October 2022
From: Brian and Tobias
Subject: Security, edge computing, and what's ahead with AI
Happy Fall! Tobias and I are back with our October edition of our infrastructure newsletter, where we share reflections, trends spotted and opportunities to collaborate.
Lots has been going on over the last month. There is certainly a lot to cover. It’s amazing that when we sent out this email last month, the renewed frenzy around AI was just in its infancy. We’re going to focus on something much cooler to kick things off (although we’ll get to AI later): Kubecon! Tobias and I attended Kubecon in Detroit for the first time and came away with tons of learnings and continued energy around all things infrastructure and developer tools.
Before we dive in, we have to give credit where credit is due. Kubecon can be a place for laughs, not just the latest and greatest in the world of Kubernetes.
Every business is a security business:
More than ever, it seems like every business needs to offer a security value proposition. The old expression in startups to make something “better, faster, cheaper” should probably be amended to “better, faster, cheaper, safer” for 21st century infrastructure companies. In a world where data is more sensitive than ever and methods for maliciously accessing that data are increasingly sophisticated, being able to provide a secure solution is not just a nice to have. This point hit home in a conversation we had with a founder who is building a Kubernetes security business. She remarked that for the first time ever, Kubernetes security was cool. Companies that previously didn’t touch it were now touting it as a core part of their platform.
This trend is not coming out of nowhere. We now produce a staggering 2.5 quintillion bytes of data every day. That is 2.5 followed by 18 zeros. Our data is everywhere and uncontrollable, meaning that businesses are at greater risk for cyber attacks than ever before. In 2021, cyberattacks and data breaches increased by 15%. 41% of executives don’t trust that their security postures have kept up with digital transformation. Corporations are expected to spend $172B on cybersecurity solutions in 2022. The numbers are really just staggering.
Security has become a core business imperative across all industries, and there’s no sign of that stopping. CISOs are also generally inclined to spend on tooling to solve their problems, making it a great sector for startup activity. A cybersecurity investor recently framed the difference to us in the following way: “Devops budget is 10% tools, 90% people. Security budget is 90% tools, 10% people.” Devops businesses are realizing this. They are increasingly targeting security use cases and teams because this is where the budget and urgency is, and there is real logic to your software infrastructure solutions being security solutions too. We expect the lines here to only get blurrier.
Moving compute to the edge comes with challenges:
Edge computing is coming in a big way. There were many sessions that discussed innovations around edge and how more data will need to be stored and processed on the edge in the years to come. More specifically, edge will be uniquely powerful and important for large, distributed businesses. At Kubecon, We talked to someone at Starbucks overseeing the edge servers deploying apps across North American retail locations. That’s a simple example, but edge will come to touch use cases in a wide range of industries, including healthcare and the industrial economy. Monitoring a patient in their home or a conveyor belt in a warehouse should happen onsite, not in a server hundreds of miles away.
Unsurprisingly, this shift has hit speed bumps and will continue to be a challenging infrastructure buildout for large enterprises. In particular, a key question at the conference was how to effectively deploy Kubernetes at the edge. Containerization is a fantastic method for shipping applications locally, but Kubernetes comes with lots of other powerful abstraction and optimization functionality (the stuff that really makes it so magical) that isn’t really relevant on the edge. Typically, Kubernetes enables engineers to abstract away the machines and focus just on building and shipping apps. On the edge, this paradigm is totally reversed. Instead of abstracting away machines, the goal is to run applications on one specific machine. Is a different container infrastructure required for those applications? We are in the early innings here, and the crowd was unsurprisingly very bullish on Kubernetes still being the right solution. As edge becomes more pervasive, we’ll be tracking the problems and opportunities emerging in this area. We’re looking forward to continuing to learn more.
Platform engineering – here to stay?
There were a lot of startups at Kubecon marketing themselves with platform engineering, a natural evolution from DevOps. Platform engineering works in service of the developer team, building a holistic platform that allows developers to have a frictionless and self-serve experience to help them do their jobs more effectively. In the same vein, lots of companies at Kubecon were marketing themselves as the “developer control plane” – tools that allow developers to control and configure the development cycle more efficiently.
Both platform engineering and the developer control plane point to a broader trend around abstraction that we discussed in last month’s post. They are both newly branded ways of expressing a simple truth that has existed for a long time and was true of DevOps and SRE before: developers need tools to do their jobs more effectively and a “platform” from which discrete tasks can be consolidated and simplified. Whether platform engineering wins the day as the name for this important function within the enterprise remains to be seen. That doesn’t really matter. But it is interesting to note how ubiquitous this terminology has become.
In addition to the learnings from the conference, here are some more insights and announcements:
BI, the low hanging fruit of AI: We have seen close to ten businesses in the last month seeking to transform natural language into SQL queries. This seems like extremely low hanging fruit from all of the innovations happening in generative AI. This is interesting conceptually, but we’ve just seen so many of these same businesses, with relatively little differentiation. We believe a better BI tool needs to exist, and we also believe the generative AI wave will fundamentally change SaaS applications as we know them. AI may disrupt BI. If it doesn’t, it won’t be because people didn’t try.
Cost optimization, and a race to the bottom in cloud: Before the wave of smart SQL generators, there was a surge of infrastructure cost optimization businesses that raised VC money. We spoke to a bunch. Our general point of view is that the businesses optimizing some part of the data stack (e.g. data warehouse, observability) are more interesting than the ones optimizing cloud spend. For the most part, the cloud providers compete based on price. There are some small differences, but server space is for the most part a commodity. The cloud vendors win by offering an accessible, easy, and inexpensive solution. Because cost is a key competitive driver, we think the cloud vendors will bring optimization in-house. They won’t be obvious about it, and they’ll help with optimization slowly and over time, but they can’t afford to fall behind significantly on price. On the other hand, pieces of infrastructure like Snowflake and Datadog do differentiate from other data warehouses and observability solutions, respectively. They choose to compete on lots of other things other than price, so they’re less likely to optimize cost themselves. The warm relationship between Snowflake and Bluesky is evidence of this point. So overall, we’re skeptical of cloud optimization startups, but still interested in other infra optimization.
Our data fellowship: Every quarter, we run a fellowship for aspiring founders in NYC looking to start companies. This quarter, we’re focusing on data and are working with ~35 founders on the precipice of starting something new in the data world. If you’re interested in participating in some way or learning more, just let us know.
We now have over 100 people in this community receiving this email. If you have a suggestion for someone who you think would like it, please introduce us over email.
Until next time,
Brian and Tobias