Gokul Rajaram on the SaaSpocalypse: why lightweight SaaS is dead, DoorDash's moat, and 30-40% corporate layoffs coming

tasks. Synthesizing data into a single view or bringing creative projects to life. And without further ado, we have Gok Roger Ram from Marathon Management in the studio. How are you doing? What's going on? Good to see you. Great to see you, John. Welcome to the show. Uh, is software dead? First, introduce yourself.

Sorry, we'll get there. Hey guys, I'm huge fan, longtime fan and listener to to you guys to VPN. Uh, I'm Gokul. I've been an operator and investor uh for 25 years in the valley and uh yeah, excited to be excited to be here with all of you. So, yeah, it's about time. Long overdue.

Um yeah, walk us through how you have been processing the the the the SAS apocalypse, the Catrini article, uh all all the different narratives that have kind of shaken the tech industry both to the upside in many cases and also to the downside in others uh at this throughout the start of this year.

Look, I think uh software is going through a change. I don't think it's dead. Software is still going to be around. In fact, every industry is going to use software at its core.

I think what is going to be challenged is companies that are surface level where they're going to be light which are there's two kinds one of them are lightweight seat based easy and replace example I always take I don't want to pick on them is zenesk zenesk is customer support software they sell on the basis of seats and use it for delivering outcomes for customers the problem with zenesk is that if I'm a customer of Zenesk.

I'm buying 50 seats. I could reduce it to 20 seats and use 30 seats for AI. In the same budget of 30 seats, I can use AI. So, I can literally compare Zenesk's performance to AI agents in parallel and Zenesk won't even know this is happening.

So, you might be running an AB test and you can literally run Zenesk in parallel. And that's a challenge for company like Zenesk because it's not deeply rooted part-time for only part of your employee, part of your customer service request.

The other kind of software is stuff that is not outcome based that is truly running your company that's truly running a process. A good example is Netswuite which is an ERP which is literally at the core of your general ledger. It's running your accounting a lot of your business is running off it.

You can't really run another version of Netswuite in parallel to Netswuite. You've got to run and so Netswuite is in a much better position than Zen does.

Now at the highest level the reality is every piece of software has to reinvent itself over the next but the time frame it's much more urgent for Zendes to do that in the next 6 to 12 months while something like Netswuite probably has years 3 to 5 years to think about it.

So if I were Netswuite what I would do I would create AI agents which I'm sure they're already doing and I would as they say commoditize my compliment.

I basically make it free uh to actually have a bunch of interesting workflows which are off offered through AI and I would basically still charge for the data very very different and u how did you how did you there's there's a whole new class of AI native ERPs that uh have come online and raised over the last one to two years uh is is h how did you process them there's certainly like a long graveyard of of of startups as well.

But uh part of the challenge is there's just so much software to build with a system like that. Uh and and maybe now is the right time given that we can create software a lot faster than we could historically. Right? I think the I' I know many of those companies. I think they are great entrepreneurs.

I think the challenge for them is that there is no pressing burning reason for somebody using a Netflix or net suite I should say to rip and replace Netswuite.

So what you have to do is you have to go with a company that doesn't have an ERP today and get them to start using you as a soft as a software company as a vendor and then grow with them. And so I think it is so that's a strategy build your customer base from net new companies which don't have an ERP today.

I think going to somebody who's using a a large ERP and saying well throw out your ERP and replace it with me is hard.

So you've got to either come up with a very very very uh narrow wedge where you can inject yourself in and then use it to over time take on net street but more interestingly I think go after younger fast growing startups which will need a netsweet or an ERP system in a few months offer it to them for free.

Some of them actually some of these ERPs have been saying if you're funded by one of these VCs you can use the first tier of my product for free. Sure. And that's a smart idea because you use it for free.

you're not going to generate much revenue but in exchange they become sticky and when you actually start generating revenue a year from now they're there for you. Yeah. Um what what do you think is happening in the private markets for software companies where their public comps have declined by 50%.

There we we sort of went through this and I'd love to know how you process the uh the end of ZERP. I remember reading Logan Bartlett on this at Redpoint a lot.

He was pulling public comps and seeing how multiple compression had happened in the public software SAS market and the effect that that was happening in the private markets. It feels like the haven't fully digested the SAS apocalypse uh in the same way because a lot of the startups are earlier on the AI narrative.

But at the same time there's like a bigger question there. That's right. The way the private markets work is what can the private markets do? They can't rerate the company on a daily basis like in the public markets. So what they can do is what did I I that term sheet I signed is is due for a wire at what valuation?

