Kanyi Maqubela of Kindred Ventures expects 2026 to reshape venture capital through potential blockbuster IPOs: SpaceX, Ramp, Anthropic, and OpenAI could each reach $0.5 trillion to $1+ trillion valuations. He warns of concentrated market risk in Nvidia and Big Tech, massive private funding into AI foundation models, and debt risks at hyperscalers, while predicting AI will both commoditize at infrastructure level and specialize in verticals like enterprise coding.
Chapters
When you talk some tech right now, valuations. Kanye Magella is managing partner at Kindred Ventures. Want to get into all of this with him this morning as we try to understand where this is all going. The venture business has been an interesting one. This could be a huge year, huge year for some of the, the vintage, but probably what, two, five, five, six, seven years ago? What's going to happen in 26?
Yes, this is really an exciting year because you've got some of the biggest potential IPOs probably of the last decade, companies like Space X, potentially companies like Ramp, and then of course the big AI companies who are looking ahead Anthropic and OpenAI. And these are companies that are going to be going public at anywhere from a half trillion to one trillion plus dollars. So that's really blockbuster IPOs which are going to let a lot of air out of the balloon for some of this cash that's been going in without any end in sight on the other end. So it's actually a really important year for venture capital and the question is.
How much do you think that's actually going to happen in 26? How much do you think gets pushed out to 27, 28, etc. And we talk about bubbles all the time. Is this the sign of one when.
We start the bubble question first, which is a very important one. And the thing about this economy that's really interesting is there's so much concentration and you're seeing it in the public markets, of course. So Nvidia is more concentrated as a single stock than you've seen probably in 40 years. Mag 7, 30% plus of the, you know, of the S and P and the tracking error is totally off. And then you're seeing that actually in the private markets as well. And so if you look at the private capital that's gone into just the large foundation models, the Frontier Labs, that represents a higher percentage of the dollars that have gone into venture capital than almost any other time in the last 10 years. And that's because the narrative is so clear right now, which is that everybody wants to, to build the new compute stack, the new inference stack and the new demand stack for AI. So whether or not that's a bubble actually has a couple of inputs that I'm looking at personally. One of which is is there going to be enough demand for all this build out? And we should talk about that. Is the deprecation for all this infrastructure actually going to work in such a way that the unit economics make sense? Are the multiples that these companies that are concentrating all this capital that we're trading at reasonable. When I look at each one of those today, it actually seems more rational than you'd expect. There's one place that I'm tracking where there's a bit of concern, which is the bond. So there's a big debt load that a number of these big hyperscalers and companies like Oracle and Core, we've taken on where there is a point of concern. But on the multiples, on the deprecation schedules and then on the, you know, and then on the demand, I actually feel pretty confident that we're in a healthy place.
Do you think that in terms of the large language models themselves that they all become commodities? I mean part of it is that they all may ultimately, if they're as successful as they want to be, you think that they would all converge.
There's two answers to that. The first answer is yes, they become a commodity the same way that the hyperscalers made the, you know, the cloud storage a commodity. That's a cloud storage commodity that's growing at 10 to 30 billion dollars of revenue a year, 30 to 40% plus year over year. So even if that were a commodity, it's actually a really good business because there's just so much net demand. So that's path, one path to which is the path that I think we're actually seeing is that each one of these models actually starts to specialize on certain areas where they become really, really strong. And so you're seeing Anthropic has really done a really, really important, big investment into coding, into Claude, which is actually their enterprise agent. And between enterprise and coding, they actually have a true differentiating advantage in their technology and how that technology is able to deliver results for end users. Whereas chat, GPT and OpenAI is really focused on the consumer use case and they're focused on media things like Sora, things like their whisper model which are audio and video. And then of course the chat model which is really, really important for consumers. And then Google which is, you know, the big monster is kind of doing it all. But I think that actually is going to play out even more deeply as these companies further invest into each of these lanes.
Final question, is there a way to invest in this in some other way that for, for a public investor today that doesn't seem obvious to you?
Yes, if I were looking at it, I would actually be thinking hard about the agentic app layer. So the reasoning models just came out over the last 18 months and these reasoning models are finally making enterprise use cases useful. I'm seeing it insurance, I'm seeing it in law, I'm seeing it in health care, commerce. And so I would be looking for companies that are spending a lot of energy and time figuring out how to actually replace services or with AI in productive ways. Peter Orszag earlier was talking about the way the Lazard is doing it, and I actually think at a stock picker level, there's individual names where there's an incredible amount of margin expansion available.
