What to watch out for in infrastructure investing (Part 1)
Infrastructure investing has gained quite the momentum these days, not only from the number of mentions from VCs but also from the significant number of exits and notable raises they’ve had this past year (e.g., Nutanix’s IPO). This increased activity and attention, in my opinion, should be channeled towards the highest-potential areas in infrastructure. But before we go into detail, here’s a high-level overview of where I see the opportunities in infrastructure (presented in a full-stack diagram below):
Out of the five layers/opportunities I’ve listed above, I’d like to highlight two areas (serverless computing and intelligent networks) that I find interesting and may surprise us with a big raise/exit this year:
1) Serverless computing
Containers are cool (kudos to Docker), but an engineer would still need to deploy and manage them. Deployments mean devops and devops is complicated. With the rise of PaaS such as Heroku and AppEngine, companies recognized the need to make devops easier. While containers simplify that process further, the best devops for a large majority of companies is no devops at all — at least on the part of the engineer (per my friend Andy, it may eventually become an engineering specialist role). Cue serverless computing — not only will you not have to pay for devops engineers, but you can also only incur the cost of cloud compute cycles when you need them.
Thus with serverless computing, devops become a thing of the past, and code can now be executed anywhere. Developers can write code and the Infrastructure-as-a-Service (IaaS) platform will execute it for them. We already see the big players (Amazon, Google and Microsoft) dominating this market, each vying for more market share with their robust E2E PaaS. It would require a lot of resources (it’s time and capital-intensive) for startups to try and compete with these giants on a technical level.
Therefore, the opportunity for startups lies in:
- Being independent: a successful infrastructure startup will be framework agnostic, aka it will be able to deploy onto multiple (ideally, any) FaaS provider (e.g., Lambda, Cloud Functions, Azure Functions) to avoid getting locked in to any particular provider.
- Abstraction: the value will be “moving up the stack”. Startups will need to extricate themselves away from the underlying configuration idiosyncrasies of various FaaS providers. For example, while it is possible to roll your own serverless platform on top of Kubernetes, companies such as Platform9 (fission) and Redhat (funktion) are betting that most developers would rather avoid such a technically complicated undertaking.
- Monitoring: while some startups have started to include this in functionality, I’m still waiting for one that would meet my top ‘serverless monitoring’ requirements. Because topologies are inherently complex, intelligent monitoring of serverless computing will be increasingly important. In addition, engineers want solutions that are ultimately fault-tolerant (which on a tangent, is a reason a recent stage-agnostic investing partner mentioned to me, why Atlassian has performed so well).
A good example of a startup: Serverless (fka Serverless Framework 1.0). Other players include: IFTTT, Zapier, Iron.io, AWS Lambda, Google Cloud Functions, Microsoft Azure Functions.
2) Intelligent Networks — this is an area that could benefit from the strides being made both in infrastructure as well as AI/Machine Learning. I agree with investors who view AI not as its own vertical but as a necessary layer on top of any industry, with the capacity of disrupting any and all conceptions we’ve had about how operations and processes and handled today. And this also holds true for the world of networking. A conversation I had with my friend Rishi at AT&T Foundry confirmed this: networking monitoring and troubleshooting have historically been tasks that have been very costly (both in terms of time and cost) to telecoms and enterprises alike.
As a result of the high volume traffic, coupled with disjointed data servers and split geographic network locations, the E2E network monitoring process becomes quite challenging. As a former PM with Google Fiber’s network infrastructure, I can say first-hand that this was consistently a P-zero for the team and management — not only because it affects P&L but also performance of the network and service delivery to the end user.
Thus, the opportunities for startups lies in:
- Introduce AI/ML capabilities to networking monitoring: why not combine the advanced from AI/ML to make the network monitoring process that much more iterative and efficient? ML can allow the network to accurately predict where errors in the network might occur, to what degree — before they even happen, as well as intelligently suggest methods to troubleshoot with resounding success.
- Bolster network endpoint monitoring/security: these days, networks are becoming moe dispersed and more complex. Some enterprises utilize WAN, while others might use a combination of WAN, VPN and MPLS. Regardless of the combination, network endpoints are often the most vulnerable to attack.
- Mobile/IoT networks: more enterprises are instituting BYOD policies, where employees can access the intranet using their own personal devices, which means that they need to be secured and protected in the event of a security breach. Likewise, we are entering the world of connected-everything, and personally I haven’t seen an IoT security/monitoring app that has really impressed me yet. There have been good attempts, but I feel that they’re lacking the ability to truly bring everything online and onto one platform for comprehensive and secure monitoring.
Good examples of startups: ThousandEyes, SentinelOne.
I’d like to acknowledge all the awesome individuals (VCs, chief scientists, and PMs) who were so generous with their time in sharing their expertise/thoughts with me. Thanks for tuning in folks — in my subsequent posts, I’ll cover more opportunities in other layers of the stack, including public/private cloud migration and emerging endpoint security technologies. Until next time!
John Fan contributed to this post.