---
title: "\"Cloud King\" has fallen behind? Amazon's \"AI strategy\": \"low cost\" is the core, not competing with cutting-edge models, focusing on self-produced chips and customized models"
description: "Amazon's new AI head Peter DeSantis bluntly stated that \"AI has a cost problem.\" He plans to build cost barriers relying on self-developed chips to provide customized solutions at 50% lower inference "
type: "news"
locale: "en"
url: "https://longbridge.com/en/news/277354356.md"
published_at: "2026-03-01T12:18:28.000Z"
---

# "Cloud King" has fallen behind? Amazon's "AI strategy": "low cost" is the core, not competing with cutting-edge models, focusing on self-produced chips and customized models

> Amazon's new AI head Peter DeSantis bluntly stated that "AI has a cost problem." He plans to build cost barriers relying on self-developed chips to provide customized solutions at 50% lower inference costs than competitors. This shift comes as its AI business faces pressure: the flagship model Nova's benchmark testing lags behind, the core technology team is in turmoil, and AGI lab head David Luan left this week. Despite Amazon's plan to invest $200 billion to enhance AI infrastructure, its stock price has still fallen 14% since January

Amazon is attempting to reshape its competitive position in the artificial intelligence field with a low-cost strategy. After falling behind competitors with its flagship AI model, the e-commerce and cloud computing giant appointed a new head of AI and established cost-effectiveness as a core competitive weapon.

According to The Wall Street Journal, Amazon's new head of the AI department, Peter DeSantis, candidly stated during his first public appearance after taking office: "AI has a cost problem." He believes that **to truly achieve a comprehensive transformation in artificial intelligence, the existing cost structure must undergo fundamental adjustments.**

DeSantis's **core strategy relies on Amazon's self-developed chips to develop generative AI models at costs lower than competitors, focusing on providing customized, cost-effective solutions for enterprise clients.** This approach sharply contrasts with the "high-profile route" taken by competitors like OpenAI and Anthropic, who release cutting-edge models every few months.

He emphasized that Amazon's goal is not to chase the speed of model iteration but to provide customers with a more cost-effective way to meet their AI needs while keeping technology updated. This shift comes at a time when Amazon's AI business is under pressure. **Its flagship model, Nova, has fallen behind competitors in independent benchmark tests, while the core technology team has experienced turmoil**: former AI scientist Rohit Prasad left in December last year, and this week, AGI lab head David Luan also announced his departure.

The pressure at the capital level is also intensifying. **Amazon plans to invest $200 billion in capital expenditures this year, with most of it directed towards AI infrastructure, raising concerns among some investors about the speed of cash consumption.** The company's stock price has fallen about 14% since January.

## New Leader Takes Charge, Low-Cost Strategy Becomes Core

Veteran technology executive Peter DeSantis took over the AI department in December last year. DeSantis joined the company in 1998 and has served for nearly 28 years, being a core driver of Amazon Web Services (AWS) infrastructure development, recognized internally for his deep technical background.

Amazon CEO Andy Jassy stated when announcing the appointment that DeSantis has a "track record of solving problems at the boundaries of technological possibilities," and this promotion is seen as an important signal of Amazon's commitment to enhancing its foundational AI capabilities.

He emphasized that **Amazon's self-developed Nova model has entered a mature stage, capable of reducing deployment costs while accelerating development.** This move aims to address enterprise clients' general concerns about the cost-effectiveness of current AI model training and chip investments, as high computing costs are becoming a major barrier to adopting AI services.

## Self-Developed Chips Are Key to Cost Reduction

**This strategy is based on Amazon's self-developed AI chips, Trainium and Inferentia, with the former specifically designed for training models and the latter for querying results.** Amazon stated that **its chips are cheaper because they are tailored for specific tasks, being up to 50% less expensive than similar products from competitors.** \*\* DeSantis stated:

> "If we can build models on our own chips, we can do it at a fraction of the cost of pure AI model providers,"

Based on this, Amazon launched the Nova Forge product, allowing enterprise customers with specific needs to build customized generative AI models without paying for the premium versions of ChatGPT, Claude, or Gemini. Technology consultant and veteran CIO Tim Crawford noted that more and more enterprises are being attracted to specialized models like Nova because they are cheaper, faster, and can be customized for industry-specific tasks such as cybersecurity threat detection. He pointed out that **enterprise CIOs are increasingly focused on the "relationship between results and price," placing greater emphasis on cost-effectiveness.**

The actual use case from Boston-based drug development company Nimbus Therapeutics confirms this logic. The company's director of computational chemistry, Leela Dodda, stated that after testing multiple AI models, the company chose Amazon Nova, partly because it is easier to train, costs less than competitors, and has a faster response time. He noted that one impressive aspect is Nova's low price; in the company's tests, Nova's accuracy was comparable to a version of Anthropic's Claude, but at only one-tenth the price.

