AI Spending Remains Concentrated as Lower-Cost Models Gain Popularity

**SPEAKER_1** (0:00)
Joining me now to discuss trends in AI adoption is Ara Kharazian, who's the lead economist over at RAMP. Ara, thank you so much for your time, and happy Friday. Thanks for coming down to the New York Stock Exchange as well.

**Ara Kharazian** (0:10)
Thank you for having me.

**SPEAKER_1** (0:11)
Absolutely. Just walk us through exactly what RAMP does.

**Ara Kharazian** (0:15)
Well, RAMP is an AI finance platform, but you can really think about our data set and the research we do through our products. It's a corporate card and bill pay invoicing products. So the data set that I work with as an economist is that universe of business spend, particularly on AI trends where this is a very closed off market, the companies developing the technology are privately held, is very little to go off as far as how this industry is developing. You can use surveys, but those are a little bit in conflict with the actual developments in the market. So we use actual business spend data to understand what are firms buying and how.

**SPEAKER_1** (0:47)
It's so fascinating because all we've got to go off is the corporate earnings, which obviously we've been getting a lot of this week. I mean, to gauge the scale of this AI trade and the spend behind it, what is your data showing right now?

**Ara Kharazian** (0:59)
Well, it's been a transformative market with a lot of shifts both over the last year, all the way starting in 2023 You saw at the end of 2025, beginning of this year as well, a massive run-up in the share of businesses that are using AI at all. Now, more than 50 percent of businesses in the US are using AI. The question is now shifting toward the intensity of that adoption, and then how far this market can grow going forward.
That's where you see a lot of differences between the kinds of businesses that are investing it. The top 1 percent of firms are spending 7.5 thousand dollars per employee per month on AI.

**SPEAKER_1** (1:32)
Wow.

**Ara Kharazian** (1:33)
But the median is only about 12 dollars per month. So big differences between the kinds of firms that are driving this kind of investment, and the vast majority of firms that are just getting started.

**SPEAKER_1** (1:43)
So where are you seeing the most aggressive spend right now across company or industry?

**Ara Kharazian** (1:48)
AI spend follows this power law interaction where the vast majority of spend is happening amongst a very small set of firms. Now, I will say there is still a lot of room for growth clearly, especially amongst the median firm. But there are some suggestions that there are cracks forming in the AI thesis, in the AI trade, and businesses are hitting a limit as far as how much they are willing to spend on AI.

**SPEAKER_1** (2:09)
Yeah.

**Ara Kharazian** (2:10)
The two major factors there, Fable 5, Anthropic's most powerful model, the most powerful AI model to be released to date. So powerful that the government briefly banned it.

**SPEAKER_1** (2:21)
Yes.

**Ara Kharazian** (2:21)
Has underperformed our expectations in the market. Only about 11 percent of business spent going to Anthropic is now on Fable 5 So businesses aren't really willing to spend on Fable 5, the most performant model, but also the most expensive.
Alongside that, you have increasing adoption of open-source models and the kinds of platforms that enable access to cheaper models, sometimes Chinese models.

**SPEAKER_1** (2:45)
I was about to go there. I mean, what are you seeing as far as appetite for adoption of Chinese models, given that obviously this is something that I'm sure the folks in DC are watching very closely.

**Ara Kharazian** (2:53)
Today, we see about businesses spending on AI, about six percent of them are using either open-source models or Chinese models with direct payments to Chinese model companies. Now, it's relatively small. It is growing. Six percent isn't really enough to freak out about.
I will say within the group of businesses that are spending on these models, they continue to spend and increase their spend on American model companies, OpenAid and Anthrobic.

**SPEAKER_1** (3:18)
But the point is, it's cheaper and it's open-source. That's what makes the Chinese so attractive right now. I'm wondering then, you mentioned the companies maxing out their spend. What was the high-profile one back in April? Was it Uber who said it had spent an entire AI budget? How widespread is this across industry right now?

**Ara Kharazian** (3:35)
Well, it's common, but part of the issue of analyzing this in the market is that, if you're to look at a token maxing period in our data set, you wouldn't really be able to find it on a chart. Because even amongst the top 1 percent of firms, yeah, on the margin, we're finding that they are becoming more cost-conscious, but they continue to increase their spend month over month as well. Part of that is because the top 1 percent of firms spending on AI are also very fast growing segment firms. So their AI spend, although it is growing, is not crowding out other investments for the firm. They're generating so much revenue that they have money to spend.