Mantle’s Amar Varma on US Election Impact, Big Tech’s AI CapEx Race, and More

In this engaging conversation, Mark MacLeod sits down with Amar Varma, founder of Mantle, to explore the critical developments reshaping the tech landscape. They discuss the emerging concept of sovereign AI as a national security imperative, the unprecedented infrastructure investments by tech giants requiring nuclear power solutions, and what these seismic shifts mean for startups and investors in today’s competitive environment.

Drawing from his chip design background, Amar offers unique perspectives on why computing power has become the new global battleground. Let’s get right into it

Mark MacLeod:
We’ve got a different episode for you. This one’s sort of inspired by Tim Ferriss, and Kevin Rose, a random episode that they had recently. Now they did it with a bartender making them cocktails. We don’t have that, but I get together with my good friend Amar Varma pretty regularly. Amar’s here on the show with me and we often talk about what’s happening in the startup world and we thought maybe we should just do that on some kind of cadence here on this show. And so that’s what we’re going to do today.

And it’s an experiment. We’ll see how it goes. Please let us know how you think it went. Amar, welcome back. How are you doing man?

Amar Varma:
Well, I’m great. Thanks for having me. I’m excited for this episode.

Mark MacLeod:
Yes. It will either be amazing or horrible and nothing in between, that’s for sure.

So this episode’s going to drop on US Election Day. And since I’m the guy who gets to ask the questions, even though I’m going to do more than that here, let’s just jump right in. What are your thoughts? What are your predictions? What’s going to happen on Tuesday?

Amar Varma:
Yeah, we’re heading into the weekend before the election and it’s become even clearer who’s going to win next week. According to your favorite poll piece, whether it’s journalism in the newspaper, social media, or folks that you follow influencing your opinions, it’s one of those darned if you do, darned if you don’t type of feelings, everybody’s going to be excited if their side wins and extremely, not just upset, but action-oriented if their side loses.

So not having a vote myself, it’ll be interesting to see what happens. I think we’ve seen what happened after the 2016 election and the reaction. We saw what happened after the 2020 election and the reaction. So I would expect people not to be standing still after the election between now and early January.

Mark MacLeod:
I mean, a big lesson for me, I think back to 2016, I was on a flight from Canada to London, England, on the day of the election. I arrived at 9 pm London time, which was still the middle of the day in the US and I was like, I’m sure Trump’s going to lose, Hillary’s going to win it, whatever. I’m going to go to bed.

And then I woke up the next day in complete stunned disbelief and just sort of like, stumbled my way like a zombie around London, wondering what on earth happened. And so if there’s a lesson there, for me, it’s don’t take Trump for granted. You know, I totally dismissed his candidacy the first time around and it was real.

Amar Varma:
Yeah, I don’t think you can take anything for granted in this election. We’ve had attempted shootings, and we’ve had candidates sitting for president, so it’s just interesting to see all of this in 2024. I’m reminded I’m a huge appreciator of the series Narcos on Netflix, season two of the Mexico version.

You’ll see the 1988 election was taken, according to the show, by the Cartel. They had a piece of software which was to predict the election results. And the key piece was early in the day, they were showing who was winning, and it would have mimicked the rest of the day. They were able to influence those early poll results.

It kind of changed the outcome of the election. Look at that as a long time ago and a different lifetime ago. And I’m looking at what’s ahead of us, and hopefully, everything plays out smoothly and there are no complications, but we’ll see.

Mark MacLeod:
And you and I were chatting about the candidates this week. Obviously, we know what Trump stands for. He’s pretty vocal about it. And of course, we’ve seen one round of this already. And the thing with Kamala is we don’t know what she stands for. We just know that she’s not Trump. If I were an American, I’d be voting because she’s not Trump. And so that’s probably not sufficient.

Amar Varma:
I feel we’ve taken for granted our privilege and right to vote for a long period of time in our lives, and voter turnout is not 100%. And you see the outcomes of that, and you see countries where people can’t vote and the ensuing crazy chaos and disruption, violence that sometimes occurs. So, if you are watching and you can vote, get out there and vote, make your opinion count. What diligence you do have, and do you resonate with the person, the party, the policies, your neighbor, whoever you are, that is your right and privilege. Exercise it.

And I think the outcome of which, if one side wins, what’ll happen in the market and kind of to society, and if the other side wins, what will happen in society, in the market, like, we’re going to find out. There’s going to be a very interesting dynamic either way for the next, at least two to three years.

Mark MacLeod:
Yeah, listen, without trying to come up with the canonical, ripping apart of the US political system, which is way beyond my pay grade, I just think this whole two-party system is deeply polarizing. It’s just like, it’s like a party over country.

