. The company declined to say how many people were affected, but LinkedIn posts from departing employees suggest the cuts hit teams focused on model customization and post-training.

3 mins read

5 sections

Key Highlights

Amazon cuts AGI and post-training roles while planning $200B in AI capex
Leadership shifts from research focus to production-grade infrastructure execution
Nova’s roadmap becomes the key signal for Amazon’s next AGI moves

What’s Actually Happening Here

This isn’t a retreat from AI. It’s a reallocation.

Amazon has been trimming headcount across the company since late 2022, shedding more than 30,000 roles following a pandemic-era hiring surge. The AGI cuts follow that same pattern: targeted reductions running in parallel with aggressive infrastructure investment.

The company has forecast $200 billion in capital expenditures for the year — up more than 50% from 2025 — and is raising tens of billions in debt to fund its AI build-out. That’s not the posture of a company pulling back.

The Nova Problem

Amazon’s AGI unit made its most visible move in 2024 with the release of Nova, a family of foundation models. But the unit’s own leadership has been candid about where things stand.

Peter DeSantis, who took over the AGI group last December, acknowledged in a recent interview that Amazon’s models “haven’t been at the very frontier for the very largest, most demanding workloads.” The goal, he said, is to get to one of the “most capable intelligent models out there.”

That’s a meaningful gap to close — especially against OpenAI, Google, and Anthropic, all of whom are moving fast.

Leadership Turbulence Adds Context

The cuts don’t happen in a vacuum. In February, Amazon lost David Luan, the head of its AGI lab, who had joined through the acquihire of his startup Adept just a year earlier. Leadership churn at this level, combined with role eliminations in post-training, suggests the unit is being reshaped — not just downsized.

DeSantis stepping in from the cloud infrastructure side signals a possible shift in emphasis: less research-lab energy, more production-grade execution.

What This Means If You’re Watching the AI Tools Space

A few things worth tracking:

Nova’s roadmap becomes more important to watch. If Amazon is sharpening focus, the next model release will say a lot about where they’re actually placing bets.
Post-training cuts are notable. That’s the layer that makes foundation models useful for specific tasks — fine-tuning, RLHF, alignment work. Reducing that capability in-house could mean leaning more on partners or shifting the work elsewhere.
Infrastructure is the real signal. Amazon’s $200B capex commitment dwarfs any headcount reduction. The compute layer — chips, data centers, cloud capacity — is where Amazon appears to believe the durable advantage lives.

The layoffs are real, but the spending tells the actual story. Amazon is betting that winning the AI race is more about infrastructure than headcount — and it’s willing to restructure around that belief even when it’s uncomfortable.