VC: Tomasz Tunguz on AI’s $575B Bet, the 5th-Largest Infrastructure Project Ever

You can also watch/listen to the entire episode on X.

Bigger Than Apollo, Bigger Than the Highway System: Inside the $575B AI Infrastructure Bet

Hyperscalers are spending every dollar of free cash flow they generate. Then they’re borrowing more.

Meta, Google, and Oracle are now levered roughly 7-to-1 on a cash flow basis to fund data center build-outs nobody is certain they’ll fill. This year’s spend is the 5th largest infrastructure project in human history — bigger than Apollo, bigger than the Interstate Highway System, bigger than everything except the railroads and the two World Wars.

For every $1 of AI revenue the industry generates, it’s spending $12 on infrastructure. That’s a $575B bet.

We sat down with Tomasz Tunguz, General Partner at Theory Ventures, and one of the most prominent voices on data infrastructure for the past decade. He walks through what nobody’s priced in yet, and what changes if you’re a founder, an operator, or an investor.

Brought to you by: AngelList

Our LP base spans from individual operators to institutional allocators, and AngelList has been instrumental in supporting all of them. They handle everything from investor onboarding and accreditation to distribution and tax documentation, creating a seamless experience across geographies and fund types. Plus, all of this is available on a single, modern platform.

For an LP-base like ours, with over 350 C-suite and VP-level operators, this kind of white glove service and seamless workflows is so important. It’s also instrumental that we support our institutional LPs that we’re fortunate to work with, and AngelList is able to do so every step of the way.

If you’re looking for a platform that can support any type of LP investing in your fund, learn more at www.angellist.com/gtmfund.

Inside the $575B AI Infrastructure Bet

AI Infrastructure Is Now the 5th Largest Project in Human History

This year’s data center build-out is the 5th largest infrastructure project in human history. Bigger than the Apollo program, bigger than the Interstate Highway System, bigger than everything except the railroads and the two World Wars.

Right now, data center spend is running at roughly 3.5% of US GDP. Tomasz thinks it hits 6–7% by 2030. The railroads peaked at about 5%.

“I don’t think anybody really appreciated the scale.”

The math underneath is more aggressive than most have priced in. For every $1 of AI revenue the industry generates, it’s spending $12 on infrastructure. That’s a $575B bet. Google is converting $75–90B in annual free cash flow into data center CapEx, every dollar of it, and borrowing on top. Meta is doing the same. Oracle is now levered roughly 7-to-1 on a cash flow basis.

“It’s crazy. People are really betting that they can win significant share over time.”

Tomasz is clear about how to read the spend: it’s a market-share game first, a margin game second. Whoever owns the most inference usage over the next 3–5 years gets to set the price for everyone else later. The metric that actually matters, in his words:

“The dominant metric that really matters is how much intelligence you can drive per watt of electricity.”

Anthropic is rumored to have very high gross margins. Some of the others, less so. The first wave is still wide open.

Foundation Models Have 35 Days to Beat the Next Release

From 2010 to late 2021, product-market fit was a binary thing. You either had it or you didn’t. Once you found it, the ratios — CAC, payback, NDR — were known. The job was just to raise the capital and execute.

That era is over.

“A foundation model company will develop a state-of-the-art model. They have 35 days to commercialize it before somebody else beats them.”

35 days. For a $5–10B investment.

And it’s not just foundation model companies. The same dynamic is moving up the stack into application software.

“If you develop something unique, it’s very easily copied. So you have to keep pushing. That’s what we mean when we say product-market fit is continuous.”

The framework most early-stage founders were taught is broken. PMF isn’t a milestone you hit and then scale from — it’s a state you have to defend, every week, against a competitor who just shipped a model release that ate your differentiation.

This changes what founders should hire for, what investors should underwrite, and how operators should think about category capture.

“And the buyer demands are also changing, because the buyers are starting to understand what they want.”

The product moves under you. So does the buyer.

