AI's Reality Check: Financial Bubble or Foundation for the Future?

Episode Transcript

00:00:00 Mike:
Welcome to Growing Ecommerce. We are your hosts. I’m Mike Ryan.

Chris: My name is Chris. We’re really glad to have you back for another episode.

Mike: Hello, sir. Chris, always a pleasure. Chris, I’m going to address the elephant in the room. Can I? I noticed something. Anyone who listened to the last couple of episodes knows Chris was in the middle of a stressful move.

00:00:19 Chris:
Yes. And my closet wasn’t set up. No, not at all. But it looks like you’re getting there. I’m getting there. It’s not that trippy, but I’m on my way. I feel way more confident today.

Mike: All right. Good, good. You should feel confident. We have a big debate about to happen. Let’s see. We’re going to continue a conversation we had last time about the AI bubble—bubble or lack thereof. Why don’t you give your take on it? For those who didn’t listen last time, we did an intro to make a statement on where we are at.

00:00:57 Chris:
Did you change your mind, or are you still—

Mike: Well, there are so many dimensions to this, and no one has a crystal ball. I’m always “Mister Nuance,” so let’s be honest. I think a lot of AI attention has been directed at topics like workforce replacement and productivity gains. Recently, Salesforce said they let go of 4,000 customer service roles. But let’s see where that goes, because Klarna did the exact same thing and then had to backtrack massively.

I’m always wondering about this. Productivity has been flat for the past couple of decades despite the boom in software and all kinds of productivity tools. Meanwhile, we have an aging population and we need productivity to keep on growing. People worry about job replacement, but I would love to see it boosting productivity. This is one of the biggest reasons why I’m a bit bearish: we haven’t seen this impact yet. It’s always “the next six months” or “next year,” and I’ve heard that for a couple of cycles. Maybe it’s really true, we don’t know, but that’s what I’m missing—the bigger impact.

00:02:51 Chris:
It’s a strong statement which I don’t really want to fight, because we see it at a very small scale in our company. We’re roughly 100 people and we want to be an AI-first company, which I think we truly are. But we see on a daily basis that those low-hanging fruits we thought were everywhere with AI are not that low-hanging. I’ll give you that. AI is not that easy to implement to really have a massive ROI or productivity impact right away.

Some big companies have to crawl back to their old ways of work because they were too bullish. I give you all of that. But that’s why you’re absolutely right that there are so many nuances. How do we define a bubble? I’m coming from the financial perspective. Are we in the midst of a financial bubble? Is this AI dream we all invested in about to burst? I think we aren’t—we are not even close to that. I have some facts which we can discuss from a financial perspective before jumping back to the operational layer.

For instance, the last bubble we really had was the dot-com bubble in 2000. Back then, the S&P 500 was trading at a price-earnings ratio of 32. Today, the S&P is trading at 25. From that very meta level, I don’t think there’s that much of an over-investment going on. As I mentioned last time, this whole game is now driven by fundamentally sound companies: Microsoft, Apple, Google, and Tesla to a certain degree. These companies are fundamentally sound; they are not overvalued. We can discuss Tesla, probably, but if you look at NVIDIA and Microsoft, these are sound companies with massive CapEx going into AI. The companies driving this AI game are way healthier than those in the last big bubble in the early 2000s.

This is why I trust it from a financial perspective. I think it’s not the prototypical bubble. I think we are just at the beginning because the return on investment gains will come. That’s what I really believe in. One thing I will grant you from this financial perspective is that even big companies have to be a bit careful. If you look at the CapEx to EBITDA ratio for the biggest companies in the S&P 500, that rose from 0.5 to 1.3. So, companies are investing way more into an AI vision where we don’t fully know how it will play out. There might be some risk that some companies are slightly over-invested and the returns are not coming immediately. But again, the companies driving the AI game are so fundamentally sound that I just don’t see a bubble from a financial perspective. What’s your take on that?

00:06:26 Mike:
I just feel like at this point, this is a bigger part of the economy than all of consumer spending is. It feels like we’re moving from conversations about needing to keep consumer spending rising to needing to keep compute rising, and that feels odd to me. We just don’t quite know where this road is headed. What happens if we reach a point where there is an AI winter?