Amar Varma:
I think you’re right. I think it’s become this divisive tool in the modern day, versus it being a forum where you can kind of resonate and express yourself, where you sit on the spectrum of policies in society and people in society. And now it’s become you’re either with me or against me, or you’re against me or you’re with me. You’re darned if you do, darned if you don’t, back into that again.

And it’s like, if I’m going to one person’s house on Friday, is my tune changing versus who I was interacting with in the office earlier in the day? Like it’s getting people to have to be chameleons and not being able to let them really be themselves. I commend those who’ve been open about their opinion, their reasoning, their stance, whatever it is. I really commend those people because that’s really hard to do, and it’s hard to really stick out, and you’re going to face critique, and you’re also going to face the things you don’t hear behind your back.

And that’s the one that I really think is interesting. I’ll give you an example. I did a call for an LP who’s an investor in my current company Mantle, and at the end of the call, they had this five-minute conversation with me asking about the political views of the GP, and I was like, where did that come from?

First of all, it’s none of my business. But second of all, why don’t you ask them? And really, the crux of the questioning was, did I feel like their political opinion was helping or hurting me as a portfolio company in their fund? And I was like, oh, very interesting. I had to pause for a minute, going, well, they’ve been incredibly helpful, and I can’t, so I said, it’s probably helping.

Like, that was my initial response. I’ve had some time to ruminate and kind of let it marinate, but I don’t have something on the downside. But the problem is the downside is probably something that I’m not being told, I’m not seeing, I’m not seeing feedback on, or there’s a closed room. I just don’t know.

Mark MacLeod:
That’s fascinating. I’ve never heard that ever come up in diligence before. That’s a sign of the times, if anything. Maybe the last thing I’ll say on this topic, but then happy for anything else you care to bring up about it is I try not to watch the news because I find the news is polarizing, and divisive, it’s focused on the negative, and there’s so much goodness in the world that isn’t newsworthy.

But I did happen to just catch Elon Musk endorsing Trump at a rally last week, and it really pissed me off. Here’s a guy who I don’t think he is anymore, but for some period of time was the world’s richest person. Like, does he need more freedom? Does he need lower taxes? Like, where does human decency rank in his algorithm for picking a candidate? And a lot of folks have come out in defense of Trump because presumably, Republicans and Trump are for more freedom and for entrepreneurs. But where does decency show up?

Amar Varma:
I’ll say this. I’ve had, at a cursory level, many friends, and colleagues talk about how they’ve loved their Teslas for a decade, and they’re contemplating what else to get. Of course, they tell them a Ford electric Mustang is the number one, as you should…

Mark MacLeod:
For anyone listening and watching, by the way, Amar sold a company to Ford, hence, that’s that allegiance. So I just wanted to give that background.

Amar Varma:
Thank you. The concept is that Tesla is a computer with wheels for me, and my Mach E is a car with a computer. So it’s like a different dynamic.

Human decency is all we can exemplify as individuals every day. So I think it’s very important.

I’ll give you a quick aside. One of my previous companies, the Acquirer was run by this gentleman who was the number three guy at Microsoft when Bomber got the job. So, Bomber got the job. He left eventually, and I started. He ran VMware through its epic run, and Pivotal, which we did to Bubble Atkin and got bought by VMware and Broadcom. Long journey.

Anyway, the moral of the story is he said, look, we’re going to win this race, when we were contemplating joining, and he goes, that’s a for sure thing. He goes, but how we win will be what we’ll be measured by. Were these good, decent people conducting things in a manner that was representative of the image we wanted out there?

That was the takeaway for me for life, where I was like the privilege of the journey is in conducting yourself day to day with how you feel is exemplary of yourself, how you want others to perceive you, and how you want to treat others and vice versa and how you want to be treated. And I just don’t think there’s anything more core or critical to that in the day-to-day.

I’m by no means a politician or understand much about politics, clearly. So, I don’t know how that translates, but that’s my fundamental core fabric of people I interact with and the things that I’m involved with.

Mark MacLeod:
Yeah, listen, I won’t pretend to know what the reality is for a politician. I think many enter for the right reasons and then just get caught up in the system. But that’s another topic.

Let’s stay in America for a second. For the entire time that I’ve been alive. America’s been the leader of the free world. So much of global trade is US dollar-based. It’s kind of like the currency of reference, I don’t know if it still is, maybe China is, but I think it was the world’s largest economy for some reasonable amount of time. But do we still bet on America? The BRICS countries just had this big conference recently and Putin’s bringing in all of his so-called friends and there’s just troubled waters, wondering what you think.

