Capital One reported Q2 2026 results that exceeded expectations, posting $5.81 in adjusted EPS versus a $4.75 consensus and $15.85B in revenue versus a $15.77B estimate. The outperformance was fueled by stronger non-interest income linked to the Discover integration and by improved credit metrics. Net charge-offs declined in domestic card and consumer banking.

U.S. earnings season accelerates this week, with IBM and ServiceNow (NOW) reporting Wednesday after the close and Intel due Thursday after the close. IBM plunged 25% in a preliminary report that missed revenue and earnings expectations, while NOW is trading around its 50-day moving average near $103 and remains below its 200-day moving average around $128.40. Intel is down more than 30% from a late-June high, with its 50-day moving average near $117 cited as a key level. The S&P 500 has been under pressure for two straight weeks, sharpening investor sensitivity to whether AI-related results deliver.

GE Aerospace fell more than 4% in a single session after reporting better-than-expected second-quarter results and raising its full-year outlook. The stock is now down almost 9% from its record close on July 6. Orders rose 17% year over year, cooling from 87% growth in the first quarter, while the company’s backlog stands at more than $210 billion—more than four years of revenue at the current pace. GE Aerospace is the only remaining industrial company after General Electric spun off GE Vernova in April 2024, following the earlier separation of GE Healthcare.

Morgan Stanley has singled out GE Vernova, Lam Research and United Airlines as stocks it expects to beat second-quarter results and potentially lift full-year profit guidance. The bank said GE Vernova is benefiting from AI-driven demand for energy infrastructure and expects its gas-turbine order book to be sold out through 2030, while Lam Research is seeing improving AI chip-equipment orders. It also cited resilient travel demand and falling oil prices as supportive for United Airlines’ profit outlook, with earnings due July 22 for GE Vernova, July 29 for Lam Research and Wednesday for United Airlines.

HSBC downgraded AstraZeneca (AZN.L) to hold from buy and cut its target price to £137.50 from £165, a reduction of more than 16%. The move follows the failure of a Phase III trial for a heart disease drug, which sent the shares down as much as 9% in a single session and left them down 12% for the week. Analysts said the stock’s premium valuation has long rested on confidence in its late-stage pipeline, and that consecutive R&D setbacks could trigger a reassessment.

Meta shares rose 15% this week, marking their strongest weekly performance in two years and turning the stock up more than 2% for the year. The company released the Muse Image image-generation model and introduced Muse Spark 1.1 for agentic and coding workloads. Meta also confirmed its first in-house AI chip, codenamed Iris, is expected to begin manufacturing in September. Investors read the updates as signs of faster AI execution, tighter cost control and more paths to revenue beyond ads.

U.S. payroll growth slowed sharply in June and prior months were revised lower, prompting markets to trim the implied chance of a September Federal Reserve rate increase to about 55% from more than 60%. Spot gold held around $4,174.66 an ounce, close to a two-week high, while August U.S. gold futures rose 1.5% to $4,186.70. Bullion posted a weekly gain of more than 2%, snapping a four-week losing streak. A firmer dollar added pressure, but shifting rate expectations continued to dominate short-term pricing.

The dollar index DXY slid to a two-week low of 100.9, while USD/JPY stood at 161.57, close to the 40-year low set in 1986. Markets pared expectations for a Federal Reserve rate hike this year after June payrolls growth slowed sharply. At the same time, rising concern over possible action by Japanese authorities has heightened traders’ alertness to near-term dollar softness and sharp swings in the yen.

Palantir CEO Alex Karp criticized the token-based pricing used by U.S. AI labs OpenAI and Anthropic, calling it “completely wrong” as costs rise, in an interview with CNBC. He said enterprises are moving toward more controllable, lower-cost open-weight models. Palantir also said it is expanding its partnership with Nvidia to build custom AI models for U.S. government agencies.