Why Infrastructure is Forever

When I returned from vacation this month, inevitably people asked, “How was your time off?” The conversations would go something like this: 

Me: “I took two weeks and it was glorious.” 

Other person: “Wow, great. Where were you?” 

Me: “Wisconsin!” 

Other person: “....Oh.” 🤔

I think they were expecting Paris or Saint Tropez. 

I get similar responses when asked these days what sectors I am excited about: 

Me: “Infrastructure!” 

Other person: “....Oh.” 🤔

They were expecting Artificial Intelligence. 

I started my career in infrastructure software when I joined Oracle in 2009. I have been investing in infrastructure companies ever since: technical founders building technical products for technical audiences. I have come to love the founder persona -- the types of people who are bold, forward-looking, drawn to the foundations, and the often hard, distributed problems. 

We all have things that command our loyalties long past the time when they are en vogue. One of mine is a family cabin in Wisconsin that remains unchanged from my childhood years and even from my grandmother’s. Like the pyramid of waterskiers, it’s a sturdy icon from a previous century. (Though I took that photo last month, look closely and it could be from 75 years ago.) 

Infrastructure has a similar timeless durability. I’ve evaluated and re-evaluated investing in infrastructure software, especially as other themes and trends emerged. I have always come back to the staying power of infrastructure. It is timeless and it's because it has enduring structural advantages: a core strategic position, horizontal potential, and -- it’s worth saying again -- durability.  

Let’s examine each of these advantages in detail. 

Strategic Position 

Platforms can pull the rug out from software built on top and not the other way around. We all think in terms of layers of the stack. Networking, storage, database, middleware all supporting the application above. Like the waterskiing pyramid, it all starts with an ultra-reliable foundation.

A recent example: HashiCorp switched to a Business Source License (BSL) from what was previously the Mozilla Public License (MPL). HashiCorp now has more control around commercialization. The opposite is also true: companies built on top (e.g., Spacelift, Pulumi), have considerably less control. Saying nothing about whether it was the right move or not it wreaked havoc for those built on top.

I look at founders today, like Chris Lattner at Modular, like Erik Bernhardsson at Modal. Each wants to take on a big, audacious problem (a variation on “fix AI infrastructure”) and where did they both land? At the foundational layer, by requirement. They are satisfied to build abstractions above only after they resolve and rewrite the foundation. 

Horizontal Potential

The term “platform” is overused, and the real test is breadth. How much surface area does a suite of products cover? How broadly and universally? Can a founding team solve additional needs over time for their audience? 

A determined founder solves one problem, generally one that he or she has experienced first-hand. The award for solving it is not a celebration, it is the compulsion to resolve the next problem. Infrastructure and cybersecurity are structurally advantaged here (versus applications) because both are designed and built to be more universal, application-unaware, unrelated to vertical or use case: more horizontal. Case in point the Datadog tagline: “See inside any stack, any app, at any scale, anywhere.”   

See the pyramid, where it takes a full eight skiers on the bottom, eight pillars, to support the flashy gymnasts at the top. Within infrastructure we have exquisite examples of starting from a single problem-area and expanding to be multi-pillar: Crowdstrike, Datadog, Palo Alto Networks, and Cloudflare. 

Datadog had an ACV of just $37K at IPO in 2019. Fast forward to Q2 this year and 82% of Datadog customers use 2+ modules, with more than 300 customers paying >$1M annually. 62% of Crowdstrike customers use 5+ modules.Cloudflare reports a remarkable 48% of customers use 8+ modules. 

Datadog, Crowdstrike, Palo Alto Networks, and Cloudflare are representative of successful infrastructure and security companies who each figured out how to solve a core pain point and then expand into additional pillars (and get paid for it.)  

Durability

Egyptian pyramids date back to 2630 BC – read indestructible. Even a waterskiing pyramid of 15 sturdy humans traversing calm water has enough tension to maintain its shape. The ropes taut, the boat carrying the skiers along. 

Infrastructure relies on that same dynamic. Constant progress, no quick movements. Less exciting than the daily updates in AI today, no doubt, but longer term very compelling, as proven by the data. 

At IVP, we routinely consider the top exits of the prior decade to keep us rooted to reality. And we endeavor to invest in those, the 10 to 15 companies of consequence each year, the companies that will make the leaderboards when they exit. Looking back we are lucky to count AppDynamics, Crowdstrike, Datadog, GitHub, and HashiCorp within infrastructure.  

What stands out in the data is that across a broad software dataset (Infrastructure, SaaS, Digital Health, Fintech, Consumer, and Climate), infrastructure software shines. Of the ten largest exits in the last decade [1], four were infrastructure companies: Snowflake, Palantir, Crowdstrike and Datadog. 

And these companies have really performed post-exit. This cohort of 17 infrastructure companies appreciated 48% from public offering to today. Datadog, a $30B company today, went public at $7.8B. Crowdstrike, a $35B company today, went public at $6.8B. 

Infrastructure rewards those with the patience to hold. 

 That waterskiing photo makes you wonder how the heck those people managed to get up there. Did they start in a pyramid? Did they climb up while skiing? Did they just magically appear? Some days I have similar questions about the revered infrastructure companies – how the heck did they assemble it all? How did they pull it off?  

It is my genuine awe of that process that keeps me going.  

[1] Dataset includes all IPO/M&A exits, valued at >$5B at exit, within B2B and B2C software, across the United States, Europe, and Israel, sorted by current value (as of June 30, 2023)