Amar Varma:
I always resonated with some things I see, and maybe part of it was just marketing, and the frequency of seeing something. Betting on the index, on the stock market is the proverbial betting on America. That’s proven to be a great bet, at least in our living lifetime, if not before that as well. I see that still continuing.

When I go back in history, kind of prior to some of these revolutions in industry, there were other powerhouses and the GDP was distributed based on a percentage of the population of the world. I think back to India before colonial times. It was like, I think it was 27% of GDP because they were about 25% of the world’s population.

Mark MacLeod:
No way. I didn’t know that.

Amar Varma:
It’s a statistic. Yeah, it’s in one of Shashi Thur’s books. One of the existing politicians goes through this. But I think about it and I go, okay, what do you need to drive change in things now? You need something that’s the equivalent of the fabric that connects us as a society.

At least for my lifetime, it feels like it’s been some kind of energy, oil, and things that have been global distribution and that’s got a very tight ecosystem among producers, suppliers, markers, and all the contracts that go back and forth on it. I’m watching this next wave now of countries that are starting to merge where they have large populations, they’re able to have domestic markets on things that otherwise previously required support in America first. That’s interesting.

So, I started investing in India maybe a decade ago, I started going to China for a number of years prior to Covid, and just watching that whole place move at a pace that I’m like, okay, they’re living partly in the future. Like they’re just able to do some of these things, and I’m just watching this phenomenon now of AI being labeled as a sovereign AI.

It’s something critical to your national identity, critical to your national security, and something that will define whether you’re a player in the next chapter of global power or not. And it’s super interesting for me to watch the technology from an outsider’s perspective, looking at it as now something that should be relatively accessible to everybody.

There are now trained open-source models at scale. Android was an open-source operating system for mobile. I guess Meta now has gone and created these open-source models at scale for people that are pretty powerful.

But we all know that that’s what the front of what they’re showing is, and the war is being fought on the back of things you’re not seeing. Access to compute capability is a big deal. Access to energy, if you look at the energy requirements for these things, Microsoft went and fired up a nuclear power plant to power their AI.

Mark MacLeod:
Are you serious?

Amar Varma:
Yeah. And if you, I think we were talking about this the other day, you start looking at the CAPEX numbers.

Mark MacLeod:
Yeah, we’re going to talk about that for sure. We can get into it.

Amar Varma:
Yeah, like the capex numbers not only for the year but the quarter. When you see an 11-figure budget for CapEx next quarter from Big Tech, that’s a big number. You multiply it by four, and it ends up being a huge number across a one-year period. And I’m like, if you don’t have that money, you’re just not playing in this infrastructure potential.

And there are only a few places on earth that have the capability to do this. And once this is done and the spending is gone and the cycle is there and you can’t catch up, you’ve really missed out. And I’m watching bubble countries that have the resources but are kind of waiting and stuff like don’t wait too long.

You know, Canada’s one of those places, don’t wait too long. This is the birthplace of AI and it’s now the best export we’ve had that we’ve got zero return on. This is one of those types of things where it’s frustrating but that’s the story of Canada.

Mark MacLeod:
I remember we’ve got a good number of Canadian listeners here and I remember hearing a stat once that there were 400,000 Canadians living in Silicon Valley and high tech. You were one of them for years. I don’t know if it’s still true, but at one point every publicly traded technology company in the US had a Canadian in the executive team, which was just a giant amount of brain drain. And you’re right, there was so much AI kind of research intelligence in Canada.

You were telling me a pretty frustrating story about someone trying to immigrate to Canada. Maybe let’s tell folks about that.

Amar Varma:
Yeah. So anybody listening who has a line to the leaders of immigration, you need to get this person immediate citizenship. There’s an individual that I’ve known over some period of time, who runs a very prominent company in AI, was a very prominent researcher, worked at the largest ride-sharing company around, and is now revolutionizing driverless capabilities and for some reason isn’t yet a citizen of this country.

And I met this person recently, and they’re like, I recently was told my application was rejected. I don’t want to swear on your show, but who the heck else is getting this immigration if it’s not a person who’s highly educated, highly skilled, a prominent job creator, and just an all-around great person? I’m just shocked that this is somebody I had to hear that story. It was like, yet again, I was let down by the government and I’m just like, I’m not here to complain, if I don’t like it, I got to go fix it.

This is one of the ones we need to fix. So, for anybody listening, everybody is aligning, reach out directly. We need to get this person’s citizenship and get past this incomplete, whatever the status is of this thing.

Mark MacLeod:
Yeah, super frustrating. Do you think AI is back to this concept of sovereign AI, is this establishing some foothold in AI? Is this the thing that’s going to set countries up for the next generation?