00:07:02 Chris:
This will happen, by the way. I think the path will be a tough one. To jump in on that, the market valuation of this AI industry is right now at a whopping $371 billion. Projections by the end of 2032 are $2.4 trillion. The bullish investment happening now is, of course, based on something that is not materializing yet. Will these return gains come? Will these productivity gains really happen? I think there might be some corrections, but I don’t see the industry imploding or companies going bankrupt. Compared to the small, fractured structure of the dot-com bubble, it’s now way more consolidated. There are a lot of startups—don’t get me wrong—that are benefiting from AI, but the major companies driving this are super sound and they know what they are doing.

00:08:44 Mike:
There are a lot of startups out there, and we’ll see what happens with them; a lot of them won’t exist in five years. If you’ve ever seen the classic Martech landscape chart, it nearly doubled recently. It absolutely took off. I think in a worst-case scenario, the big companies can withstand this. I don’t think we’re going to need a big tech bailout like the bank bailout in the housing crisis. But the market is at all-time highs and feels overall frothy, and it feels like a huge amount of that is based on this giant speculative bet where we don’t know exactly what’s going to happen. That could be damaging to the whole market if something goes wrong.

00:09:34 Chris:
I follow a lot of financial meme accounts, and you always see memes about NVIDIA being the last thing holding up the whole global economy. It’s absolutely true. There is a certain concentration risk because the market bullishness is carrying the markets for the “Magnificent Seven,” and almost all of them are somehow tied to this. If we were about to witness the bubble bursting, then we would have a situation like the early 2000s. I just don’t see the bubble, but there is a concentration of risk.

One thing that maybe counters your statement about not seeing operational low-hanging fruit is that a lot of foundational investments have been made. It’s not about having one killer app; a lot of foundational investment is happening right now which puts us in a position to eventually realize ROI. A lot of what has been happening the last couple of months or years was infrastructure building. That gives me confidence that institutional investors see this and are embedding AI in their infrastructure. It’s embedded in everything we do and everything we’re building.

00:11:58 Mike:
I see they’ve done that work of embedding and made that investment. But it’s left and right inside of Google Workspace that we use, and while some features like meeting transcriptions are awesome, some of them are very much in search of a use case. I have this feeling that we’re going to go through the classic hype cycle. Maybe I’m excessively pessimistic, but I think it’s going to take time for the use cases to really be discovered and trickle through markets. There are still companies operating with fax machines today.

00:13:14 Chris:
I know one of those clients! No name-dropping, but it happens. Mike, how do we wrap this up? What I would like to conclude with is that what we are talking about with the AI vision is very much tied to the digital world. One thing that is hardly priced in is what is going to happen with AI in the physical world, talking about robotics. I don’t see that priced in at all. This is probably an even bigger market than the software or digital AI game. We are about to have a massive disruption in what AI can take over in our real physical world, and that game is not priced in yet. So I think there’s more leeway.

You couldn’t convince me, Mike, that we are in a bubble. What I grant you, however, is that this immediate return everyone envisioned is not that easy. Much work has to be put in to make AI a valuable asset. But I believe this is the next big thing for mankind and, from a financial perspective, we are just at the beginning.

00:14:48 Mike:
Well, let’s leave it on that note. We’ll talk about that again in a year when we know more.

Chris: Or after the biggest financial shock in human history.

Mike: I won’t show up for that one! Let’s move on. A quick follow-up from a topic we discussed in a past episode: Amazon leaving the Google Shopping auctions. Do you have some insights here?

00:15:35 Chris:
Exactly. They returned everywhere except the US. But the big question is, what’s the impact?

Mike: There had to be an impact, right? What if I told you there wasn’t? It was basically nothing. It’s surprising that Google Shopping’s largest advertiser turns off all their ads overnight and there isn’t a huge shock. Last time, I looked at the seven days after they turned off the ads and compared them to the median of that weekday year-to-date to see how this varied from a typical day. There was no variance. Now I applied the same methodology to their return—compared the seven days after to the year-to-date median—and it’s the same thing. It’s flat. It’s crazy. Last time, that proved to be a pretty okay methodology because we heard from advertisers that nothing changed.