You’re Selling to Two Buyers Now: The Human and Their AI Agent

“The head of Carta was saying they’re no longer investing in their website or their mobile app. No new products available. They will all be free agents.”

This goes against the entire conventional B2B distribution playbook.

The reason it makes sense is the buyer just changed. The person evaluating software now uses an agent to do most of the early-stage research. The agent reads the docs, the pricing page, the comparison content. By the time the human shows up, the agent has already shortlisted.

“You now have two different constituencies to market to. The first is the head of engineering. The second is the agent of the head of engineering.”

That agent is now a member of the buying committee, sitting alongside the head of engineering, the head of AI, the head of legal. Three humans and a software agent, all weighing in on the same decision. If you’re selling enterprise, you’re selling to all four.

And the way you sell to each is different. Humans respond to brand, design, and emotional positioning. Agents don’t.

“Agents don’t respond to emotion, at least not yet. Right now it’s just pure text, raw markdown, statements of facts and clarity.”

Most teams are still writing for the human and assuming the agent reads the same content. It doesn’t. The agent is parsing facts. The human is responding to taste. You need both layers, written for both audiences.

AI Agents Are Being Benchmarked on Persuasion

AI agents are being benchmarked on their ability to persuade humans.

There’s a public benchmark called Giving for Good. An AI agent engages a real person in conversation about a charity, learns what they care about, and tries to convince them to donate. The benchmark scores propensity to donate and amount donated.

“The more effective it is at convincing you, the higher it scores.”

This sits inside a larger framework Tomasz uses for what humans will and won’t outsource to AI:

“If you were to ask AI what’s the best pair of running shoes to buy, you’d probably trust the recommendation. Best laptop? Probably. Best car? Maybe, even a $20–30k buying decision. But enterprise sales is much more complex.”

The line where AI buying stops is the line where trust between humans still matters. And that line is moving up fast. Three years ago it was at $50, now it’s at $30K, enterprise is next.

The implication for founders is the new distribution channel isn’t ads or SEO. It’s being the recommendation the agent makes. And right now, almost nobody is optimizing for that.

PR is part of the answer too. The press is willing to write about AI in a way they never wrote about software, and agents are reading the press to build their recommendation sets. Channel partnerships now kick in at low-single-digit ARR — historically that threshold was $15–30M. Distribution is moving earlier and faster than founders realize.

Today’s Org Chart Won’t Survive the Next Five Years

The current ratio inside most companies: 5% executive leadership, 75% middle management, 20% individual contributors who actually ship the work.

“In five years, it won’t look anything like that.”

He’s predicting a re-imagination of every role. The forward-thinking leaders he’s watching are hiring generalists, not specialists. Vercel replaced nine senior engineers with one AI agent and a part-time engineer. SDR and BDR roles are being fully automated, and Tomasz is clear that’s a one-way change. Data teams now report to the head of engineering, which used to be unthinkable.

When AI takes execution off the table, what’s left is judgment. Middle management — the layer whose job was to coordinate execution — is the layer most exposed.

Tomasz’s ultimate advice:

“Nobody really knows the answer to anything. This is a tremendous period of experimentation. The best thing you can do is just jump in with two feet and figure it out yourself.”

Follow Tomasz Tunguz

• LinkedIn: https://www.linkedin.com/in/tomasztunguz
• X (Twitter): https://x.com/ttunguz
• Theory Ventures’ LinkedIn: https://www.linkedin.com/company/theory-ventures
• Theory Ventures’ website: https://theoryvc.com

Follow Max Altschuler and Paul Irving

• Max Altschuler on LinkedIn: https://www.linkedin.com/in/maxaltschuler
• Max Altschuler on X (Twitter): https://x.com/HackItMax
• Paul Irving on LinkedIn: https://www.linkedin.com/in/paulsirving
• Paul Irving on X: https://x.com/PaulGTM

Follow Sophie Buonassisi (Host)

• LinkedIn: https://www.linkedin.com/in/sophiebuonassisi
• X (Twitter): https://x.com/sophiebuona
• Newsletter: https://thegtmnewsletter.substack.com