Amar Varma:
The short answer is yes for me. The long answer is I take a lot of cues from Hollywood, so it’s a guilty pleasure. I look back to Bond villains, I look back to Terminator. I look back at some of the things we’re seeing. And I even think last year in the Mission Impossible series, it was this whole thing.

If you think the Cold War was something that just set the world stage for how the world should end up, the next version of some kind of war is this cyber security war that starts as a cybersecurity war and then evolves into something larger, where if you can win the technology game, you can leverage it into a bunch of other pieces in the political chessboard.

We’ve already seen wars being fought from computer terminals halfway around the world, that’s the thing today. Imagine those aren’t powered by people anymore, they’re powered by bots because they understand they’re humanoid robots or some version of robots that are powered by these open source models, like not trying to paint any kind of doomsday, but this is the war that’s being fought.

And when I look at sovereign levels of investment for these things, obviously they want to be at the forefront of tech. Like that’s important. We’ve seen that play out. I think if I look back 20-plus years ago, South Korea was a great example. They went and invested immense amounts of money in broadband. And look at kind of a generation later how prominent that’s become and how important that country’s become in terms of the knowledge-based economy and a lot of manufacturing.

So I think we’re seeing those groundworks being set up now. But if you look at the scale of investment and the power requirements, etc., not everybody can do this, not everybody’s going to have access to this. It’s not putting fiber in the ground anymore. It’s like getting access to the compute, having sustainable power that allows you to power this compute, and then being able to use it and harness it in usages that are important to your country.

Mark MacLeod:
So on the energy front, that feels like a strength for Canada. My dad back in the day used to install Canadian-made Candu nuclear reactors in different places around the world. And he spent a whole bunch of time actually in South Korea doing that. So this is a resource that we have.

Amar Varma:
I went in high school, grade 12 physics, we went to a nuclear power plant. I grew up maybe an hour from one and it was cool. You had the Geiger counter, you had studied the unit in advance. And when you’re young, sometimes you don’t realize how amazing it is that you’re physically close to something like this and you can actually go see it.

So for me, I was like, wow, this is really interesting, really cool. Then you watch disasters happen across the globe, in this, and then you also notice locally there’s actually a bunch of reactors that are still running and still operating and still out there. So there’s like operator error, technology error, there’s no room for error in this game.

So you got to be very precise and if this is what it’s going to take to power the next generation, like I just saw Microsoft turn on or announce that they’re going to turn back on a nuclear reactor. Like this is the start, this is the trend coming back. Total random question. Your dad now has been involved with Candu reactor installations, like he’s fine, right?

Mark MacLeod:
He’s fine. Yeah. It was a weird period of time where his hair was green but I think it was just the process of turning white. I think it was a coincidence. But my sister, her husband, so my brother-in-law both work at Ontario Power Generation now. So it’s a nuclear family. And so far they still have 10 fingers and 10 toes. So yeah, I think they’re good. And to your point about the Geiger count, they get checked. They measure that stuff on some recurring frequency.

So you brought up Microsoft and so let’s go back to that, right, CapEx and what you called it. You and I were playing tennis this week and riffing about this stuff, and we call it that kind of the haves versus the have-nots. And so Alphabet, they spent $24 billion on CapEx last year, but they announced they’re going to spend $13 billion just in this quarter alone. Meta spent $37 billion on CapEx last year. Microsoft spent $24 billion. Like how on earth does a startup compete with this?

Amar Varma:
So I think it warrants mentioning why the power requirements are so high because I think this is sometimes lost. I’m going to try and simplify it and I’m going to oversimplify it so there’s going to be some mistakes. So in the comments, please adjust if you feel like I’ve missed something.

But I happen to be a chip designer to start my career in Silicon Valley, so I understand a thing or two about compute cycles and things like that. I worked on graphics cards, which are compute. So one of the things that’s very interesting, if you look at the order of magnitude of increase in compute, it’s just gone up the order of magnitude, it’s logarithmic now.

Moore’s Law was a doubling every 18 months. We’re far surpassing that based on the measurements. The challenge with that is the power requirements to train these models have been so great. So if you think about the physical space of a data center, the compute capability required that was capable there was not.

It was big, but until the cycles and operations went way up and the power consumption leveraging technology went down, you couldn’t get enough of the compute cycles to go train these models. The second they were able to do that, the ability to do, I think Jensen put it in his 2024 keynotes. I think it’s something like the compute capability went up 300x, and the power requirements went down basically in order of magnitude. And basically by leveraging that and shrinking, shrinking, shrinking, they’re able to now train these models.

And when they started to be able to train these models, it was like years, months, weeks, days, and it just starts coming down. This is why power requirements at a very abstract level are there. And when you think of it like, hey, we want to create models, train models, use models for inference, like these are expensive operations from a power envelope and you need a lot of it.

