Goldman Sachs CEO David Solomon Talks ‘Very Robust’ Demand for AI Deals

David Solomon · Goldman Sachs

AI infrastructure spending will not rise in a straight line; expect stops, starts and recalibrations.

Solomon cautions that the market expects AI infrastructure spending to be constant in a straight line, but every other technology super cycle has been non-linear, with calibrations and stops and starts. He adds that new technologies and evolutions could change the trajectory of the build-out, and that it is very hard to know how much will ultimately be spent building infrastructure worldwide.

“there's an expectation that it's gonna be constant in a straight line. And if you look at any other technology super cycle, it's never constant. There are calibrations. There are stops and starts.”at 3:29

David Solomon · Goldman Sachs

Compute demand far exceeds supply; robust financing will come over the next 12-24 months.

Solomon argues that the demand for compute right now far exceeds supply, which is why he expects robust financing to meet that demand over the next twelve to twenty-four months. He calls the moment a super exciting time for AI infrastructure build-out but says it will not go without bumps. His evidence is the current gap between compute demand and supply that Goldman is financing.

“...the demand for compute right now far exceeds the supply. And certainly in the next twelve to twenty four months, there's gonna be robust financing, you know, to meet some of that demand.”at 3:51

David Solomon · Goldman Sachs

Capital is not the constraint on mega AI deals; higher rates and government debt raise cost, not availability.

Solomon says the constraint on how large AI financing deals can get is not the amount of capital in the world. He points to public equity market capitalization of about $160 trillion, and says investors are still looking to participate, with debt investors demanding higher returns when markets get more challenging and equity investors taking risk for upside. He concedes that higher rates and governments issuing more debt make capital competitive, which affects cost of the financing rather than its availability.

“I think the constraint is not the amount of capital in the world and the size of these deals on a relative basis.”at 5:09

David Solomon · Goldman Sachs

Five to ten years out, AI should drive significant productivity gains across the economy, despite bumps.

Solomon says that stepping back five to ten years, he is very excited about how AI can be deployed in the economy and drive productivity gains that can be very fruitful. He acknowledges he cannot say the build-out will go without bumps in the coming years. He also says that the application layer will be built over the next decade or two, leaving room for substantial upside for investors.

“But I can say when you step back and you look kind of five, ten years from now, I'm very excited about how this technology can get deployed in the economy and really drive productivity gains that can be really, really fruitful.”at 4:10

David Solomon · Goldman Sachs

Regulation allowing deals at scale is driving broad consolidation and M&A activity in the US and Europe.

Solomon argues that M&A activity around the world, particularly in the US and Europe, is more shaped by a regulatory environment that permits deals at scale. He says everyone recognizes the importance of scale in almost any industry, which is leading to broader consolidation across industries and more M&A. He says Goldman has benefited from being in the middle of that given its leading M&A franchise.

“when you look at M and A activity around the world, particularly in The US and Europe, it's more shaped by a regulatory environment that's allowing deals at scale.”at 7:49

David Solomon · Goldman Sachs

AI needs governance: government sets rules, builders and deployers are accountable for their use.

Solomon's op-ed proposes three principles for AI governance: government sets rules of the road, the people who make the technology act responsibly, and enterprises that deploy it are accountable for how it is used. He cites Goldman's roughly twenty years of AI trading platforms, with no platform losing hundreds of millions of dollars, as evidence that accountable deployment works. He argues this mirrors the financial-system model built since the financial crisis, and that humans must remain accountable even as models act without oversight.

“The first is government has to set some rules of the road. The second is the people that make the technology have to be responsible and thoughtful about what they make and how they sell it into the market.”at 11:32

David Solomon · Goldman Sachs

Open-weight models will proliferate across many sources and bring AI costs down.

Solomon expects a big role for open-weight models, which will proliferate from many different places and bring costs down. He says Goldman still needs frontier models for certain uses, particularly cyber, but many productivity and process tasks do not require the frontier. So he says Goldman does not have to be first to deploy open-weight models, given the high governance standard it is held to.

“there's gonna be a big role for open weight models in the world... open weight models, and I think they will proliferate from lots of different places, will bring costs down.”at 14:50