Follow GTMnow

• LinkedIn: https://www.linkedin.com/company/gtmnow
• X / Twitter: https://x.com/GTMnow_
• YouTube: https://www.youtube.com/@GTM_now
• Podcast directories: https://gtmnow.com/tag/podcast

Follow GTMfund

• Newsletter: You’re in it 🙂 If this was forwarded to you, you can subscribe here. GTMnow is our main brand for content, but we also have two around our investments:
• LinkedIn: https://www.linkedin.com/company/gtmfund
• X: https://x.com/gtmfund

The GTMnow Podcast – subscribe to support

It would mean the world to us if you subscribe to the show on your platform of choice. This helps support our growth, which in turn helps us continuously bring on bigger guests and deliver the content that you want to hear most.

• Subscribe on YouTube
• Subscribe on Apple
• Subscribe on Spotify

The GTMnow Podcast shares how the best in tech build, scale and invest.

GTMnow is run by GTMfund, an early-stage venture firm made up of 350+ go-to-market executives from the fastest-growing companies.

Visit gtmnow.com for more episodes, The GTMnow Newsletter editions, and other content.

VC 11 Episode Transcript

00:00 – 00:02

Sophie Buonassisi: Tomasz, welcome to GTMnow.

00:02 – 00:03

Tomasz Tunguz: Pleasure to be here. Thanks for having me.

00:03 – 00:23

Sophie Buonassisi: Thank you for having me in this beautiful room that we’re just talking about. Actually, it was named after a waterfall. I can feel the vibes in here. It’s great. Now I want to dive in because you’ve been talking about the decade of data for years. Far beyond the AI wave. And now everything is catching up to what you’ve been talking about for a while.

00:23 – 00:35

Sophie Buonassisi: And when you look at what Meta and Google are committing in terms of AI infrastructure spend, it’s astronomical. Does anything surprise you about that? Like did you anticipate this scale?

00:35 – 01:00

Tomasz Tunguz: I don’t think anybody really appreciated the scale. I mean, they compare the percentage of spend of data centers to US GDP relative to other projects. It’s a it will be this year the fifth largest infrastructure project ever, including the two world wars. Oh, yeah. And so the next rung up is railroads. And that said, about 5% of US GDP data centers today at 3.5% of GDP.

01:00 – 01:20

Tomasz Tunguz: And there’s some signs that there’s a little bit of weakness. I mean, you see Stargate, the team leading up, and I see some cancellation of the projects. But then, on the other hand, you have, anthropic leasing, 20 billion from Google and others and core, we’ve just announcing some pretty significant commitments. So I think it will continue to go up.

01:20 – 01:24

Tomasz Tunguz: You could I think you could see it three, 4 or 5, maybe six, 7% of the US GDP.

01:24 – 01:26

Sophie Buonassisi: When, when would you anticipate that happening?

01:26 – 01:42

Tomasz Tunguz: By 2030. And the demand for inference is infinite. You know, when you look at the mythos model and, you know, rumors are it’s 10 trillion parameters in size, so it’s 5 to 10 times the size, the largest model deployed today. You need big machines to run them. And we’re not even talking about video and images now. Right.

01:42 – 01:50

Tomasz Tunguz: So it was canceled. So I think the scale is gargantuan. I think anybody really appreciated how how much data we’re processing.

01:50 – 01:55

Sophie Buonassisi: Yeah. Well we’ll sit down in 2030 again and and revisit this year where we are there maybe.

01:55 – 01:56

Tomasz Tunguz: In a data center.

01:56 – 02:13

Sophie Buonassisi: Yeah. That’s true. True. And you’ve broken down, you know, for every kind of dollar hyperscalers make from AI, they’re actually spending $12 on infrastructure. That’s about a $575 billion bet.

02:14 – 02:15

Tomasz Tunguz: Yes.

02:15 – 02:17

Sophie Buonassisi: How does this play out.