So you’re starting to see that data centers that previously came with a small power generation capability are like, we need backing of nuclear plants. So this is just the scale, and order of magnitude thinking that is required to understand this CapEx.

So the CapEx on buying H100 from Nvidia is immense. You’re seeing it in Nvidia’s quarterly call. The backlog’s there, the demand is there. The ability to train models now is mind-blowing versus three, four, or five years ago. The fact that they’re open source now, you can get them at scale, I think that’s something that’s unbelievable.

Now the race is to make that resource available to people, to leverage it. And doing that is an arms race now and there’s only a handful of people with the balance sheet to go do it, which is why you’re seeing the immense investment. And the winners of that by proxy will be the folks, Nvidia, its suppliers, and likely even AMD and some of its suppliers as it goes through it. I’m sure there’s a bunch of people I’m going to miss in that shout-out, but you’ll see it.

But like in cloud and some of these other super scalers where there’s just a Microsoft, an Amazon, and a Google and you just can never spend to leverage against these people, you’re going to see that in this AI infrastructure and you’re seeing the investments that were like quick to happen in a lot of the infrastructure companies. The scale is now just so large that raising $500 million is an immense amount of money.

But when you hear an announcement like we’re spending $13 billion next quarter or $24 billion next quarter, you’re just going to be like it’s orders of magnitude off and it’s just like how can you win in that situation? So that’s the interesting part and the way to win is not to go after the core investment piece but to go after the next piece of the roadmaps that they have that they may not be able to fulfill internally and stuff like that. And you get big wins that way, and you’re seeing it.

But it’s just at such a scale now where if you think of infrastructure of construction, of roads, airports, these things, we’re now putting that in the AI infrastructure. It’s just unbelievable.

Mark MacLeod:
I was not a chip designer. I’m buzzword-compliant at best. And so I’m curious, like, do you take from this that anything foundational, anything horizontal at this point is just the exclusive purview of the big players and you have to kind of go vertical? Like I’m dumbing it down, but like is it as simple as that?

Amar Varma:
Yeah, it’s a good question. So I think the question is really, is that like the opportunity to go invest in this infrastructure still available? For a select few it is. The folks who know some of the tricks of how to get the 10x gains or the 100x gains. You’re still seeing companies being formed, you’re still seeing investment going into those folks.

But for the average person to go in now and say I’m going to go join like a company making a model, well, there’s going to only be the super scaled folks and maybe a handful of others, maybe even less.

The next wave now becomes, okay, train models. The next to me is now the inference. Can you go win the inference layer, and that’s where you’re seeing a lot of investment and a lot of really smart people who are spending their time on it.

The luxury and benefit of all of those people putting their effort into it is you’re seeing a lot of productivity-capable applications that are starting to come out. Fully integrated is obviously ChatGPT, but like the consumer side of winning this, is amazing resources at hand.

Seeing it now in an enterprise capability where we’re working on one of them for ourselves, kind of seeing a vertical model where it’s integrated, helping people do their job better. Not just like multiples better, but orders of magnitude faster, better, smarter, is just, okay, we’re seeing that and we’re just at the start of it.

And every vertical has some play of like, hey, we use AI to fix your workflow or ingest documents and give you this output. It’s coming, it’s here and it’s in trial, or you’re already using it, you don’t realize it. So it’s coming. Those will be a lot of the productivity gains we’ll see over the next two to five years.

So if I were to start over today or start again today, I would really be looking in that area and I’d really be looking for a knowledge surface area where I’m an expert, I have really good relationships with design partners or somebody in the industry who knows what they’re doing.

I’ll pick health, tech, healthcare. Everybody’s like, hey, doctors should have a transcription or use AI models to help with billing, coding, and these things. And it’s fun. Yeah, of course they should. But do you know enough people practicing that could help you form and kind of figure out how to optimize all those things, do you understand how the clinic works, do you understand at scale, how a regional thing, like a New York Presbyterian, how does it operate? How does it do all this stuff? Is it outsourced to a third party? Do you need to work with a third-party farm like this?

There’s a lot in here that’s super exciting to go through. Other areas I think are super interesting. As you’re seeing. I’m maybe biased because I’m an enterprise-focused person, but you go into these large organizations and there’s like $100 billion being spent on the general IT budget and nobody knows what some of the programs or some of the software repositories are doing.

There’s really an amazing opportunity to go in there and be like, okay, let us use these capabilities to tell you what the code bases are doing, which applications are powering, which stuff is like sliced-off code that has no functionality that you can stop maintaining. Like, there’s just a bunch of these obvious things that are out there.

