Goldman Sachs CEO David Solomon on Running a Bank in the Age of AI
Goldman Sachs CEO David Solomon joins the Odd Lots podcast to discuss AI's impact on banking jobs, arguing that the 'AI job apocalypse' is overblown. He covers Goldman's technology investments, the importance of human relationships in banking, recent major deals including the Alphabet equity raise and SpaceX IPO, and his views on market valuations and the shift toward fewer public companies.
Summary
David Solomon, CEO and Chairman of Goldman Sachs, appears on Bloomberg's Odd Lots podcast to discuss his New York Times op-ed arguing against the narrative of an AI-driven job apocalypse. Solomon frames his optimism around a 5-10 year horizon, acknowledging that while AI will disrupt certain jobs, it will ultimately increase productivity and economic growth rather than cause mass unemployment. He draws on historical parallels, arguing that technological disruption has always shifted rather than eliminated labor demand.
On the specific question of junior banking talent, Solomon describes Goldman's current intake of approximately 2,400-2,500 interns and new hires spanning client-facing, operational, and engineering roles. He argues that AI will not eliminate the need for young bankers but will free them from rote analytical tasks — like the six-hour stock comparison charts he produced manually early in his career — allowing them to spend more time with clients and build relationships. He sees this as a genuine opportunity to reverse a trend where junior bankers became increasingly desk-bound over the past 30-40 years. The key challenge he identifies is how to apprentice young people and ensure they absorb foundational knowledge even when AI provides instant answers.
Solomon reflects extensively on his own career origins, from cold-calling 100 people a day as a Merrill Lynch intern in 1981 to building client relationships at Drexel Burnham and Bear Stearns. He argues that the interpersonal skills developed through these experiences — phone communication, relationship building, emotional intelligence — are not going away and may actually become more valuable as AI commodifies analytical knowledge. He explicitly states that Goldman is not a 'star system' and has a strong culture of data and knowledge sharing, which he believes positions the firm well to leverage AI on clean internal datasets.
On Goldman's specific AI deployment, Solomon describes the '1GS 3.0' initiative, which involves remaking client onboarding, anti-money laundering, and KYC processes. He gives a concrete example: a process that previously involved 3,800 people touching it will be reduced to a few hundred. He is candid that measuring productivity gains broadly across investment banking is difficult, but points to Goldman's 65% revenue growth and 140-145% earnings growth since their January 2020 investor day as evidence of improving productivity over time.
Solomon discusses Goldman's role as sole advisor on the Alphabet/Google equity raise, described as potentially the largest secondary follow-on equity offering ever at approximately $85-90 billion. He argues this is a landmark data point on investor appetite for AI-related capital raises and predicts more large tech companies will issue equity rather than relying solely on debt to fund massive AI capital expenditure plans over the next five years. He also addresses the SpaceX IPO mandate, clarifying that Goldman's role was built over 20 years of relationship-building with Elon Musk and his teams — starting with SolarCity and Tesla — rather than any recent DM pitch, though he confirms he does communicate with Musk via X direct messages.
On market valuations, Solomon pushes back on the 'unprecedented' narrative, noting that 30%+ concentration in the top 10 S&P 500 companies also occurred in the 1920s, 1960s, and late 1990s. However, he distinguishes the current period by noting today's top companies trade at roughly 30x forward earnings versus 40-50x during the dot-com bubble, and actually generate substantial real earnings and cash. He describes the other 490 S&P companies trading at 17-20x forward earnings as potentially attractive if AI-driven efficiency improvements materialize. He acknowledges seeing 'greed' behavior and fear-of-missing-out crowding into a narrow group of stocks.
On private vs. public markets, Solomon argues the shift toward fewer public companies is a structural trend driven by regulatory friction, market structure changes, and the availability of private capital — not a fundamental threat to capitalism. He believes companies going public now are doing so because they genuinely need the capital at a scale that private markets alone cannot efficiently provide. Finally, Solomon discusses AI-generated music, drawing a parallel to how Logic Pro and Ableton already commodified music production, and arguing that AI music tools face unresolved IP compensation issues for artists and that human 'voice' and creative judgment remain essential for producing resonant work.
Key Insights
- Solomon argues the '16% decline in entry-level hiring' statistic is misleading because it only tracks a handful of industries that fit a social narrative of what the 'right' job looks like after college, excluding large swaths of the labor force.
- Solomon claims Goldman Sachs is less susceptible to knowledge-hoarding problems than other firms because it has a deeply embedded culture of sharing and collaboration rather than a star system, which he believes positions it well to leverage AI on internal datasets.
- Solomon argues that AI is most powerful when applied to clean, structured datasets — such as Goldman's 40-year trading history in its SECDB system — but produces 'cockamamie answers' when sent out into the broader internet, citing a Tiger Woods Masters fact error as a concrete example.
- Solomon predicts that the rise of AI will shift Goldman's hiring mix away from engineering talent back toward client-facing roles, reversing a 10-year trend where engineering headcount grew materially relative to operational and client talent.
- Solomon described Goldman's '1GS 3.0' operational reengineering as a measurable productivity example: a client onboarding/AML/KYC process that previously involved 3,800 people touching it will be reduced to a few hundred people.
- Solomon argues that the Alphabet $85-90 billion equity raise is the 'first concrete tangible data point on investor demand at this scale' and predicts other large tech companies will follow by issuing equity rather than relying solely on debt for AI capital expenditure over the next five years.
- Solomon states that Goldman spent five months as the sole bank advising Alphabet on its equity raise before other banks were brought in, describing it as potentially the largest secondary follow-on equity offering ever.
- Solomon argues the SpaceX IPO mandate was earned over 20 years of relationship building starting with SolarCity and Tesla, and was not won through any recent single pitch — explicitly denying that he DM'd Elon Musk on X to solicit the deal, though confirming he does communicate with Musk via X DMs.
- Solomon distinguishes the current market concentration in top 10 S&P 500 companies (~30x forward earnings) from the dot-com bubble (~40-50x forward earnings), arguing today's top companies generate substantial real earnings and cash, making the comparison to 1990s valuations imprecise.
- Solomon argues that the structural shift toward fewer public companies is the result of decades of policy and market structure decisions that made going public unattractive for smaller companies, and that companies going public now are doing so primarily because they need capital at a scale private markets cannot efficiently provide.
- Solomon contends that human emotional intelligence, voice, and interpersonal skills will become more — not less — valuable as AI commodifies analytical knowledge, arguing that a CEO's instinctive reaction to a banker's pitch ('I just didn't like the way that guy talked to me') is a human judgment no model can replicate.
- Solomon expresses concern that while large banks are well-invested in cybersecurity, mid-sized banks and critical infrastructure operators like regional utilities and water companies lack the resources to defend against AI-enabled cyber threats, warning that a cyber attack on a mid-sized bank could create systemic ripple effects similar to SVB.
Topics
Full transcript available for MurmurCast members
Sign Up to Access