I'd like a 50% discount please cuz the cop is down 50%. Yeah. You can't exactly do that.

you know that happened in 2008 where you did see when the 2008 lemon thing happened you did see people pull term sheets right after that happened that were outstanding but those kind of thing that happened on the edge but the biggest thing that's happening is there are certain classes of business models that have maybe raised a seed or an A that are now unable to raise a B or will be unable to raise financing and those are business models that are primarily application layer I think if you're just doing workflows it better be a deep and complex workflow that is deeply deeply deeply embedded Because now with open emerging you guys might have seen this open actually does the work it's not just workflows it actually goes and does the job.

So we're just putting software workflows that's enabling other people but not doing the job yourself.

it becomes harder to justify that you will exist because remember the VC time frame is not not a year it is 5 to 7 years and if you think about compounding over 5 to 7 years that's insane why because the capabilities of these models are doubling what once every 6 months say and if it's once every 6 months it's quadrupling every year that means in 5 to 7 years it's 2 to the^ of 2 to the^ of 12 which is a thousandfold better than what it is today and my human brain is unable comprehend what it means for these models to be 500 to a,000 times better than they are today.

what it means for software.

So I think everybody's trying to figure out over a venture time frame what does this mean if you extrapolate the performance of these models and make them a thousand times better what survives and what doesn't what survives things that are deeply rooted that have atoms we think as software investors that things that have atoms yes there are humanoid robots but I personally feel general human robots that are machinists or hwack repair people whatever the case is that's going to be slightly further out so things that have atoms are generally protected over the 5 to 7 year time frame things that have regulatory modes are protected things that have money flowing through them are protected.

So there's a few classes of things that are protected. Yeah. Uh if somebody came to you and wanted to vibe code Door Dash, what what advice would you give them about the challenges that they might face on uh building a competitor? You know, I'll give you I'll give you a story. Uh I'll give you two stories actually.

Assume it's June 2028, right? Which is what I think Catrini had for 30% drop in stock. Assume you're hungry. You open your phone. You ask Claude, say, or or whatever your new startup is, the the food ordering startup to order your sandwich with gluten-free bread.

It routes you order directly to the restaurant because it's figured out how to connect to all the restaurants and it saves you $2, but it takes an hour and a half to show up. It was cold and they forgot to make the bread gluten-free, so you can't eat it anymore. Who owes you a refund?

Does Anthropic have a customer service number? No. So basically, you open Door Dash, the sandwich is the first recommendation, you tap on it, 90 minutes later, you're holding your sandwich, still fresh, and guess what? Remember the gluten-free bread.

So that's one story where if something goes wrong in the physical world, who does who who takes care of it? Door Dash is a managed marketplace. I'll get back to that. The better, more pertinent story is run from 2016 or 2017. Google launched a product called Google Food Ordering.

You might have seen this Google food ordering. can actually order from a restaurant and it literally it was driving at that point multiples of traffic to any restaurant compared to what Door Dash was driving because Google still massive back then was massive.

You could search for a restaurant, you could order from a directive from Google. However, the retention of people who ordered Google food ordering was basically zero. It was it was negligible. Why?

Because after you place the order, there are tons of things that go wrong in the physical world and you couldn't really call Google and the restaurant didn't take responsibility there. So that product doesn't exist today. Huh. And and I think that's the fundamental reason.

There are things that happen in the digital world that are not enough to replicate something like Door Dash. Doash manages not just aggregation and supply of restaurants. It manages everything afterwards. It manages figuring out how to get you out of the right right dasher.

It manages how to basically refund if there's something that goes wrong. It manages customer support in case you want to. There's a collection of systems that work together. Discovery is something that I think agentics will do well and hopefully do dash will incorporate many of those. But discovery is not enough.

If you think about Google and digital, what did Google disinate? What do we use Google for today? It did really well with Expedia and flights and hotels and also things like mortgages.

So the types of marketplaces that Google has been able to disrupt our search are things that are more lead gen or just I would call light marketplaces where payment happens to Google. You can book hotels or flights to Google. You don't have to take care of what happens afterwards.

But a managed marketplace in the physical world is the most durable thing. It is not a piece of software. Right? This is the thing I think people get wrong about Doash. Doash the soft Doash the first piece of software they built after the website. If you remember Doash had a stall to delivery.

It had this website with a PDF. But the first piece of software they built was not a consumer app. It was a dasher app. It was the app for dashers because that was what was needed to get the food properly to make sure the dasher got the order, accepted the order, picked up the food, delivered it before the consumer.

Can you talk a little bit about what it takes to build liquidity on a marketplace? It feels like that's also underrated if you're if you need to spend billions of dollars to actually create the scale of a marketplace that that's something that can't necessarily be vibecoded. Exactly.