And then the other one I talk about, if I were to ever do a service business again, is like, there’s so many applications people need, but if you could start from some base layer of a platform that’s built for half of the stuff we know we need to do and the other half is a service engagement to get it right with the customer, I think you could have infinite business right now. I’m just seeing it in some of my investments, my friends doing this stuff.

Mark MacLeod:
Fascinating. I’m curious. We’ve been talking about a whole bunch of macros. We have two questions that may be more specific to you and your company. One is, how much do you feel you need to be aware of, kind of these moving pieces in the broader environment in order to formulate your strategy for your company?

Amar Varma:
I think you need to go down the rabbit hole. You start reading stuff, you start using the models, you start talking to friends, you start contributing. There are depths and layers to this. My co-founder, Dwayne Ford, is a CTO here. He put out a piece on our site on how to think about models. At this point, unless you’re super skilled, it’s this tame, not train, and understand the parameters, understand what you’re doing with the different parameters. And he’s not the only one. Lots of people have published good things on this.

So, like, you can go read pieces like that, and understand the state of the art, but getting deep has given me one other piece that I think is important. It’s like, these models were great at consuming information and having a lot of readily accessible stuff. I say like in their cache memory equivalent, but their reasoning capability was very small to start with.

And I have four kids, so I’m watching these kids every day, every week, every month, every year, gaining motor function, intellectual function, speech, all these things. And I’m watching the inferences of these models grow in leaps and bounds. So what would have been a newborn to a toddler is like a newborn toddler, a high school student, master’s students, professionals at the top of their game to, like, emeritus-type stuff.

And then once you can get past that, you end up with something like an expert out of the box at some point. That’s kind of the bit of the holy grail of all these people working on it. But, like, I’m seeing it in the early innings. But I think we’re a long arsenal date myself. But I think we’re not one or two models away from every worker who sits in front of a computer being obsolete.

I don’t think that’s happening. We have another series on our website around leveraging AI and code, and there’s a bunch of great resources out there. I think for myself, what was really eye-opening is there’s a cohort of people who’ve written code, understand code, and they’re using these tools and they’re getting way more efficient.

Whether it started out writing test infrastructure or writing out simple things that would have been tedious tasks. Okay, great. But then there’s this category of people who’ve never written code and never will write code.

And now you have these, like you had these no-code tools for a period of time. Now you have these tools that are like writing code, and if you know how to prompt and understand things, you can get it to write things in production, things with tests.

It’s really quite intriguing to watch because we’re trying to leverage some of these tools every minute of every day. Because every time we do something, I go, can we automate it? Can we take our knowledge and go to a higher order in our business and our understanding? And I think if you evaluate everything in that automation capability, you’ll really start to see efficiencies and gains in these things.

The follow-up question for me always is, so what’s the quantifying the gain? I think for us as a skilled group that’s worked together over five companies, I’m probably seeing productivity gains in at least a third, 33-35% on a few things or most things on average.

And then there are some places where we’re getting like 100x, like 100 times or 100%. We’re just getting automation of tasks or things that are just happening. So it’s very interesting if I had to compare maybe to a decade ago for what we’re outputting, what would that have been? It’s probably 3x that. It doesn’t correspond with the third productivity game, but just we’re able to get the tools more efficient as well, and all those things. So there’s a bunch of things.

Mark MacLeod:
I’m curious, you’ve raised money recently, if there was a pie chart of the mantle story, how much of that story was AI-centric?

Amar Varma:
In terms of fundraising? None. Because we raised it in the fall of 2022, maybe two years or so. It wasn’t really a driver. The wave we caught was obviously the models coming through and providing us with a really good opportunity for a lot of automation. Now when we talk about our products, in all of our interactions with customers, partners, and prospects, we are very upfront. We’re trying to automate everything.

And we’re very clear we’re not using your data to train anything. Your data is not going into the models and being trained in any capacity. And it’s very important that people understand that data piece for obviously understanding piece of mind and a bunch of other things. But they really are encouraging us to be like, okay, you can do that. And they’re seeing it reflected in speed, time to market, and then cost and quality. They’re just seeing a lot of those things getting tightened and tightened and tightened.

Mark MacLeod:
Maybe a random aside, but again, I’m completely unencumbered by any technical knowledge whatsoever. What role do D-Wave and quantum computing have to play here at all?

Amar Varma:
I’m probably likewise.

Mark MacLeod:
Does that lower the burden? I feel like we’re brute-forcing our way through things now. Does this change that at all?

Amar Varma:
I see, like, well, I’ll likewise say I’m not fully qualified to comment here. I will say quantum computing has always been an interest of mine, and just in terms of reading and curiosity, I never felt skilled that I could go work in it because it felt like a pretty company to build or something like that.