00:17:45 Chris:
To paraphrase: the biggest e-commerce player and probably the biggest Google spender abruptly leaves the auction and comes back basically to all markets within a day, and there’s no shock to the system? No changes to crucial KPIs like impressions or CPCs?

00:18:15 Mike:
Correct. And it’s shocking. It might speak to the amount of control that Google has over this fully automated system—that they’re able to somehow absorb this or modulate it. If Amazon exits the auction, you would expect a system shock because all of a sudden there’s a huge amount of ad inventory available. If there’s enough latent budget sitting there ready to spend, it could flow in, but not in a 1-to-1 ratio because there was a reason that demand was latent—they weren’t willing to bid as high as Amazon. We would have expected some dip in the CPC.

00:19:32 Chris:
Google looks at multiple factors, not just the bid. It’s not purely an auction; they look at predicted click-through rate and website quality. Amazon is very strong on pricing and delivery; they’ve got their reviews in order. These things support their dominance. A couple of years ago, we were working with a big book retailer when Amazon shut down their shopping ads in one specific country and then activated them again. It sparked everything up immediately. Every crucial KPI changed the moment Amazon entered the auction again. So this current situation feels very weird.

00:22:15 Mike:
I don’t know what more to say about it. I could be wrong, or my methodology could be wrong. I’d love to hear what other advertisers find in their accounts. Shout out to the listeners: if you have anything that supports or contradicts what we discussed, we’re happy to talk. Contradict me—make me smarter! But for right now, that’s what we’re seeing.

Onto the next topic. Chris, do you hear about Merchant Center account suspensions often?

00:22:45 Chris:
Yes. There are not many things more critical to online retailers than hearing a whisper of a potential Merchant Center shutdown. Why are you asking?

00:23:09 Mike:
Well, you’d think it’s super easy to fix a problem, right? Just one phone call? Just kidding—it’s the worst case. We’re dealing with a combination of automation and bureaucracy that creates a “hellscape” for advertisers. I remember a large U.S. sports brand we worked with; their main channel was Amazon, but they spent a ton on Google Shopping. They had an account suspension that took ages to fix.

00:24:19 Chris:
As far as I know, the team which decides if a Merchant Center can be reactivated is completely detached from the teams working with the clients or partners like us. It’s a structural thing with Google. The teams supporting the retailers do everything they can, but they are isolated.

00:25:16 Mike:
There’s a conflict of interest because it would be in Google’s financial interest to look the other way, but those teams are isolated. I also think there’s a tremendous backlog and a huge amount of individual variance and judgment involved.

I want to drop some numbers on you. Google, like all large online platforms in Europe, is required by the DSA to share moderation data. Google Shopping is the most heavily moderated platform in the Western world. The second most heavily moderated is Amazon, which makes sense given the volume of products and sellers. But Google Shopping is about seven times more moderated than Amazon.

00:26:59 Chris:
Seven times? That sounds crazy to me.

00:27:19 Mike:
In Europe alone, there are between 30 million and 70 million products every single day that receive limited visibility due to policy violations, real or perceived. This results in an average of 14,000 account suspensions every single day in Europe on Google Shopping. That’s why it might take you a while to hear back.

00:28:33 Chris:
So it’s not just about the structural separation; there are just too many to handle. What about the level of automation in detection and decision-making?

00:28:53 Mike:
More than 99% of detection is automated. To reactivate, you have to go through a human team, but 62% of those final decisions are also fully automated. I hope it’s solved by AI in the near future so we can make better decisions.

00:29:52 Chris:
Let’s see. You got me, Mike! I enjoyed the podcast, though.

00:30:15 Mike:
Me too. Thanks for chatting, Chris. That’s another episode of Growing Ecommerce. This podcast is brought to you by Smarter Ecommerce, also known as smec. You can learn more at smarter-ecommerce.com. As always, please leave a review or a rating. Thanks for listening and we’ll see you next time.