Amazon's latest earnings report reveals a stark divide in the AI economy: cloud infrastructure providers are thriving, while AI labs and startups struggle to convince investors of their value. The company's stock surged nearly 10% in after-hours trading after reporting better-than-expected second-quarter results, with net sales climbing 20% and AWS revenue hitting $42 billion for the quarter — a 37% year-over-year increase. Main Developments Amazon's capital expenditures underscore its aggressive bet on AI infrastructure. The company spent $173 billion on property and equipment in the fiscal year ending June 30, up from $107.65 billion the prior year, and raised its 2026 capex forecast from $200 billion to $220 billion. This spending covers GPUs, natural gas turbines, and land for data centers. To fund the buildout, Amazon has begun drawing down cash reserves. The company ended the quarter with $7.6 billion less cash than 12 months ago, marking its first period of negative free cash flow this year. Under normal circumstances, such ballooning expenses would alarm investors, but AWS's revenue growth provides a compelling justification. Read also: Why Apple's Services Growth Stalled Despite Record Hardware Sales The same pattern played out at Microsoft and Google, whose shares also rose after reporting strong cloud revenue. In contrast, Meta — which has significant capex but no clear AI revenue source — saw its stock fall 8% after quarterly earnings, as investors focused on its cash flow crunch and continued spending. Background Amazon's AI strategy extends beyond data centers. The company is investing in custom chips like the Trainium TPU and Arm-based Graviton processor, which don't appear in capex numbers but can meaningfully improve margins for AWS. CEO Andy Jassy said during the Q2 earnings call that "the AI business follows very much the same margin trajectory we saw in the core business before." Jassy also emphasized that AWS and its Bedrock AI platform can succeed without owning a frontier AI model, arguing that "there's not going to be a single model to rule them all." This positions Amazon differently from AI labs like Anthropic, which rely on cloud hosting for their own operations. Notably, Amazon's hosting revenue is directly tied to AI lab spending — in Anthropic's case, the same money flows between the two companies. This interconnection means cloud hosts are not entirely insulated from demand problems facing AI labs and their clients. Why It Matters The divergence between cloud host valuations and AI lab skepticism reveals a critical dynamic. Investors are treating cloud providers as the most reliable part of the AI stack, while doubting the underlying economics for AI labs and startups. However, if AI spending by labs and their clients isn't sustainable, cloud revenue won't remain stable either. The situation echoes what venture capitalist David Cahn framed as the "$3 trillion question": whether enough demand exists to justify the massive infrastructure buildout. Cloud hosts may be several steps removed from that demand problem, but they are not immune to it. The years-long lag between breaking ground on a data center and selling its capacity adds further uncertainty. What's Next Amazon's raised capex forecast to $220 billion for 2026 signals continued aggressive investment. The company will need to demonstrate that demand keeps pace with supply, especially as it dips into cash reserves. Investors will watch upcoming quarters for signs that AWS revenue growth can offset spending, and whether AI labs can sustain their own spending on cloud services. The broader market will likely continue differentiating between cloud hosts with clear revenue streams and AI companies without them. Meta's stock decline after its earnings suggests the skepticism toward AI labs without monetization paths will persist, while cloud providers with proven demand enjoy investor confidence — for now.