So it's the the biggest mode here the the one of the modes is density. Density leads to liquidity which means the more orders you have in a given geography at a given time the better you can bash deliveries which increases driver utilization and reduces cost per order. And that's what you have to do.

If you were to actually play the act pay the actual cost of delivering it would be 101 12 $15. The reason you're only paying $2 or $3 is because these orders are batched and you're basically not having a single person. You're basically batching it to the best Dasher driver.

That's why you might see in in busy busy times that there are drivers waiting outside certain restaurants because they literally know that that order they can get pick it up and there is massive density of orders. So density and liquidity, that's what drives marketbased economics.

And so whoever wants to replicate Door Dash doesn't just need to improve aggregation at the consumer level, but build a massively dense geography by geography marketplace in every city that and every suburb that Door Dash operates in. Uh and that's a massive endeavor.

That's what Door Dash spent a few billion dollars uh building. Yeah. Just a few. Uh what does it take to be a great public company board member? It's interesting. I think you know it has changed over the last several years. I think a few years ago, decade ago, say the profile of a board member was quite uniform.

It was people from the finance or corporate worlds and that's still important. I think you need a finance person who runs your audit committee. Uh you need a CEO. I think every public board member, if you would go to any public CEO and say what is the archetype you want in your board, most of them want CEOs.

They want to see you on the board because you want a CEO who you can confide in and you can ask for help from. But there are new archetypes that have emerged over the last few years. One product engineering leader, someone who's led product and engineering at a company. Why?

Because every company, like I said, has become a software company, a technology company. So every public company wants someone with product engineering chops. Second, you want someone who represents your customer. At Door Dash, for example, sorry, at Square, we had the CEO of Shake Shack. And why?

because Shake Shack was a prototypical customer for Square. Uh no, Danny's it was Randy Garuti. Danny was a chairman. So Randy was Randy was a CEO. I mean yeah Danny was a founder and Randy was a CEO and so he was excellent.

He always he would really assume and then typically what happens is especially if it's a company that came through venture capital typically one or two of the VCs that have been early in the journey of the company stay with the company and the CEO trust them. In square's case it was Rolof Ba at Sequoa.

In Door Dash's case, it's Alfred Lyn, ironically again from Sequoa, but at Pinterest's case, it was Jeff Jordan from Andre. So each company has one VC probably who who's and then in in uh Coinbase's case, it's Fred Wilson and Mark Andre from from Union Square and Andre.

So VC, product engineering, chops, um voice of the customer, CEO, finance person. Those are the five archetypes. One of the best practices today that I've seen companies using is a notion of a board buddy.

What it means is every exec at a company should actually be paired up with a board member who's an expert in the same area. So for example, I typically will work with the head of product and the head of engineering at the companies I'm on the board of and I meet with them once a month outside of the board meeting.

Those meetings are actually really good. I am essentially listening to them in a judgment free zone and trying to be helpful and getting feedback from them. How can we be helpful? And the same is happening with the finance person, the finance oriented board member.

they meet with the CFO, the customer board member, they meet with the CRO, etc. , etc. So, I think uh that board buddy system I've seen helps dramatically because that's what your board member should be helpful not just the CEO but to the exec team.

It also helps gets exposure for us to the exec team who they are and as we think about succession planning and stuff like that, we can evaluate who these people are and who should succeed the CEO who are the people. Yeah. Uh, what are your forecasts around M&A in in 2026?

Last year was the year of the of the licensing deal, the zombie acquisition was a ghost ship. Yeah, the the ghost ship acquisition. Uh, but it still feels like we we could see a lot of activity uh this year. There's a lot of cheap companies out there. GoPros at 150.

There's a lot of Yeah, in the public markets, but also but also public companies that want to bolt on fast growing teams. Yeah, I think there's going to be both.

I mean you saw for example my fitness pal bought this company Cal AI and that thing is 18 months sold right 50 million and so I think there's going to be a tremendous amount why because like you said there's a few things happening one software companies are valued and the valuations has gone down and so I myself in a portfolio I've seen if the CEO has courage they can actually your the stock market is already telling you that the terminal value of your business is zero so there's really no downside to burning the hordes and going all in on AI and I have a company Podium uh I mean you should actually get the guy he's amazing he similar to intercom very similar to intercom he had a business that grew to 200 million was kind of growing very slowly and he said you know what I'm going to go build an AI agent I'm going to is basically a product for SMBs and he's basically gone from 0 to 100 million in 24 months and is accelerating by just saying you know what I'm not going to worry about the legacy business we're going to start a new business so I think but the assets are there these companies have assets.