If you plot the gains and you plot the charts, it’s kind of the holy grail. There are some folks that are out there now with commercial opportunities that seemingly are starting to crack the nut. Ah, it’s exciting. I would love for it to be here today because this…

Mark MacLeod:
It kind of scares me, to be honest. But yeah, it does sound cool.

Amar Varma:
It does. But I haven’t gone and checked out the interviews with Jetson and others on it, so they probably have a much better take on where they think quantum computing is versus today’s state of the art and where they see an intersection of this stuff.

I will just comment on a few things. We had done a research chair in AI and bioinformatics as a family at the hospital, and this was six-plus years ago now, maybe closer to seven. And I was like, hey, if you guys could leverage AI to uncover things quicker, etc., like molecules that shouldn’t be tried as candidates. Like that’s going to be revolutionary in terms of speed to market of things. And you’re just watching how many days did the COVID vaccine take? Three?

It was like they just were able to do stuff like that. And I was like, that’s the kind of stuff where you can see generational changes in outcomes in terms of knowledge of information and leveraging it. You know, we’re seeing now a combination of CRISPR CAR T cells. There are a couple of other alpha folding problems that have now been solved. Not just the first, but the second, and the third one is solved now.

You could start seeing these types of things that are jumping, and I think in terms of isolating reasons and root causes for defects or aging or longevity, like you’re starting to see a bunch of people spending a lot of time and money in this area. That’s super exciting to me. If that can be unlocked by quantum computing, like we should all be working on that, because that would be cool.

Mark MacLeod:
Certainly into learning how to slow down aging. I have a vested interest in that. Maybe the last big topic, you and I were marveling at the set of funds that General Catalyst announced last week. $8 billion in new funds. These guys have been around since 2000 and for anyone tracking this is like a 40% increase in their assets under management. Like, a giant expansion. What’s your take on that?

Amar Varma:
Yeah, so Mark, I have been around the VC game for a long time. I joined my first firm, which was early in the mid-2000s. It was a billion-plus dollars. It was giant. Now today that is not a giant firm. And I’m watching, I think Lightspeed’s gone and done it and recently has gone and done it. Sequoia has gone and done it, all the big firms have gone and done it and raised billions and billions of dollars.

And when I saw General Catalyst do $8 billion, I was just like, what are they investing in? And to me, there has to be some thesis of like we’re going to go invest in some kind of public market equivalents.

Like the good videos just outperformed everybody. It’s just like if you just had the index versus some venture portfolio, I think the returns were ahead over the 10-year time period. And you’re just like, how does that feel for paying 2 and 20 or 3 and 20 or 3 and 30 or whatever deals?

Mark MacLeod:
And having your money tied up for at least 10 years.

Amar Varma:
And your liquidity premium for that too. So it’s just like, I’m really amazed at the scale of money going into this. It’s really impressive to see. I’m just watching to see, like, when you have to manufacture opportunities to put money to work, it feels a little less authentic.

Mark MacLeod:
Yeah, I feel like SoftBank did that.

Amar Varma:
SoftBank did. I watch a lot of PE people over the last decades with large amounts of capital. They weren’t privacy manufacturing outcomes, but they were able to go find mispriced assets at scale and put a formulaic here’s how we’re going to buy it, here’s how we’re going to tighten all the screws and we’re going to create this rosy picture at the end and somebody else is going to buy it. Like, I’ve watched them do that at scale. That is not the craft or the experience of most people in startups and venture. Like, that is just not the DNA of the person. That is not apprenticeship.

So I’m interested to see the openings of which partners start joining, and which partners start leaving. Because if you’ve worked at a firm and you’ve been at a $300 million fund for 10, 12 years, having three or six or $10 billion is a much different game. Is that why you got into this business? Was that your passion? Is that where you really see your best superpower being flexed? Like, I just don’t think that’s true. But then there are some who’ve really adapted to the times. They’ve become, they’ve become great investors across decades at scale. And those are very few. I’m like, how old is the Vino coastline?

Mark MacLeod:
He’s still at it. Still a baller.

Amar Varma:
I remember Cisco buying Soren for it was $6.7 billion when I lived in Silicon Valley. I was like, oh, my God, this is amazing. Like, that was him. That was decades ago. He’s still doing it, right? But if you put in a $6.7 billion exit and the $8 billion fund plus growth vehicle, I don’t think you get all the money back.

Mark MacLeod:
You know, I had a $50 million fund back in the day, which I would need to turn into 150 million for the math to work and to deliver on the promise that LPs expect. And I was stressing out, figuring out how to do that. Let’s say you own 10% of the companies. You have to generate one and a half billion outcomes. In order to get the 150 million for the venture fund math to work, I was saying I would literally just have a coronary if I was responsible for generating a return on $8 billion.

