Amazon has closed its general artificial intelligence lab in San Francisco, created in December 2024, as part of a restructuring that prioritises applications with immediate customer impact. The company has not disclosed the number of affected employees.
Amazon confirmed on July 22, 2026, the closure of its general artificial intelligence (AGI) lab in San Francisco, a unit that operated for just 18 months since its establishment in December 2024. The company has not detailed how many employees were laid off, but stated in an official announcement that "we have been building large AI models for several years, and it remains one of the most important things we are working on."
The lab was led by David Luan, who left the company in February 2026, less than two years after taking on the role. His departure was an early indication of the changes to come. According to sources from Reuters and CNBC, the cuts mainly affected roles related to post-training, model customisation, and internal functions, but not the development of models themselves.
The AGI unit now reports to Peter DeSantis, following a series of leadership changes that included the prior departure of Rohit Prasad. Amazon emphasised that it is "sharpening our focus on the initiatives that matter most to customers," implying a reallocation of capital and talent from experimental AGI research to applications with clear returns.
Is Amazon abandoning the AI race?
No. The company continues to invest heavily in large AI models, specialised chip development, quantum computing under the broader AGI organisation, and customer-oriented initiatives with measurable impact. What changes is the priority: from pure AGI research to products that generate short-term revenue.
This move comes amid industry questioning whether the $200 billion in capex that Amazon has allocated to AI is generating sufficient returns. The restructuring responds to the pressure to demonstrate tangible ROI, something that investors are increasingly demanding.
As Amazon restructures, its competitors maintain their own approaches. OpenAI remains the benchmark in frontier AI, Google competes in language models and AI search, Anthropic leads in safe and aligned models (with Amazon as a significant investor), and Meta bets on open models. Amazon's advantage lies in its real customer data and immediate use cases in AWS, e-commerce, and logistics, but it faces scepticism about whether it can compete in pure AGI research against AI-native players like OpenAI.
Lessons for startups and founders
This move by Amazon offers three practical lessons for those navigating the AI ecosystem in 2026. The first: prioritise applications over pure research. Even a giant with nearly unlimited resources is pivoting towards initiatives that solve concrete problems with clear value metrics. For a startup, this means avoiding "AGI washing" in the pitch and focusing on what can be offered today.
The second lesson: timing matters more than technological ambition. Amazon's lab lasted 18 months before being closed. For a startup, that is critical time to validate product-market fit in 6-9 months, no more. Investors in 2026 are asking "what problem do you solve today?", not "what could you do in five years?".
The third: talent is an early indicator. David Luan's departure in February was a warning sign. Founders should pay attention to leadership movements in their own organisations and in the market to anticipate strategic changes. As sources from CNBC point out, Amazon is "keeping pace" in AI, but reallocating resources where the impact is more immediate.
For startups working with Amazon Web Services (AWS) or relying on the company's AI infrastructure, this change should not affect them in the short term. AWS AI services, such as SageMaker or Bedrock, continue to operate normally. However, founders seeking investment in AI should prepare for tougher scrutiny regarding return on investment and the path to monetisation.
In summary, the closure of Amazon's AGI lab is not an abandonment of AI, but a sign of maturity in the sector: the era of research without a return date is giving way to the era of applications with measurable impact. For entrepreneurs, the key is to demonstrate value today, not promise it tomorrow.
You may also like