Yeah, they need they need that. And so it might be PE buying these companies for cheaper, but the PE can no longer have the playbook of we'll operationalize. No, it's not about operationalizing or improving costs. It's about building a new product, a net new product, an AI, AI native product.

Then like you said, I think there's a tremendous number of for example inference clouds are a new thing, fireworks, modal, base 10, there's multiple 5 to10 billion companies.

I have to assume the hyperscalers are not sitting around saying inference in AI is happening on these completely different cloud platforms different than what I have. What the hell am I? I'm going to be a commodity cloud platform.

So almost certainly I think you're going to see one of the hyperscalers buy one of these neo clouds. I think core was one kind of cloud but there's together there's like six or seven different next generation cloud companies.

So cloud software you saw chips with Nvidia buying grog there's a bunch of nextgen chip I think you're going to see and of course the whole Warner Brothers thing at the application level the consumer side there's a tremendous amount now that I think we are in an administration which is much more uh open to this for lack of a better word you're going to see I I think there's going to be amazing activity next year uh do you think that the square you were at square for almost six and a half years do you think the square riff from last week will be an anomaly or are you expecting uh 40% to be the new 10% uh riff?

The latter. I'm almost certain there's going to be multiple 20 to 30% cuts uh maybe even 40. The reality is nobody's hiring. I think almost every public company has basically net new hiring is almost zero. Why? Because you can do much more. Every function can do more with AI.

So it's equaling to a 30% boost in productivity. And that's the first year. Next year, if you get another 30% boost, you need 30 people. So over the 18 months, almost certainly every public company is going to have a 30%. If they don't, I question I question the leadership.

So yeah, so a lot of the so so uh a lot a lot of people's reactions to the the Square riff was, okay, AI is just a convenient excuse. It it it kind of is a narrative that the market will be able to process. Do you think there's more?

You you think they're actually getting seeing the efficiency uh to be able to make a decision like that and it's not just marketing? 100%. I think they realize that they could basically do the same job and probably more efficiently with fewer coordination.

The fewer people you have easier it is to coordinate as you know. I mean the two of you run the shows assume you had four hosts how hard it would be like some of the ESPN shows right? It's just the two of you. Every time you add a person it's m* n.

There's like like it's it's it increases exponentially the coordination problem. Yeah. So, I think a 12,000 person company, 6,000 company, night and day, I bet they're going to move faster, be more innovative, and drive more productivity than they were doing before. What's the update on Marathon?

Uh like where where are you guys focused uh in Q1? Uh what's most exciting for you? We are an early stage software and fintech uh focused firm. Uh we focus some of the best companies to do are at the intersection of the two. But in software, we look at infrastructure, security, and application.

Uh I think the themes we're looking at are some of the themes that we talked about. One is what are durable companies at the application layer. Um there are lots of interesting and you know things on the security side we're looking at. For example, fishing is something which is going I mean I have been fished now.

I mean I feel almost on a daily basis I'll get an email that looks like it's coming from a uh coming from a human being but it's actually AI. Uh and then uh the third area is around infrastructure.

Like I said, there's a lot of new infrastructure that has been built because every piece of infrastructure that existed till 2 years ago was built for humans. Mhm. Every piece of infrastructure has to be rebuilt for AI agents. On fintech, um, you know, stable coins is just I think I I'm I'm in love with stable coins.

I think USDC uh I have so many people I know outside the US now who basically want all their money to be not in their local currency but in USDC, a stable coin because it's peck to the US dollar.

It gives the opportunity for everyone in the world to not suffer a 10% daily inflationary basically reduction in their purchasing power but to hold money in in in a uh in a in a way that prevents basically respect to the US dollar.

So I think stable coin is one of the greatest you know so far I think it's the best real outside of tokenization it's the best real world application of of of the crypto industry in some ways and I'm extremely bullish on it. Amazing. Well, thank you so much for taking the time to come chat with us.

Yeah, great to finally have you on the show. It should have should have happened a long time ago. Yeah, always huge fan. Yep. Always welcome. It's been amazing. Has it been two years since you started? It's been incredible. One year. Isn't that crazy? Yeah. That is incredible. Congratulations, guys. Thank you so much.

We'll talk to you soon. Cheers. Have a good rest of your day. Let me tell you about Sentry. Sentry shows developers what's broken and helps them fix it fast. That's why 150,000 organizations use it to keep their apps working.

Board member Coinbase, board member Pinterest, board member the Trade Desk, spent four and a half years at Door Dash, 6 and 1/2 years