Goldman Sachs
Pivot complete: exit consumer banking, double down on institutional AI. OneGS 3.0 embeds AI across trading, IB, and asset management. Stock surged 56% in 2025.
Scenarios
Institutional moat (proprietary deal flow, counterparty relationships) is precisely what AI cannot replicate. AI amplifies analyst productivity asymmetrically.
Goldman's 2025 and 2026 earnings growth of 20.8% and 12.6% respectively (consensus) is almost entirely dependent on M&A and capital markets volumes remaining elevated — if dealmaking activity falls 30% in a risk-off environment (as it did in 2022, when GS net earnings fell 48%), AI tooling provides zero revenue cushion. GS AI Assistant, deployed to 10,000 employees initially, improves productivity but does not create a new revenue line — it reduces headcount needs, which is deflationary to the talent-intensive model. Open-source AI models are already closing the capability gap on financial analysis, and boutique advisory firms can now match bulge-bracket research quality at a fraction of the cost, threatening the research-led client relationship that justifies GS's fee premium.
Key Factors to Watch
- ●GS AI Assistant deployed firm-wide; OneGS 3.0 covering all major business lines
- ●Marcus fully wound down; capital redeployed to AI-adjacent private credit
- ●2025 stock +56%; 2026 EPS consensus +12.6% YoY
Score History
| Date | Score | Direction | Note |
|---|---|---|---|
| 2026-03-08 | 7.1 | Positive | Score 7.2→7.1 (formula reweight: sb 0.25→0.15, are 0.20→0.25, md 0.20→0.25, dr 0.20→0.25, aam 0.15→0.10) |
| 2026-03-08 | 7.2 | Positive | Score 7.5->7.2 (are 9->6, dr 5->3, aam 9->8). External research cross-ref: 95% of S1 prospectuses AI-drafted but revenue is advisory/trading (are=6), elite relationships defensible (dr=3), GS AI Assistant >50% adoption (aam=8) |
| 2026-03-08 | 7.5 | Positive | Initial assessment from batch 2 research |
Finance Peers
Last researched: 2026-03-16
This is research and analysis, not financial advice. Scores reflect AI impact potential, not investment recommendations.