EBay boosts investments in GPUs in push to harness AI
Summary
The online marketplace is carving out more space in its capital expenditure budget to fund its push into generative artificial intelligence.EBay is carving out more space in its capital expenditure budget to fund its push into generative artificial intelligence.
The online marketplace during the first quarter doubled the amount of computing capacity that it can produce from all of its graphics processing unit hardware compared with the end of 2023. GPUs are the computer chips behind generative AI programs.
EBay isn’t disclosing how much it is spending on GPUs, but investments aimed at supporting generative AI are accounting for a greater share of the company’s capital expenditures, according to Chief Financial Officer Steve Priest.
The company expects total capital expenditures in 2024 to remain steady compared with last year at between 4% and 5% of revenue. In 2023 it generated $10.1 billion, up 3% from a year earlier.
EBay last fall launched what it calls its magical listing tool, which helps sellers create listings by extrapolating details about a product from an image. Other uses of generative AI at the company include customer-facing features that, for instance, recommend apparel to shoppers, and internal company tools aimed at helping software developers improve their productivity.
The Wall Street Journal’s CFO Journal recently sat down with Priest and Nitzan Mekel-Bobrov, eBay’s chief AI officer, to talk about the company’s investments in AI, and how it is managing AI-related risks. Here are edited excerpts of the conversation, which was recorded for the Journal’s Executive Insights podcast.
WSJ: Nitzan, tell me about the generative AI tools that eBay is working on. What’s available already, and what do you have in store for the year ahead?
Mekel-Bobrov: We have a three-pronged approach to our generative AI strategy, the first one being the listing process. So really, these are seller-facing tools, including description generation, listing generation, as well as some of the image modification and enhancements that generative AI enables sellers to do with our tools to create more engaging and more studio-quality photography and listings.
We also have a number of buyer-facing features currently live. And then finally, our third prong is really around enabling productivity internally for our employees.
WSJ: Steve, what’s been your role at the company in getting some of these generative AI investments off the ground?
Priest: The way I approach being a CFO at the company has been a thought partner to the leadership team and really thinking about, how do we continue to invest to drive benefits for our customer, while at the same time continue to manage the financial architecture of the company?
WSJ: Let’s dig into some of those numbers. EBay said this year that it will spend between 4% and 5% of its revenue on capex.
Priest: As a company that generates around $10 billion in revenue, that 4 to 5% of capex, as a percentage of revenue is in that $400 to $500 million range. So that’s our total capex.
Now, as it pertains to 2024, we will be doubling the GPU capacity by the end of Q1 from the end of last year. We’re leaning in hard. We continue to invest.
WSJ: How much are you spending on GPU capacity? How much of your capex budget is devoted to that?
Priest: We don’t specifically talk about the GPU expenditure per se, but it’s integrated into the whole 4 to 5% of revenue that we alluded to, along with other sorts of investments that we will make to continue to generate benefits for our customers.
WSJ: What are some of your other AI-related expenses, in addition to GPU capacity?
Priest: It’s about reallocating spend across the business. There’s the capex side in terms of GPU capacity. We’re also investing in the right levels of tech talent to really drive innovation at the company.
We’re putting investment into areas outside of the sort of customer experiences in terms of driving better efficiency and velocity. We’re also looking at investing in our customer support areas to ultimately invest in those products in order to save money in the short term.
WSJ: Have you seen yet an impact on top-line growth or an impact on bottom-line profit?
Priest: Certainly there’s elements of momentum that we’ve seen from these investments in generative AI. Nitzan talked about the magical listing, whether it’s item description or the other stuff that we’re doing in beta. We are seeing customer satisfaction go up significantly from the investments that we’re making.
We’re seeing customers leaning in and using those AI-generated descriptions when they drive additional selling. So it’s clear to me that friction is coming off the marketplace, and it’s enabling sellers to list more products to make sure there’s more healthy inventory on the overall marketplace. So, inevitably, we’re seeing benefits under the hood.