Listen to the audio version here.
Big Cloud’s dominance is a given when it comes to the most foundational and capital-intensive services.
But as I proposed in my previous essay, AWS, Google and Microsoft have not conquered the entirety of the cloud market, and Castles in the Cloud, Greylock’s new interactive data project, will aid us in identifying and understanding the areas where startups can and may thrive.
Of course, the sheer number and range of business divisions operated by cloud castles means it is just not possible to avoid all head-to-head competition. Nor is it realistic to expect to avoid the competition from the countless other startups battling the Big 3. This is why the challenger companies emerging as the forerunners of this new era of cloud innovation are implementing the following strategies.
To start, we’ll examine how some challenger companies have managed to achieve relevance in spite of the cloud castles.
First, Avoid the ‘Castles’
The best way to attack the big clouds is to simply avoid them.
This is a well-worn, first principles strategy of many challenger companies, and most do not compete directly with any of the cloud castles.
Instead, go up the stack or identify different users for whom APIs, tools, and applications can be developed on top of the foundational big cloud offerings, like storage and compute.
This is how cloud API-based businesses like Shopify, Stripe ,and Twilio have been able to thrive. All were able to get a firm hold on e-commerce, payments, and communications, respectively, by targeting developers and use cases to build an atomic unit of almost every application.
Most of the apps we use every day need to handle payments, subscriptions, or other financial transactions, and almost every app needs to email, text, or call their customers. Developers should identify the other atomic units that underlie other apps.
If the Big 3 cloud vendors have already won the heart and souls of the infrastructure developer, then the next battle will be for the hearts and souls of different APIs. If Shopify, Stripe, and Twilio combined are over $300B in market capitalization (private and public) then there should be another generation of great API companies. APIs and higher level developer abstractions are the spiritual successors to the Platform-as-a-Service (PaaS) market. Outside of Heroku, acquired by Salesforce, the popularity of Infrastructure-as-a-Service (IaaS) has outgrown PaaS.
To date, the big cloud companies have instead prioritized releasing sets of API tools to make the production and consumption of APIs easier and safer. A few startups have similarly focused their efforts on API tools, but not as many as we would have expected.
From the data we can see that roughly $400M in venture capital financing have been directed at new API tools, which seems like a lot, but is still underfunded compared to the billions going into AI/ML ($4 billion), analytics ($3 billion), and security ($3 billion).
In this analysis we are not tracking the investment in API businesses like Stripe, because by and large, the Big 3 cloud providers aren’t investing as much effort into these areas (but potentially could be in the near future), and therefore do not present an opportunity to do a side-by-side comparison with challenger startups.
Own the Community
This often overlaps with all of the previously mentioned strategies, and is of paramount importance if you are building an open source project.
The Big 3 and open source projects are co-opting the developer community faster than startups can hope to.
As I mentioned earlier, going up the stack will help you identify the community of product users who inform your go-to-market strategy by narrowing the persona you sell to. Increasingly the persona of the cloud user is not a developer but a data scientist, a revenue ops manager, a business analyst, etc.
Startups are no longer building products to sell to an enterprise CIO, but for the actual practitioner within the company. Whereas in the past the CIO was the gatekeeper and monopoly provider of technology to employees, the cloud has enabled companies to reach users directly with low friction.
To investors and founders, this “product-led growth strategy” has become the de facto go to market selling motion.
The key for startups in this selling motion is to create awareness of their solution and then reduce friction to trial and usage of the product.
One way to analyze market needs is to disaggregate applications into atomic parts like messaging, identity, etc. Another potential analysis is to look at the different potential users of the cloud.
Every enterprise is full of departments, lines of businesses, and functions that are underserved by technology today. Software historically represented the digitization of a business process like order to cash or hire to fire, or a system of record like CRM, or served a distinct user like a knowledge worker toiling over a spreadsheet. Cloud has reduced the cost to build new applications and expand the number of departments, LOBs, and personas it can serve.
For example, startups like Coda, Figma, Airtable, Notion and Asana have thrived in the productivity space during the pandemic-induced shift to working from home. But that’s because they were ahead of the game before remote work was a thing — they had developed tools for knowledge workers coupled with prescriptive templates and workflows..
Figma’s focus first on the designer persona and then expanding to product and other knowledge workers is a strong example of owning the community and owning the practitioner.
What are the other practitioners inside a company that are underserved? A product cloud for PM like Productboard? Engineering productivity tools like Jellyfish for engineering leaders to measure productivity of their remote teams? Revenue Ops? SRE tools like Jeli, Transposit, and Blameless?
The cloud has enabled startups to focus on increasingly more distinctive roles inside a company.
Find the White Space
The demands of an increasingly cloud-based world means there are tons of different startups offering very similar services such as databases, DevOps, and storage — and still thriving thanks to the sheer size of the market. (Note: Success in these areas is dependent on the strength of your moats such as deep IP, ownership of a user community, and more which I will get into later).
Niche or Focused Markets
Big Cloud is focused on these common infrastructure services like storage, but that leaves an opportunity for startups to focus on niche areas like vertical markets or more focused technical or business problems.
This has led to a few successful companies like Samsara and Keep Truckin in IOT for fleet management or a company like Pragma in gaming. Serving the demands of a focused market requires a careful evaluation before plunging into these markets to understand what are the vertical specific channels, buying behaviors, and needs.