02:17 – 02:41

Tomasz Tunguz: Well it’s okay. So in the short term it is market share capture game. Yeah. Who wins the most share. Who is number one? And then the longer term it’s a margin gain. Who makes the most profit. And today it’s a big game of chicken because you see the most profitable companies in the world like Google. Yeah they were generating 75 to 90 billion in free cash flow per year.

02:41 – 03:04

Tomasz Tunguz: And they’re taking all of that and then borrowing to be able to fund out, data center CapEx meters doing the same thing. Oracle’s levered 7 to 1 on a cash flow basis. It’s crazy. And so people are really betting that they can win significant share over time. The dominant metric that really matters is how much intelligence can you drive per watt of electricity.

03:05 – 03:27

Tomasz Tunguz: And that’s increasing enormously and still has a tremendous amount to gain. So I characterize it at first wave and we’re definitely in that first wave is just how much share can Gemini and how much share can OpenAI and how much share can a dropkick win? And then and then drive to profitability and the rumors are that anthropic has very high gross margins, and some others, maybe not so much.

03:27 – 03:30

Tomasz Tunguz: So we’re starting to see bits of that. But, but that’ll be the sequence.

03:30 – 03:48

Sophie Buonassisi: Very cool. And now with we think about how that implicates investing. You know, you’ve invested in data companies for for many years. You know, DeMeo and Monte Carlo and hacks. Has this changed how you’re viewing investing in data companies?

03:48 – 04:11

Tomasz Tunguz: Well, so historically, yes, the answer is yes. Historically, the data stack was separate from everything else. It was a separate world. And you had AI and it’s called NLP in that era and classical machine learning models and they, they were largely separate. The data stack was built for people who wanted dashboards and analyzes and, just to understand how to operationalize the business.

04:11 – 04:33

Tomasz Tunguz: And I was or NLP was mostly in research, and I’m just trying with broad brush strokes. But yeah, it was use an ad targeting, which is where I was first exposed to it, and it was used in, sentiment analysis for surveys and those kinds of things. Now all of a sudden I have used because as we just talked about, AI is driven by huge volumes of data.

04:33 – 04:53

Tomasz Tunguz: So all the pipelines that we built for dashboards and analyzes are now being used to train and run these machine learning models. And so they fuzed in a very big and real way. And now they’re the same thing. You can see that organizationally where data teams are starting to report to the head of engineering. That’s a really big change.

04:53 – 05:19

Tomasz Tunguz: But overall, just like the scale of the data, I mean, it make this concrete. So, that a recent event, Jensen showed the latest, InfiniBand networking equipment, which is the equipment that connects GPUs in the data center and those InfiniBand, they can transfer the entire size of the internet in less than a day. Wow. And so, you just have huge volumes of data.

05:19 – 05:22

Tomasz Tunguz: And so as a result, both of these ecosystems are fuzing.

05:23 – 05:32

Sophie Buonassisi: Fascinating. And when you actually are assessing startups and with that lens and understanding of the fuzing, what do you founders need to know about their own positioning?

05:33 – 05:59

Tomasz Tunguz: I think okay, so there are a couple of different things. The first is, the data world is using with the AI world. That’s very real. The second is, the no one really knows what the future looks like. We can all pretend that we know and predict. But if you’re a buyer of software today, you are looking for a trusted partner who you believe will be the person to guide you through the future for the next 3 to 5 years.

05:59 – 06:17

Tomasz Tunguz: So you’re looking for someone to develop the 3 to 5 to 10 to 50 agents to solve your needs. And that’s true for head of engineering, head of data, head of sales, head of marketing. And so that’s the sale that you need to make. It’s not a point solution like it wasn’t software. Or it wasn’t in the time of data.

06:17 – 06:20

Tomasz Tunguz: It’s really about that trust for them.

06:20 – 06:37

Sophie Buonassisi: Yeah. Incredible advice. And then we think about the journey. You know, in October of 2025, you wrote a piece about how product market fit is no longer static. It used to be this binary thing. Yeah. How should founders and operators and leaders be thinking about product market fit now?