Amar Varma:
What’s happening now, and I’m watching this is when you watch these portfolios, a lot of the times the best companies out of the gate get picked up early. And when they get picked up early, the fund’s like, shoot, you were supposed to be the fund maker. Let me give you another round. Let me preempt this deal. Let me give you more money at whatever valuation they can wrangle that’s proportionate with the amount of money and go for that next rung. And when you see that, you’re like, okay, there’s a defined probability-weighted outcome the entrepreneurs and investors have to get in on together. And it’s like, okay, I’m not, I’m not exiting here. I’ve gotta, I gotta take the money at this point and 10 exits from here so that I can make your fund, you can fundraise on this. And I’m taking all the risks. This is why you’re seeing a flurry of these secondaries that have happened over the last number of years.

Mark MacLeod:
I was just gonna say that. That, to me, is just giant secondary territory.

Amar Varma:
And they’re like, hey, you can’t get your head around being an entrepreneur. Here’s a huge secondary. It’s fine, carry on and keep doing you. And it’s just like, okay, that’s fine, it’s the game. But it doesn’t feel like the acquiring capability is out there to go consume things at scale right now. Especially with Big Tech in a purgatory zone, right, with DOJ. But it’s just like with that wheel not turning, all of this money’s going in and it’s kind of creating this cascading effect. If there’s no returns, there’s no recycling of that money. And if you look at DPI, I’m in dozens of funds, like, I wish there was just more DPI. They’re not that funds, they’re just, they’re in a vintage where there’s not enough exits.

Mark MacLeod:
So true. Yeah, so true. Do you connect the dots? I think you kind of said it here with kind of doubling down on the ones that are winning. But connecting the dots here between these $8 billion funds and what we were talking about earlier, and this just dynamic of haves versus have-nots, and just the magnitude of the dollars that it takes to play now.

Amar Varma:
Yeah, one thesis that we were referring to the other day is that your startup is now more of an infrastructure-type deal than a whiz-bang software or system that you could put in a network or something like that. The venture investment here is something of an infrastructure scale that requires dollars that are commensurate with having not an $80 million fund, but an $8 billion series. And we’re just picking on eight because that’s what they announced. But they commenced with that kind of scale. If it’s working and you’re putting in money, you have a clear ROI for each dollar spent, okay, got it.

But there are a lot cheaper costs of capital when you start getting at that scale that like don’t have venture people behind them. It’s interesting to see. I have a purview through my investments on a lot of this. But a lot of the backers are really important in institutions around the world. And if they’re directionally wrong and they’re wrong at scale, watch out.

Mark MacLeod:
I worry about that, to be honest, and it’s probably out of scope and out of time to unpack it fully, but we just came through a drunken orgy of investing in 2021 and now I feel like this notion of deploying 8 billion forces, artificial behavior and is sort of just like SoftBank all over again. I have to artificially and prematurely anoint kings, and I worry that it just doesn’t turn out well.

Amar Varma:
Yeah, let me flip one thing for us here. Like, I don’t want to make it feel like we’re of a different generation. It was harder in archive.

Mark MacLeod:
Those old guys, they don’t get it.

Amar Varma:
Well, it’s very much. There’s a lot of promise, a lot of opportunity. I want to be very clear that that is a thing. I just feel like as the landscape is shifting and the game is changing and the order of magnitude of the scale is going up by another order of magnitude, it’s like, let’s not forget what makes these ventures successful.

And it’s the people who run them and work through them that they are putting their entire life and time into it. So there’s a lot of optimism. I think really now we’ll see some outcomes over the hill here, but let’s just be mindful of the scale of those outcomes. Is there a world in which you start seeing multi-hundred billion dollar exits?

Mark MacLeod:
Fascinating stuff. As I expected, I knew we would run out of time before we could get through all of these topics, so I guess we’re just going to have to do it another time. And we’re about to enter the never-ending Canadian winter, so maybe the next time has to be in person with a single malt and a smoking jacket or some kind of dipshit thing like that. But Amara, fascinating stuff, and I really enjoyed the conversation.

Amar Varma:
Thanks for having me. Always a pleasure and see you soon.

Mark MacLeod:
See you soon.

Hey, thanks for listening to the Startup CEO Show. If you’d like to connect with me, be sure to visit my website at markmacLeod.me, or follow me on LinkedIn at The Mark MacLeod, or X account @markmacleod_, and if you want to tune in again next week, be sure to subscribe on YouTube, Spotify, Apple, or wherever you get your podcasts. See you next time!