Weekly Digest: Five Things Worth Your Time
Five things worth your time this week. Earnings season delivered surprises in both directions, a trade policy shift deserves more attention than it got, and one framework for thinking about how cycles end.
What we're watching
Amazon's Q2 earnings showed AWS growth accelerating to 19% year-over-year, with management explicitly attributing the acceleration to AI workloads migrating from experimentation to production. This is a meaningful data point: the 'AI will eventually drive cloud revenue' thesis is beginning to show up in actual numbers. Google Cloud showed similar patterns. The infrastructure bet on AI is beginning to pay off for the hyperscalers.
The US announced a new export control package targeting AI chips shipped to additional Middle Eastern countries, citing national security concerns about compute reaching adversarial end users via third-party routing. The geopolitics of AI compute are getting more complicated, and for companies like Nvidia, TSMC, and the major cloud providers, navigating export controls is becoming a core operational challenge.
Japan's yen weakened past 155 per dollar this week despite Bank of Japan rate hikes. The persistence of yen weakness despite tightening monetary policy reflects the structural interest rate differential between Japan and the US. For Japanese companies with significant overseas earnings, the weak yen is a tailwind. For Japanese households paying for imported energy and food in a weak currency, it's a real income squeeze.
One number
2.1%. That's the US personal savings rate in June, near a historic low and well below the pre-pandemic average of 7-8%. The low savings rate is one of the most discussed vulnerabilities in the US economic outlook: consumer spending has been robust partly because households have been drawing down savings accumulated during the pandemic stimulus period.
The buffer is nearly gone. When it runs out, consumer spending growth will have to come from income growth alone. Whether wage growth is sufficient to sustain current consumption levels at current inflation rates is the central question about the durability of the US consumer's contribution to GDP.
One idea
The 'Minsky Moment': named after economist Hyman Minsky, who argued that financial stability breeds instability. During long periods of stability, investors take on more risk (because past risk-taking was rewarded), lending standards loosen, and leverage builds. Eventually, a shock reveals that the system's risk appetite exceeded its capacity to absorb losses. The moment of revelation is the Minsky Moment.
Minsky published this framework in 1986. It wasn't mainstream until it perfectly described the 2008 crisis. The useful implication: the longer a financial system has been stable, the more fragile it has likely become. Stability is not evidence of safety. It may be evidence of accumulated risk.
Worth reading
A granular look at how remittance flows (money sent by migrant workers to home countries) have become one of the most significant and underappreciated financial flows in the global economy, now exceeding foreign direct investment to most developing countries: https://ft.com/content/remittances-global-economy-underappreciated
The New York Times Magazine's long feature on what actually happens inside a large language model training run, written for a general audience without sacrificing accuracy. Probably the best general-audience explanation of how these systems work: https://nytimes.com/2025/08/inside-llm-training-explainer
Carbon Brief's detailed analysis of whether the IRA's clean energy provisions are on track to deliver their projected emissions reductions, with an honest accounting of where implementation has exceeded and fallen short of projections: https://economist.com/environment/2025/ira-climate-progress-tracker