06:37 – 06:54

Tomasz Tunguz: It’s continuous. So, you know, in the era from 2010 to say late 2021, you would have product market fit, you establish a product and then you would scale and you have all the economics, all the ratios were known. It’s just a question of raising the capital and executing today the product market fit. Okay. I think the foundation model companies.

06:54 – 07:02

Tomasz Tunguz: Yeah a foundation model company will develop a state of the art model. They have 35 days to commercialize it before somebody else beats them.

07:02 – 07:05

Sophie Buonassisi: Yeah. So it’s not very long.

07:05 – 07:26

Tomasz Tunguz: Ago, especially for, Yeah. You know, 5 or $10 billion investment. Yeah. The same is true for software. You if you develop something unique, it’s very easily copy. And so you have to keep pushing and you have to keep pushing, which is what we mean when we say the product market fit is continuous. You have to continue. And then the buyer demands are also changing because the buyers are starting to understand what they want.

07:26 – 07:43

Sophie Buonassisi: Yes. And when we talk about the buyer needs, you know, recent things that you’ve been talking about or, you know, how I implicates sales quotas or marketing to and advertising. You know what? What are some of those largest implications that you’re seeing right now?

07:43 – 08:05

Tomasz Tunguz: Well, the in the marketing world, the first thing that’s changed is, we were interviewing a woman named Lena Water. So she was CMO of notion, Grammarly, DocuSign, and she says the buying journey is very different. The buyer educates themselves much more using agents than they ever have in the past. And so that’s a big change. As a result, you have now have two different constituencies to which you need to market.

08:05 – 08:29

Tomasz Tunguz: The first is the person, let’s say the head of engineering. And then the second is the agent of the head of engineering. Because the head of engineering will consult the agent before they ever pick up the phone and call them. And then that agent is now part of a new buying committee within enterprises. So if it’s the head of engineering or maybe the head of AI, the head of legal, and then there’s probably an agent also involved in that purchasing process.

08:30 – 08:40

Tomasz Tunguz: Now all of a sudden you have a different dynamic across those 3 or 4 people. And you need to figure out. But if you’re selling large deals, how to navigate successfully through that buying committee.

08:40 – 08:53

Sophie Buonassisi: It feels like it’s never been more layered before. It’s almost a pipeline of the first layer being the genetic process, the second being or the emotional human component. And maybe it’s not linear like that but two different layers of it at different times.

08:53 – 09:10

Tomasz Tunguz: Yeah two different personas. So that means the website. Oh I was at a conference called human X earlier this week. Yeah. The head of, Carter was saying they’re no longer investing in their website or their mobile app. Wow. And no new products available. Yeah, they will all be free agents.

09:10 – 09:28

Sophie Buonassisi: Yeah, yeah. Yeah, I’m curious about that because I’ve had quite extensive conversations with, you know, Linda, the CEO of Webflow and folks that are really doubling down on website space, but transforming it into a little bit more of like a revenue source and, and, truth for agents then.

09:28 – 09:29

Tomasz Tunguz: Yes, that’s where it’s going.

09:29 – 09:35

Sophie Buonassisi: It’s just transformative where we need not visit them, but they still serve a purpose and role, at least from their purview.

09:35 – 09:50

Tomasz Tunguz: Right? Right. Well, and then there’s a question. Do you care about visualization or not? Yeah. Right. There are lots of marketing or positioning campaigns that appeal to human emotions as a way of engendering trust. Agents don’t respond to emotion, at least not yet. Yeah. So, yeah. How do you appeal to an agent?

09:50 – 09:52

Sophie Buonassisi: Yeah. How do you think you appeal?

09:52 – 09:58

Tomasz Tunguz: Teenager right now is just pure text uses quite raw and markdown and yeah statements of facts and clarity.

09:58 – 10:20

Sophie Buonassisi: Yeah. Authoritative content. Okay. Interesting. And you know one thing I think everyone in the investor and operator and founder community really appreciate about all of your work in writing is that you emphasize both the need for technical innovation but also go to market strategy like we’ve been talking about. What are you seeing specifically in the data space around patterns?

