亚马逊在过去两年加大内部模型、专用芯片和数据中心的投入，并在Anthropic上追加80亿美元，以维持其在AI时代的云端主导权。Matt Garman强调AI正从独立项目转向嵌入式功能，AWS以Bedrock、Nova模型、自治代理与Forge训练平台作为核心卖点，主张其在成本、可靠性与规模化上优于微软与Google。尽管Azure与Google Cloud在ChatGPT和Gemini整合后增速更快，Garman将AWS第三季度优于预期的表现视为趋势反转信号，并强调企业需求已由探索性试验转向大规模部署。
AI在Amazon内部引发显著组织调整，包括14,000人的裁撤与工程流程重构。Garman引用一项内部重写项目，原规划30人需18个月，借助AI由6人在71天内完成，显示出工程效率的数量级提升。不过内部出现环境担忧，超过1,000名员工警告Aggressive AI扩张可能带来生态成本。Garman认为AI代理的效益依赖于任务可理解性，定位其为生产力放大器而非替代品。外部客户也呈现收益，如Reddit利用Forge在数百万审核决策上训练模型，形成可扩展的“social intuition”。
Garman对行业泡沫保持警惕，点名数十亿美元的零代码早期实验室为高风险案例。相对之下，AWS在过去12个月新增3.8 GW基础设施，并宣布最高500亿美元的美国政府AI数据中心投资，声称所有新产能上线后即被市场吸收。Garman认为企业正获得可量化回报，因此无减速迹象，而AWS选择在超级智能竞赛中保持基础设施与企业价值导向，其能否借此继续领跑云市场仍未定。
Amazon has expanded internal foundation models, custom chips, and data centers while investing USD 8 billion in Anthropic to preserve cloud dominance in the AI era. Matt Garman argues that AI has shifted from standalone projects to embedded features, positioning Bedrock, Nova models, autonomous agents, and Forge as cost-efficient, reliable hyperscale offerings superior to Microsoft and Google. Although Azure and Google Cloud accelerated after integrating ChatGPT and Gemini, Garman cites AWS’s stronger-than-expected Q3 results as evidence of a turning tide and claims enterprise demand has moved from experimentation to broad deployment.
AI is driving major internal restructuring, including 14,000 layoffs and engineering workflow redesign. Garman cites an internal rewrite project originally planned for 30 people over 18 months, completed instead by six people in 71 days using AI, indicating order-of-magnitude productivity gains. Environmental concerns surfaced as over 1,000 employees warned that aggressive AI expansion may carry ecological costs. Garman states agent effectiveness depends on task comprehensibility, framing them as productivity amplifiers rather than replacements. External customers also show gains: Reddit used Forge to train on millions of moderation decisions, producing scalable “social intuition.”
Garman remains cautious about industry exuberance, identifying multibillion-dollar zero-code labs as high-risk. By contrast, AWS added 3.8 GW of infrastructure in 12 months and announced up to USD 50 billion in US government AI data-center investments, asserting all new capacity is absorbed immediately. Garman sees measurable enterprise returns with no signs of pullback, positioning AWS as pursuing an infrastructure-centric, value-driven path amid the superintelligence race, though its continued cloud leadership remains uncertain.