In the past 10 years, Snowflake (NYS:SNOW) has culminated into a force that is far from melting, in our view. As enterprises continue to migrate workloads to the public cloud, significant obstacles have arisen, compromising performance of data queries, creating hefty data transformation costs, and yielding erroneous data. Snowflake seeks to address these issues with its platform, which gives all of its users access to its data lake, warehouse, and marketplace on various public clouds. We think Snowflake has a massive runway for future growth and should emerge as a data powerhouse in the years ahead.

Traditionally, data has been recorded in and accessed via databases. Yet, the rise of the public cloud has resulted in an increasing need to access data from different databases in one place. A data warehouse can do this but still does not meet all public cloud data needs—particularly, in creating artificial intelligence insights. Data lakes solve this problem by storing raw data that is ingested into AI models to create insights. These insights are housed in a data warehouse to be easily queried. Snowflake offers a data lake and warehouse platform, which cuts out significant costs of ownership for enterprises. Even more valuable, in our view, is that Snowflake’s platform is interoperable on numerous public clouds. This allows Snowflake workloads to be performant for its customers without significant effort to convert data lake and warehouse architectures to work on different public clouds.

We think that the amount of data collected and analytical computations on such data in the cloud will continue to dramatically increase. These trends should increase usage of Snowflake’s platform in the years to come, which will, in turn, strengthen Snowflake’s stickiness and compound the benefits of its network effect. While today Snowflake benefits from being unique in its multicloud platform strategy, it’s possible that new entrants or even public cloud service providers will encroach more on the company’s offerings. Nonetheless, we think that Snowflake is well equipped with a fair head start that will keep it in best-of-breed territory in the long run.

Data lake, but no moat

We think Snowflake benefits from switching costs and a network effect, protected by its unique multicloud strategies throughout its data lake, data warehousing, and data sharing offerings. However, we cannot say with the complete certainty that these moat sources will lead to excess returns on invested capital 10 years from now, given the limited public history of the recently public company as well as its significant lack of profitability today.

Snowflake is a fast-growing provider of data lake, data warehousing, and data sharing solutions. The company’s value proposition lies in overturning the faults of existing data storage architectures and even more recent methods of storing data in the cloud through its combined data lake and data warehouse platform. Traditionally, data has been recorded in and accessed via databases such as the Oracle database or SAP’s HANA. However, the rise of the public cloud has resulted in an increasing need to access data from different databases in one place. A data warehouse serves this need by gathering data from various databases in one place.

The data warehouse alone does not fulfil all public cloud data needs today—particularly, the need to create insights from artificial intelligence and big analytics. As a result, many companies now depend on having a data lake as well as a data warehouse for these needs. Data lakes are used to store not just structured data within a database or data warehouse, but also data that is not yet fully indexed (either unstructured or semi-structured data). This data in the data lake is what is ingested into AI models to create novel insights. These insights are indexed and therefore are considered structured data, which is then housed in a data warehouse to easily be queried.

Data lakes and data warehouses can be created in a number of ways. Enterprises can hire robust technological teams to build out data lakes and data warehouses, or use out-of-the-box platforms, such as Amazon’s data warehouse offering, Redshift, or Microsoft Azure’s SQL Data Warehouse. However, the issue with out-of-the box platforms is that they tend not to be cloud agnostic; Redshift only runs on AWS, while Microsoft’s product only runs on Azure.

a picture of showing lines of code on a computer screen

The amount of data collected and analytical computations on such data in the cloud will continue to dramatically increase

Snowflake solves this problem because its data lake and data warehouse combo can be deployed on various public clouds. We believe Snowflake’s solution provides tremendous value to its customers in two ways. First, it allows for high-performance queries for companies using multiple public cloud vendors. Second, Snowflake offers flexibility in the future if a company wants to change public cloud providers. IT departments are often fearful of vendor lock-in, so this flexibility provides value.

For example, if an enterprise had all of its public cloud workloads on AWS, and in turn chose Amazon’s Redshift for its data warehouse, it would be an extremely arduous task for the enterprise to move its data warehouse away from AWS and Redshift and into a competing cloud vendor. We think this process might take several years. However, if the enterprise’s data lake and warehousing needs were met through Snowflake’s platform, the shift would be far less painful as Snowflake’s platform has built in cloud vendor interoperability such that even a company on multiple public clouds could use the Snowflake platform across all. We still think that switching cloud providers is a painful task that IT departments won’t take lightly, but it still may occur. For example, after making an acquisition, an enterprise might want to consolidate the workloads from the acquired company onto its cloud vendor of choice. Or in other cases, an enterprise may want to spread its bets across multiple cloud vendors out of fear of lock-in or unfavourable pricing. In any of these instances, Snowflake’s platform-agnostic data warehouse and data lake products make it easier for IT departments.