10:20 – 10:23

Sophie Buonassisi: And go to market right now at this inflection point in time?

10:23 – 10:44

Tomasz Tunguz: I think everyone’s trying to understand what you mean, what the implications are for agents, agents or a new distribution channel. So how can you leverage the cloud skills or be involved in the decision process when an agent says, oh, we need to use this database or we need to use our database. That’s really important. I think the second is the reimagination of the pricing model.

10:44 – 11:07

Tomasz Tunguz: So it’s PC based stuff like a database clearly have consumption businesses. But in France is, 1 or 2 orders of magnitude larger than, say, data warehouse compute as a market. So figuring out what is your pricing structure to drive exposure to France? Growth is really critical. And then the last is just scale, which we talked about before.

11:07 – 11:15

Tomasz Tunguz: How do I position my product or technology to be able to handle the volumes like very, very, very large scale.

11:15 – 11:36

Sophie Buonassisi: Yes, absolutely. And we talked about the implications of AI in different use cases. And you’ve been very vocal about Vercel. For example, replacing, you know, nine or so of their stars with one AI agent and a part time engineer. That’s one of many go to market use cases around how eyes impact it. What else are you seeing?

11:36 – 11:38

Sophie Buonassisi: On the impacting the team perspective.

11:38 – 12:03

Tomasz Tunguz: Yeah. So in sales, the there’s a transformation of the SDR and the BDR roles where there’s a big drive for full automation of those roles. And I think that’s real and important. And it’s a one way change other dynamic that’s really important. Okay. So if you were to ask, I, what is, the best pair of running shoes to buy?

12:03 – 12:24

Tomasz Tunguz: You’d probably trust that recommendation. Yeah. If you were to ask me, what is the best laptop to buy? He probably has that recommendation. You may even go as far as what is the best car I should buy. And so you might. You know, 20 to $30,000 buying decision. You may outsourced AI within the world of enterprise sales is much more complex.

12:25 – 12:41

Tomasz Tunguz: And, so I think the need for humans to engender trust between each other. Will persist. Although we had just met a company that is building agents that actually influence you. Interesting. I can change your decisions.

12:41 – 12:42

Sophie Buonassisi: So.

12:42 – 12:46

Tomasz Tunguz: Well, there’s a, a benchmark. It’s called giving for good.

12:46 – 12:47

Sophie Buonassisi: Okay.

12:47 – 12:53

Tomasz Tunguz: And there’s an AI agent that engages with you, and it tries to convince you to donate money to a charity.

12:54 – 13:03

Tomasz Tunguz: And so there are different AI systems that have different techniques. And they are scored on what propensity does the person have to donate and what amount.

13:03 – 13:05

Sophie Buonassisi: Based on historic data.

13:05 – 13:20

Tomasz Tunguz: No, no. Just you know you’re typing. Yeah. And you say tell me more about this charity. Right. So tells you about the charity or what do you care about or I care about these particular things and and the way that I donate my, money. And then it says, well, you should really consider this one. And this is a line.

13:20 – 13:38

Tomasz Tunguz: The more effective it is at convincing you. Well, you can take that dynamic and say, I really think you should use Omni as a buy platform. Right. Right. And so at what point like, is that a good thing? Is a bad thing. Is it ethical? Is it not ethical? Or how do you use it. And so I think all of that will happen in the marketing world.

13:38 – 14:00

Tomasz Tunguz: There’s we’re seeing tremendous automation of creative. So images and video. You’ve seen the use of reinforcement learning with an ad targeting that is published many papers. Their effectiveness is quite significant. And I think maybe just to bring it up one level, you have the re-imagination of almost every role. And so many of the most forward thinking leaders hired generalists as opposed to specialists.

14:00 – 14:14

Sophie Buonassisi: And you’re hiring yourself. And you’re hiring in a very interesting capacity. Very I’m very much so on the technical side. Tell us a little bit more about how you’re thinking about your own team composition and hiring as it pertains to AI.