## Huge Capital Expenditures Raise Investor Concerns

Amazon's planned capital expenditures of $200 billion this year—roughly equivalent to the total of the past two years—have raised alarms among some investors. Analysts expect that in the first quarter alone, Amazon will consume about $9 billion in cash. The company's stock price has declined since January, reflecting market doubts about whether this spending pace can translate into sufficient new business.

In response, DeSantis referenced Amazon's historical precedents. He stated that in Amazon's early days, critics believed the company could not compete with large chain retailers; similarly, when investing in AWS data centers, analysts were also pessimistic. "I don't think anyone views these two things that way anymore," he said.

Notably, Amazon also announced a $50 billion investment commitment to OpenAI this Friday, demonstrating that while it is betting on its low-cost approach, it is still diversifying its AI ecosystem. In terms of talent competition, Amazon faces significant pressure as well; according to data from market research firm Levels.fyi, the average base salary for Amazon software engineers and research scientists is lower than that of Meta, OpenAI, Apple, and Anthropic, and the company has recently undergone two rounds of layoffs, with approximately 30,000 white-collar employees leaving. DeSantis expressed confidence in the existing team and believes the company can continue to attract top talent

### Related Stocks

- [AMZN.US - Amazon](https://longbridge.com/en/quote/AMZN.US.md)
- [AMZW.US - Roundhill AMZN WeeklyPay ETF](https://longbridge.com/en/quote/AMZW.US.md)
- [AMZD.US - Direxion Daily AMZN Bear 1X Shares](https://longbridge.com/en/quote/AMZD.US.md)
- [AMZZ.US - GraniteShares 2x Long AMZN Daily ETF](https://longbridge.com/en/quote/AMZZ.US.md)
- [AMZU.US - Direxion Daily AMZN Bull 2X Shares](https://longbridge.com/en/quote/AMZU.US.md)

## Related News & Research

| Title | Description | URL |
|-------|-------------|-----|
| AWS 工程师允许一个 AI 工具进行操作，结果导致云部门瘫痪 | 亚马逊的云计算部门 AWS 遭遇了与其 AI 编码工具 Kiro 相关的两次服务中断。12 月份的一次事件导致了 13 小时的停机，原因是 AI 自主决定删除并重新创建一个环境。AWS 员工指出，这些停机是可以预见的，引发了对 AI 自主性 | [Link](https://longbridge.com/en/news/276561043.md) |
| 亚马逊的 AGI 实验室负责人即将离职 | 亚马逊人工通用智能实验室负责人 David Luan 宣布将在本周末离开公司。Luan 于 2024 年 12 月通过收购他的初创公司 Adept 加入亚马逊，并一直负责专注于开发人工智能代理的 AGI 实验室。他的离职是在 AGI 部门重 | [Link](https://longbridge.com/en/news/276791386.md) |
| 股价狂飙四倍后，SK 海力士承诺继续扩大 AI 芯片产能 | 在股价过去一年飙升逾四倍、创下盈利纪录之际，SK 集团董事长崔泰源承诺大幅扩产 HBM“怪兽芯片”，以应对全球 AI 基础设施建设引发的供应短缺。受益于英伟达等客户的强劲需求，SK 海力士 2026 年产能已售罄，盈利预期最高调升至千亿美元 | [Link](https://longbridge.com/en/news/276582791.md) |
| CoreWeave 接近达成 85 亿美元 “AI 芯片 +Meta 合同” 担保融资 | AI 云计算公司 CoreWeave 正在寻求 85 亿美元的债务融资，以巩固其在 AI 基础设施市场的地位。融资将以 AI 芯片和与 Meta 的云服务合同作为抵押，预计利率约为 6%。此次融资有望获得 A-评级，标志着公司信用认可度提升 | [Link](https://longbridge.com/en/news/276915771.md) |
| 英伟达的 “神秘芯片” 背后–推理时代开启 “四大算力新趋势” | 英伟达将整合 LPU 技术推出全新推理芯片，OpenAI 重金跟进，标志 AI 算力主战场从训练切换至推理。申万宏源研究指出，推理时代正催生四大新趋势：CPU 部署场景增多、LPU 专用架构崛起、国产芯片加速突破、算力需求从训练转向海量 T | [Link](https://longbridge.com/en/news/277352378.md) |

---

> **Disclaimer**: This article is for reference only and does not constitute any investment advice.