ROI Poses Ultimate Test to AI Spending Boom as Volatility Climbs

SPEAKER_1 (0:00)

I do want to shift gears here, though. Turn our attention to the economics of AI. Bring Sam back in and welcome in our next guest. Joining us, Anik Bose, the managing partner at BGV. Thank you so much for being with us today. Now, we are seeing enormous amounts of capital flow into AI infrastructure, but you say that the ultimate economic test sits with the enterprise customers. You pose the question whether using that intelligence generates enough value to justify continued spending. So our question to you is, does it?

Anik Bose (0:32)

Yeah, I think it's still early days. What we're basically seeing is that in order for enterprises to buy the capacity that's being built out in these AI factories, they have to basically be able to prove that those AI investments they're making to buy the capacity is generating either revenue growth or profitability growth or cost reduction.

So far, it's very, very early days.

SPEAKER_3 (0:58)

So is that the value? It's the revenue growth and the profits through cost reduction. Are those the metrics we should be looking at? I mean, how long do you anticipate? How many quarters of earnings do we need to sift through? Do you think until we start to see evidence of that, how patient can this market be, Anik?

Anik Bose (1:17)

Yeah, I think if you dig a little bit deeper into it, if you're a large enterprise, fundamentally, what they're looking for is lower costs per successfully completed workflow, because they're using AI to redesign their workflows, and either they're going to get higher output with the same level of resources or they're going to get better decisions or they're going to drive incremental revenue. When they make these investments, they're looking for paybacks, whether it's in three quarters, four quarters, five quarters, but it's definitely not in one quarter. So fundamentally, what that really means is the benefits for any investment that they make has to exceed the full cost of AI, whether that's kind of the capacity they're buying, whether it's integration they have to do, whether it's the human oversight that's required to make it work. So unless these enterprises are generating tangible ROI from the AI investments, it's going to be tough.

SPEAKER_1 (2:14)

And Anik, you brought up a good point there with tangible ROI. I've read a few notes lately about the revenue growth that we're seeing across the AI ecosystem, but that doesn't necessarily mean that these customers are actually getting a return on their investment. Are you finding that investors are confusing spending on AI with proof that AI is actually paying off? Because I feel as though I've seen hints of that in many articles of late.

Anik Bose (2:43)

Yeah, you're absolutely correct. When you think about it, what we're seeing today is infrastructure spending. That does create revenues for suppliers, whether it's Broadcom or AMD or NVIDIA, or it's the hyperscalers, whether it's Microsoft or Amazon or Google. But what's happening there is, while their revenues are growing up, what's really interesting also is some of these companies can also be one of those investors, suppliers, and customers. There's a little bit of a circular economy.

For example, if you look at Anthropic, they have a commitment to basically spend close to $580 billion over the next 10 years, and 80 percent of that is non-cancellable. It's with folks like AMD, folks like Microsoft and others. What's important is while that's building out the supply side of these AI factories, there's not today, if you were to drill into it and say, how many enterprises have signed up for long-term contracts to use that capacity?

And there's very little visibility into that.

SPEAKER_3 (3:48)

Have you got any examples or could you point to any particular companies or sectors that are seeing some evidence of the productivity now, where obviously this is starting to show up in the numbers as far as strengthening the business case? Because I remember last earnings, there was a lot of talk around Airbnb, for instance, just to pluck something out of thin air that they had been showing the efficiencies that the market was looking for, given the amount of job cuts that we'd seen from that company, but what they had actually managed to do with the use of AI.

Anik Bose (4:27)

Yeah, I'll give you two examples. The first example I'll give you is Amazon. One of the things that they've introduced is something called Rufus, which is basically an AI shopping assistant.

The problem they're trying to solve for is fundamentally, that they have a high abandonment rate of carts when consumers go to the site. What basically Rufus does, it basically provides information, engages the customers on specific recommendations, on products based on the history and what they're looking for to improve that conversion rate.