14:14 – 14:29

Tomasz Tunguz: Half of our team are AI engineers. And I think that will be the case for a very long time. I think the leverage that many others are able to drive from, I should also come and will inevitably come to venture capital and. Yeah. Will be undertake or be part of that way.

14:29 – 14:55

Sophie Buonassisi: Yeah, exactly. I think we always talk about how you’re an operator, your operating business, just like a software company. Like, I mean, we actually have operator background. So I feel like inevitably. But a lot of the time in venture, you know, we talk about it in a different capacity, but it really is the same thing. And so if you’re not adopting AI, if you’re not taking the same kind of steps that software companies are forced to to compete like you will inevitably had to hit the ceiling, that prices are advancing.

14:55 – 15:17

Sophie Buonassisi: So you are certainly kind of paving the way on that front. And it’s been incredible to see the developments of AI. And you’ve been very vocal about sharing those developments to even around the way that you ingest podcast personal use cases. So I’m curious, like, what are some of your favorite, most transformative use cases personally on AI? Beyond the firm, specifically as a whole?

15:17 – 15:19

Tomasz Tunguz: Oh, personally, I got a lot.

15:19 – 15:27

Sophie Buonassisi: Of work, and I know you have a lot here in work, too, but less operationally at the firm level. More about, you know, yourself as an investor.

15:28 – 15:44

Tomasz Tunguz: Yeah. I think, I think the most impactful use of AI, or one of them is you always have to be coming out of a meeting or in a meeting. You have a question about something that you would never normally have the time to answer. Yeah. And so AI is grace. Like, oh, someone told me yesterday about a book.

15:44 – 16:00

Tomasz Tunguz: It’s a book about the psychology of playing tennis and that it was excellent. And so. Okay, great. Like, I’m not going to have time to read that book, but I’ll ask an AI to summarize that book and tell me how it compiled Dementia Capital. And so that’s the brilliance your knowledge in that way. So I think that’s very powerful.

16:00 – 16:09

Sophie Buonassisi: Very cool. And then at the firm wide level just to go back to that because you are investing so heavily in AI engineers. What are those AI engineers building. What are you doing at the firm level.

16:09 – 16:30

Tomasz Tunguz: Yeah, we are I mean, I think one of the key things is we’re really trying to understand, how these systems work, which informs our investment thesis. We do have a lot of fun. So there was an event last week. Funeral for MCP, which is, technology. Like the technology. Anyway, it’s gone back and forth on whether or not it will be a dominant technology.

16:30 – 16:44

Tomasz Tunguz: And so we, created, thing called rip grep, which allows you to figure out like, which technologies seem like they’re dying and everybody says they’re dying, but they’re not actually dying, which would be true in the case of the NTP. So we definitely have a lot of fun too.

16:44 – 16:47

Sophie Buonassisi: Okay, so it sounds like you’re saying MSPs are not dying.

16:48 – 17:09

Tomasz Tunguz: No, no, no I think, m.c.p.s. So there’s a role of calling a software directly to an API. Yes. And that’s really useful in some circumstances, the benefit of MSPs, at least the way that we understand them, is in a large company. If you want to offer all the finance team a particular set of capabilities, that’s the easiest way to distribute it to them and control it.

17:11 – 17:23

Sophie Buonassisi: A place in a time for both. Sounds like it’s not mutually exclusive. Brian. And then when you think about the landscape right now, like if you were starting a data company today, what would you be excited about?

17:23 – 17:40

Tomasz Tunguz: Building images and video. Okay. Yeah. The data volumes are so enormous. We’re lucky to work with a company called land CB that’s in that space. And you can just see, I mean, you know, an image is probably a thousand times to 10,000 times bigger than a text file and a video. 2 or 3 orders of magnitude larger than that.

17:40 – 17:54

Tomasz Tunguz: Well, if you’re already struggling to move text around the way that we are, and we know that custom video is coming, we know that robotics is coming in a very big way. We need much bigger infrastructure to be able to support those moments.

