The U.S. Treasury has issued more than $100 billion in tariff refunds since May, exceeding half of the total revenue collected from IEEPA tariffs over the same period. Refunds hit $21.97 billion in May, topping that month’s collections for the first time, and surged to $49.18 billion in June, pushing net customs revenue to negative $25.56 billion. Research from the Federal Reserve Bank of St. Louis found tariffs lifted prices of pharmaceuticals and household utensils by more than 4%. The jump in refunds is undermining the tariffs’ fiscal payoff.

Netflix reported Q2 revenue of $12.6 billion, up 13% year over year, but its revenue growth rate has continued to slow—from 16% in Q1 to 13% this quarter, with Q3 guidance at 12%. The stock fell as much as 9% in after-hours trading. The company raised its expected 2026 content-spending growth to 10% from an 8% average over the past five years and announced its largest-ever quarterly share repurchase of $4.7 billion. Netflix said AI has been used in post-production workflows across 300 titles, though the figures did not change the near-term signal of softer growth.

IBM shares sank 25% on July 14, marking the company’s largest single-day drop in its 115-year history, after preliminary Q2 revenue came in at $17.2 billion, about 3.7% below expectations. The sharp selloff contrasted with blockbuster bank earnings released the same day, including JPMorgan’s $21.2 billion quarterly net income and Goldman Sachs’ $6.4 billion profit. The episode fueled arguments that the AI trade is vulnerable to an “earnings bubble,” where profits prove less durable than they appear, rather than a pure valuation bubble, according to Johns Hopkins economist Steve Hanke. Several banks later adjusted their forecasts and price targets for IBM, including BofA cutting its target to $280 from $330 and UBS maintaining a $236 target while lowering 2026 EPS estimates.

Anthropic has reported to Washington that Alibaba allegedly used API access to distill capabilities from its Claude model, highlighting that current U.S. export controls do little to constrain cloud-native AI. The dispute has helped accelerate legislative efforts such as the Remote Access Security Act, which would seek to bar foreign entities from obtaining sensitive AI compute and model weights through cloud services. While no public company is named, the episode has sharpened attention on AI infrastructure security, cloud export compliance, and protections for frontier models, with potential longer-term regulatory implications for AI chips, cloud platforms, and index constituents.