As part of the CRM Playaz "What is CRM in 2026?" excutive conversation series, for the first time we added a panel of folks investing at different levels into the enterprise application space, as well as in AI startups that are heavily impacting CRM/CX technologies. And given all of the uncertainy surrounding AI and the exponentially growing amount of investment being plied into the infrastructure buildout, one of the key questions that came up was around 2026 being the year of an AI bubble bust (or is it really "burst"?); and what impact will it have on the industry. As part of this particular paart of the conversation we were grateful to have the following folks take on this question in the clip below:
- Anand Thaker - Principal & Operating Advisor
- Keith Teare - Founder and CEO, SignalRank
- Larry Augustin - Managing Director, Augustin Ventures
- Scott Beechuk, Partner, Norwest
Below is the interaction that took place after I posed the quetion asking them each to give their take on if they see an AI bubble exploding in the year ahead. And below the clip is an editted transcript of the discussion, followed by an analysis of the conversation from Gemini for your review.
To see the full panel discussion go to What is CRM in 2026? - Day One.
Edited transcript
Brent Leary : Is 2026 a year that there is a potential AI bubble?
Larry Augustin : So I think we're gonna see a combination of winners and losers here where there's gonna be some big losers and big winners. I don't know if you fall— put that into the definition of a bubble or not, but I think some of these companies have raised a ton of money, are gonna discover ultimately the economics don't work and they're gonna crash.
At the same time, we're seeing some companies that are delivering phenomenal value, have real growth, real revenue, and that will be hugely valued. We're gonna see coming out of this, you know, another wave of mega-valuation giant tech companies. We're also gonna see some big craters in the ground.
: Right now, that's what I think is going on in the, in the venture world, is everybody's trying to figure out the difference. And in some cases it doesn't matter. You just invest in them all. The big ones will make up for the craters.
Keith Teare: I think you could characterize the next several years as the Golden Age of Margin Expansion. Costs are gonna go down. Gross profits are gonna go up. Why? Because the cost of doing things is tending towards zero. Probably never reaches zero. Platforms replace humans and marginal costs replace salaries. Society has some issues there—how to deal with the humans—and I think that's a separate conversation.
In a Golden Age of Margin Expansion, a bubble makes no sense as a framing narrative. This is not a bubble. This is Day One of making a lot of money. Some bets are gonna be bad, but that is almost not worth talking about, that's always true. It's more the good bets that are worth talking about.
Larry Augustin : The margin expansion, Keith, super, super interesting point because I see two sides of that. One is we can decrease our overall cost structure because we can use technology and that's a big opportunity for margin expansion. But on the other hand, the AI costs... I see a number of companies right now that are upside down, and in particular on gross margins on COGS, because the AI costs are actually much higher.
I mean, SaaS margins, you know, were great and we were paying for the infrastructure. Now we're paying for the AI and suddenly I see companies delivering a great product with like a 10% gross margin and all the dollars are going to the uh, LLM vendors and the infrastructure players.
Keith Teare: Any AI company making a profit today, the—their shareholders should shoot them. They should be spending every penny on growth. Every penny. Becoming profitable shouldn't be a goal. Becoming unit economics profitable, yes. That should be true from day one. I don't think you stop investing until you—until you stop seeing growth.
Scott Beechuk: I would say that we're in 1995 right now and you've got the bets laying down the fiber optic cables for the internet. And they're gonna keep laying down those fiber optic cables. And guess what those are? The fiber optic cables are GPUs. And it's a lot of other infrastructure for AI that's being laid down right now.
And we've all seen the numbers. Uh, we don't need to, you know, go into the, to the depths of all of the, the commitments that have been made through 2030 and beyond. But the question is, are we overbuilding infrastructure right now? And so it's gonna take some time before we know the answer to that. And the answer will be dependent on how successful a lot of these application companies that consume that infrastructure are.
And one thing that we do know is that the cost to deliver—Keith to your point on the unit economics—yes, you do have to have a unit economic model for your business that makes sense today. But you also have to squint your eyes a little bit and assume that the token cost is gonna continue to come down and the cost to serve is going to continue to come down over time. We don't know how much and how fast, we've got some guesses. But if we're in 1995, then we've got about, you know, call it five to six years before, you know, maybe we've overbuilt the infrastructure too much.
Or maybe it's not gonna play out exactly the same way and maybe the application companies do consume enough of the infrastructure. I tend to think that history repeats itself. We will probably overbuild infrastructure for AI. I think we probably will. Um, but it'll take some time for us to all figure that out.
And then to Larry's point, there are gonna be some massive winners here. Some companies that not only get the unit economics right but get the productivity value right. And so those companies are gonna grow at rates that we've never seen before in B2B software.
Anand Thaker: Infrastructure will always get repurposed. That's what happened with the internet. There was a lot of lines that got dropped, you know, to continue that actually story, which I do agree with. We are in like that '95, '98 kind of era. You know, or at least it feels that way, especially with laying down the ground lines. But all that infrastructure got usurped pretty quickly, even after 2008, and required an incredible amount of more expansion as a result of it.
Anand Thaker: So I can tell you from the deals that we've been working on or the things that I've been involved in, it's like storage units. You can't actually, you know, I mean yes some of those GPUs and all that or you know the technology might be aged, but the use of that infrastructure can be repurposed pretty quickly and will be found to be used in a different capacity. So um, yeah are we rebuilding, you know, building too much now? Perhaps, maybe, maybe not, but it will never go to waste.
Summary: Is 2026 the Year of the AI Bubble? (via Gemini)
The discussion centers on whether the AI market is heading toward a bubble in 2026. The panel largely rejects a traditional "bubble" narrative in favor of a complex market correction and infrastructure boom comparable to the internet era of the mid-90s.
1. The "Winners vs. Losers" Bifurcation
Larry Augustin argues that rather than a total market collapse, we will see a sharp divide. Companies with poor economics will "crash," while those delivering real revenue and growth will see "mega-valuations."
He notes that VCs are currently investing broadly ("spray and pray") because the massive winners will likely cover the losses of the craters.
2. The Golden Age of Margin Expansion
Keith Teare rejects the bubble framing entirely, calling this the "Day One of making a lot of money."
He predicts a "Golden Age of Margin Expansion" where technology tends costs toward zero by replacing human salaries with platform costs.
The Counterpoint: Larry counters that currently, AI costs (LLMs and infrastructure) are actually hurting gross margins for many SaaS companies, making them "upside down" economically.
3. The 1995 Internet Analogy (Infrastructure Phase)
Scott Beechuk compares the current moment to 1995, likening the purchase of GPUs today to laying fiber optic cables for the internet.
He predicts the industry will likely overbuild infrastructure, but emphasizes that token costs will eventually decrease.
Anand T. agrees with the 90s comparison but adds that overbuilt infrastructure (like the "dark fiber" of the past) is never wasted. It will be quickly repurposed and absorbed by the market, just as internet infrastructure was post-2008.
4. Profitability vs. Growth
Keith emphasizes that AI companies should not focus on net profit right now, but rather spend "every penny on growth."
However, the panel agrees that Unit Economics must be positive from day one. Companies can burn cash to grow, but the fundamental business model must work.
Would you like me to dive deeper into the specific debate between Larry and Keith regarding AI operational costs and gross margins?