Goldman Sachs CEO David Solomon on AI’s economic impact: ‘There’s very little that we do where we need frontier models’

Goldman Sachs committed $500 million to expand its 10,000 Small Businesses program with AI training, research and access to capital, partnering with Anthropic, Google, Meta and OpenAI. A Babson College survey found 86% of program graduates now use AI, up from 56% in 2024, but just 15% have integrated it into core operations and 77% want more training and resources. CEO David Solomon urged businesses to prioritize practical productivity returns over rushing into costly frontier models; 84% of surveyed AI users reported no workforce change.

Analysis

The investable signal is less about frontier-model demand than distribution: small businesses appear to value low-cost AI embedded in familiar workflows, which favors bundled productivity and advertising platforms over standalone, compute-intensive tools unless those tools demonstrate measurable ROI. That dynamic supports Alphabet and Meta as potential distribution winners, but the reported free access and low current spend make near-term revenue contribution unlikely to be material; treat this as ecosystem positioning, not an earnings catalyst. It also implies upside for software vendors that can package AI into existing subscriptions, while putting pressure on standalone providers to justify incremental fees.

For Goldman Sachs, the program may deepen small-business relationships and generate useful product research, but there is no evidence here that it changes lending volumes, credit quality, or earnings. The commitment’s financial timing and accounting are unspecified. Don’t capitalize the strategic narrative into a GS valuation premium without evidence of customer conversion or returns.

Over 1–3 months, watch for partner data on sustained paid usage, workflow integration, and productivity—not training reach or trial sign-ups. Over 6–18 months, the key test is whether AI expands output and hiring or substitutes labor, and whether lower-cost models are sufficient for most workflows. Contrarian point: adoption rates can look impressive while value capture remains negligible; the integration gap and willingness to pay matter more than headline usage. No standalone trade is warranted on this announcement.

Key Decisions for Investors

• Keep GOOG and META as watchlist beneficiaries of AI distribution, not an event-driven buy: seek evidence of paid conversion, retention, or higher engagement before attributing incremental revenue.
• Avoid extrapolating this program into a GS earnings catalyst. Reassess only if Goldman discloses measurable small-business customer acquisition, lending growth, or program returns; verify the commitment’s timing and expense treatment.
• Track whether SMB AI usage shifts from free chatbots to paid, embedded workflows over the next 1–3 months. If usage rises but paid conversion and core-process integration remain weak, treat the adoption narrative as a monetization warning for standalone AI providers.
• Falsify the distribution thesis if SMBs show durable willingness to pay for standalone frontier-model products, or if embedded AI fails to retain users after trials; do not initiate a pair trade without that evidence.