Internal Amazon documents reveal a growing challenge for Amazon Web Services (AWS) as AI startups increasingly bypass its cloud platform in favor of specialized AI tools and GPU-focused “neoclouds.” Once the go-to infrastructure for startups, AWS is now seeing delayed adoption, with founders investing early in AI models from OpenAI and Anthropic, developer platforms like Vercel, and GPU providers such as CoreWeave and Lambda Labs. According to March and July 2024 internal AWS documents, startups are shifting budgets away from traditional compute and storage services toward AI-specific spending, including model inference, training, and API access. The documents, marked “Amazon Confidential,” highlight a “fundamental” shift in startup behavior. While AWS remains dominant in overall cloud market share, its grip on early-stage spending is weakening. Among Y Combinator’s 2024 cohort, only 4.3% used AWS’s Bedrock AI service, despite 88% using OpenAI and 72% using Anthropic. AI startups like Cursor spend over 90% of their cloud budget on external AI APIs and GPU providers, not AWS infrastructure. This trend is reflected in revenue growth: in Q2 2024, Microsoft Azure and Google Cloud both grew over 30%, while AWS grew just 18%. Meanwhile, neocloud providers have seen more than 200% revenue growth, though from a smaller base. CB Insights data shows AWS’s share of leading AI startups dropped to 30% between January 2024 and September 2025, down from 33% in 2022–2024, while Google Cloud rose to 38% and Microsoft Azure to 7%. AWS leadership acknowledges the shift. One document warns that AI spending is less “sticky” than traditional services, allowing startups to switch providers easily. Another notes that AWS is “playing catch up” in AI, particularly in the Bay Area, where its executives struggle to secure speaking slots at venture capital events. The company also faces criticism over pricing, with early-stage startups citing AWS’s higher GPU costs compared to rivals. In response, AWS recently cut prices on its Nvidia GPU instances by 45%. Despite these challenges, AWS maintains it remains the top choice for startups. A spokesperson cited Perplexity, Luma AI, Writer AI, and others as recent AWS adopters, and claimed over 85% of top AI startup lists (CNBC Disruptor 50, Forbes AI 50, etc.) choose AWS. The company also points to its deep partnership with Anthropic, which could generate $5.6 billion in AWS revenue by 2027, according to Morgan Stanley. Still, AWS faces structural hurdles. Its startup team lacks strong venture capital ties, and leadership changes—like the abrupt departure of VP Jon Jones—have raised concerns. The company is now building a data-driven model to identify promising AI-native startups earlier, but the challenge remains: if AWS isn’t involved early, it risks losing customers when they scale. While AWS is not being abandoned, the rise of Cloud 2.0—built on AI models, APIs, and specialized hardware—means startups are building their tech stacks outside AWS from the start. Whether AWS can reassert its dominance depends on its ability to innovate faster, price competitively, and integrate seamlessly into the new AI-first development lifecycle. For now, the cloud giant’s throne is under pressure.