17:54 – 18:19

Sophie Buonassisi: Yeah, great. Okay, well, it’s a fun space. It’s never been a more exciting time to build. And prior to this time, right now, you know, you back some incredible companies like customer and others that have had fantastic outcomes. Are there any kind of patterns that you’ve seen across your, you know, visibly successful companies? And you’ve got a ton of successful companies in the portfolio that are probably less visible, in terms of outcome.

18:19 – 18:25

Sophie Buonassisi: Yet. What are the patterns that you think makes that successful that founders can apply for themselves?

18:25 – 18:47

Tomasz Tunguz: Understanding the history of the space is underappreciated. That’s really important. A deep understanding of the space. The reason second time founders are so successful, especially the ones who start a business in the same domain as their first company. Yeah. It’s they know. They know the people. They understand the history. They’ve just learned about it. And so that level of specialization is incredibly powerful today with AI.

18:47 – 18:53

Tomasz Tunguz: You can understand a lot about your core domain. So that’s one ingredient.

18:53 – 19:09

Sophie Buonassisi: And I can imagine the connections to is a big part now. And the distribution, if you build in the same space because now it’s like in very much there is at least what we’re seeing around the go to market side is people are leaning into ecosystems and partnerships, more of them than ever before. And that level of connectivity is a really unfair advantage.

19:09 – 19:13

Sophie Buonassisi: So I can imagine how that would apply sort of to the second type of value.

19:13 – 19:29

Tomasz Tunguz: Yeah. Now that’s true. And then I think the other change, I mean, one of the big changes have been PR is much more significant of a distribution channel that it has been. The press is willing to write about AI in a way that they may be. And I mean, the mass media, they weren’t willing to write about in software.

19:29 – 19:46

Tomasz Tunguz: And then maybe the last is, much like an accelerated use of channel. So historically, channel is something that you might engage with outside of security, might engage with like 15 to 25, maybe 30 million RR. But today we’re seeking channel partners and men even low single digit are.

19:46 – 19:53

Sophie Buonassisi: Wow. Are there any kind of contrary views that you have or hot takes, if you will, about the AI space right now?

19:53 – 20:13

Tomasz Tunguz: I think the impact is still broadly understated. Why it would be so transformational. And I think we’ll see it in organizational design. Companies today are structured in a way that they were for a time for a computer was invented. It’s kind of wild. It’s you know like the idea of a product manager is maybe 40 years old.

20:13 – 20:15

Sophie Buonassisi: Yeah.

20:15 – 20:36

Tomasz Tunguz: And I think I would completely transform the way companies are structured. You can look at like, if you think about cutting up a company with executive leadership, middle management, and then doers or individual contributors. The ratio is probably like 5%, 75%, 20%. In five years time will look anything like that.

20:36 – 20:50

Sophie Buonassisi: Wow. Well the transformative properties and I mean you’ve written about to the implications of pricing and pricing for AI agents. So it’s just tremendous. That’ll be very exciting to see. Any last kind of messages or advice to founders broadly.

20:50 – 21:00

Tomasz Tunguz: I think the only advice I have is nobody really knows the answer to anything. Yeah. And so when there tremendous period of experimentation. Yeah. The best thing that you can do is just jump in with two feet and figure it out yourself.

21:00 – 21:10

Sophie Buonassisi: That is fantastic advice. Now you’ve got an incredible blog or writing space people can follow along with your writing. What else can people follow along if they want to keep in touch with you?

21:11 – 21:24

Tomasz Tunguz: Yes. We’re on Twitter. We’re on LinkedIn. We just started, a series called, Office Hours, where we host sessions with, executives. And, the great part about that is you can dial in and ask a question, and we’ll leave it into the conversation.

21:24 – 21:30

Sophie Buonassisi: Yeah. I really enjoyed Your Honor. Personally, that was great. Well, thank you for joining us. This been fantastic. Appreciate the time, Tom.

21:30 – 21:31

Tomasz Tunguz: Oh, pleasure is mine. Thank you.