As my partner Reid Hoffman has cautioned about such cases, while it is possible that a highly valuable opportunity exists — but hasn’t yet been recognized by serious players — it is also possible that the opportunity just isn’t that valuable.
Security, almost by definition, is an eternal whitespace for new companies as startups continue to innovate to keep pace with hackers and new exploits. The recent spate of ransomware attacks on our energy and food supply chains combined with sovereign cyberhacking like the Solarwinds hack have only highlighted the need for new security tools.
While security startups received over $2B in venture capital financing the past two years, newer adjacent markets like compliance and governance only received $700M in venture financing, a market that we think is still under the radar but will quickly move to the front of our awareness as more data compliance laws like GDPR and CCPA are passed.
API, Serverless, SaaS security: One area that we have been tracking is the intersection of the API market with Management and governance, such as how Salt Security has used API analysis to prevent sensitive data exposure, and overall simplify compliance.
Build Deep IP
Perhaps the most recent example of an individual company succeeding against Big Cloud is Snowflake — which built a $70B market cap company, Snowflake was able to do that not with open source, but by building deep IP combined with operational excellence. Snowflake didn’t try to replicate what AWS’ Redshift was doing. They rethought the entire data warehousing process and separated storage from computing.
Twenty years ago, the three fundamental building blocks to any computer was storage, compute, and networking. Snowflake was able to exploit the fundamental advantage of cloud computing, turning storage, compute, and networking into an elastic resource to build a better database.
Will there be a next generation storage company built today, or have those categories been killed by the cloud?
We have seen a lot of the innovation and IP move to adjacent markets like databases and machine learning that take advantage of the cloud like Snowflake does.
As I mentioned in my Evolution of Cloud essay, many of today’s leading cloud companies were built by people who either experienced the cloud transition (painfully) while working at the “dinosaur” legacy enterprise companies, or who worked at tech companies that rose to prominence specifically because they were born in the cloud like the major consumer app companies like Uber and Facebook.
Their first-hand exposure to issues that arose from using hybrid infrastructure allowed them to identify not only new problems to work on, but also gave them the opportunity to develop proprietary methods to solve novel and existing problems.
Where are the next areas where deep IP is being built by startups?
First, the same database market that created Snowflake will continue to be a market disrupted by deep IP. The predecessors owning this space like Oracle and Teradata gave way to Cloudera which gave way to Snowflake and Databricks. The next evolution of IP in the data market is the move to real time data and streaming. In the streaming and stream processing markets, each of the cloud providers have created solutions like AWS Kinesis and Google Dataproc.
As discussed in the essay on VC funding trends, these markets have created multibillion dollar startups Confluent (Kafka), Fivetran, Fishtown Analytics (dbt). More recently startups like Rockset, Startree, Imply, Rill Data are all building real time databases.
For example, Rockset, which was built for the cloud-only world, looks like a database to the developer (i.e. speaks SQL) but looks like a search engine under the hood, using a serverless indexing that delivers analytics at an unprecedented speed and scale.
These are buzzwords we all hear everyday, and there’s so much investment from both big cloud and venture capital. The Big 3 cloud providers have 74 ML services between them and in the past two years around 35 ML powered startups (at over $500M valuation) have raised more venture capital than any other market. Despite all this investment, the number of multi-billion dollar ML companies is still relatively few: C3.ai, Scale.AI, and Datarobot are three of the larger public and private companies.
The ML tools category is still in the early innings, but AWS, Azure, and GCP are all trying to win the category with AWS SageMaker, Azure ML, and Google AI Platform. This is one market where two different strategies may win. The first is to own the end-to-end experience for the ML user like SageMaker or Datarobot attempts to do.
The second strategy is to go deep in an area and build deep IP. For example, Snorkel is leveraging its open source project and technology around data labeling to differentiate itself versus the market and Truera is using its IP around explainable AI to build tools for model quality and monitoring.
While billions of dollars are going to win the ML tool market, even more money can be said to be chasing the returns of applied AI/ML, as every company today uses some form of ML to build a “System of Intelligence.”
Applying ML deeply in one focused market is potentially the best way to build a deep IP moat.
The recent IPO of Uipath ($35B in market cap) is illustrative of how a company can build a large business in Robotic Platform Automation (RPA). The next generation of applied ML companies will use automation to solve the most complex business problems.
For example, Instabase is using ML to transform how enterprises work with documents and complex business workflows. Instabase leverages ML techniques like OCR and data transformation to create a set of tools that let customers build custom apps to automate previously human intensive workflows.
Examples of applied ML exist in all major SaaS applications, like Salesforce’s Einstein or Gong in sales operations technology. Security is another market where applied ML has been transformative. Abnormal Security employs various ML techniques to prevent business fraud in email and other communication channels.
As we go through the various strategies companies have employed to compete (or at least co-exist) with the Big 3, it’s important to remember that not all tactics that have worked in the past will necessarily work in the future. Furthermore, each strategy may unfold in different ways as each market we have analyzed matures – or in some cases, declines.
Additionally, we expect to discover new strategies and tactics deployed as the cloud ecosystem continues to evolve. We are already beginning to see surprising trends in our data analysis of venture capital funding, which we discuss in an accompanying essay and Greymatter podcast.
As we gather more information through the course of this project, we are eager to learn all the ways entrepreneurs are working to break up the dominance of the Big 3 Cloud Providers.