Exclusive: Goldman CEO David Solomon on how his dad’s advice shaped his career—and why interns should steal it

Goldman’s CEO David Solomon told 2026 interns to “go slow” and use AI models to stress-test views rather than treat them as oracles, while emphasizing that AI lacks human judgment about whether something makes sense. The remarks coincided with record Q2 results: net revenues of $20.34B and EPS of $20.98, up 92% year-over-year. While the AI message is measured (AI can draft ~95% of an S-1 in minutes, but judgment remains human), the overall news is mildly positive given the strong earnings backdrop.

Analysis

For Goldman, the key read-through is margin structure rather than top-line growth: management is implicitly saying AI lowers the cost of producing advice, but it does not commoditize the final judgment step clients pay for. That favors the largest platforms with deep client relationships and balance-sheet credibility, because they can spread AI-driven productivity gains across more mandates without surrendering pricing power. In that setup, the real winner is operating leverage in advisory and markets, not a new revenue line.

Second-order effects are more interesting than the headline. If first-pass work gets automated, analyst and associate productivity rises, but so does the bar for differentiating via distribution, trust, and product breadth—an edge that smaller boutiques and lower-touch competitors may struggle to match. Over the next 1-3 quarters, the market should care most about whether compensation and non-comp expense actually decelerate; over 6-18 months, sustained opex leverage could justify a higher multiple for scale banks if revenue holds.

The contrarian risk is that AI also arms clients: if every CFO, sponsor, and in-house banker can pressure-test ideas faster, some advisory fees and underwriting economics may compress rather than expand. That would show up first in weaker M&A conversion or an unexpected rise in compensation as Goldman has to pay up to retain talent while efficiency gains remain theoretical. The thesis is falsified if the next couple of quarters show no improvement in expense ratio or if capital markets activity weakens enough to offset any AI savings.

Key Decisions for Investors

• Long GS on any post-story weakness over the next 1-2 weeks; view as a quality/operating-leverage exposure into the next earnings print, with upside if the comp ratio trends lower and advisory revenue stays resilient.
• Pair trade: long GS / short KRE for a 3-6 month window. Thesis is that AI and client-relationship economics favor scaled investment banks more than rate-sensitive regionals; exit if bank revenue dispersion narrows or macro credit stress lifts regionals.
• Set an earnings alert on GS for evidence of AI-driven cost savings: non-comp expense growth below revenue growth and/or an improving efficiency ratio would validate the rerating case; absence of improvement for 2 consecutive quarters is a thesis break.
• Avoid chasing AI software beneficiaries on this story alone; the message is more supportive of incumbents that can absorb AI internally than vendors claiming immediate monetization. If buying the AI theme, prefer financials with visible opex leverage over pure-play AI names here.