Snowflake’s platform is also distinct in its ability to store diverse data sets all together as one data set, as well as the platform’s methods to optimise the performance of data queries even amid large volume data sets. These characteristics differ from common issues that arise in external data lake and data warehousing architectures. One problem with existing data warehouses is that they are unable to store diverse data types in a consistent format such that the data does not require transformation before using. Data transformation is seen as undesirable as it can lead to errors and duplicates of data. This issue has come under the spotlight as diverse data types have become more and more common, while the desire to use such data sets together is becoming increasingly preferred as enterprises are increasing wanting databases designed for both transactional and analytical work simultaneously. Other issues with existing data architectures include their speed in querying large data sets, which can be problematic as the public cloud lends itself to larger and larger data sets (because storing data becomes much more effortless in the public cloud).

Altogether, we think data architecture software is extremely sticky. More than ever, enterprises rely on data significantly to be the backbone of their business—internally and through their products. Therefore, storing and accessing such data is a critical activity for an enterprise. Changing these storage and access methods of sensitive and highly mission-critical data creates numerous headaches that aren’t dissimilar to switching costs we see in other enterprise software, such as a significant learning curve in using new data architecture and significant monetary and time-related costs associated with “rewiring” how a company stores and accesses data. We’d argue that data architectures might be even stickier than the enterprise software used to input and access this data, since the underlying data is even more fundamental and mission-critical to a business. As one simple example, one can imagine the disaster that would occur if Weather Channel data for Peru were accidentally replacing the data set for weather patterns in Ohio, which might lead to bad decisions and damages for companies reliant on such data, like airlines or scientific researchers.

We think Snowflake’s switching costs are strong even without locking in multiyear subscriptions. Roughly 93 per cent of the company’s revenue is consumption-based, meaning that Snowflake is not a subscription-as-a-service company. Snowflake clients typically commit to set consumption rates on an annual basis. However, if they do not use their agreed-upon rate, Snowflake often allows for agreed-upon consumption to be rolled over to following periods. Although Snowflake does not lock its users into multiyear contracts like many moaty SaaS businesses (many of which we view as having narrow or wide economic moats), we don’t think its revenue model makes its business any less sticky. We think that switching costs permeate throughout the enterprise software industry are most heavily tethered to the costs of implementation of a new software rather than the sheer lock-in of a multiyear contract. The company’s net revenue retention rate illuminates the loyalty of its customer core, as Snowflake boasts dollar-based net revenue retention rate of 162 per cent.

Switching costs aren’t the only moat source we see in Snowflake’s business, however. We think that Snowflake benefits from a network effect in its data sharing business. Snowflake is the only platform that allows for sharing of data sets in a multi-cloud fashion through the Snowflake Data Marketplace--examples include FactSet, the Centers for Disease Control, and AccuWeather—and this data can be purchased by other companies. One example of a data set that might be sharable would be IMDB data for the movie industry. Before Snowflake, if the Academy of Motion Picture Arts and Sciences wanted to use IMDB data, it would have to consider where the data is housed. If the academy had its internal database and workloads on AWS, but IMDB had its data on Microsoft’s Azure, it might be a tedious process, taking several months and many product managers, to funnel this IMDB data over to the Academy. Theoretically, the Academy might choose to leave the IMDB database on Azure, but workloads accessing this database would be much less performant because the data would be housed in different locations.

With Snowflake, such data sharing across different public clouds can happen seamlessly in just a matter of hours or days because the data does not need to be transformed manually. Once purchased, the data can easily be transferred to the purchasing company even if it uses a different public cloud than the provider. We think Snowflake’s network effect is a product of its data marketplace because each marketplace user benefits from each additional Snowflake data set provider on the platform and vice versa.

Public cloud providers have their own marketplace options for data sharing, but within their public clouds only. Amazon has its AWS Marketplace and Microsoft has its Azure open data sets. We think this multi-cloud platform will ultimately give Snowflake an edge in terms of network effects; data set buyers will likely turn to Snowflake because it has the widest multi-cloud selection, in turn encouraging more users and more data set providers to join the Snowflake marketplace over time.

We also note that Snowflake enables negotiable data set pricing. Amazon requires that companies that are selling data set usage on AWS publish the cost of such data sets. This allows the data sellers little flexibility in a marketplace that traditionally depends on variable pricing, dependent on the size and use cases of the data buyer.

In summary, we believe that Snowflake is a key supplier into a host of enterprises today. We ultimately foresee Snowflake achieving massive growth as it is well positioned within a large and growing Big Data market. However, the company is racking up significant operating losses as it remains in its growth phase. Such growth is a necessity in order for the company to generate excess returns on capital, and the company is a bit too early in its lifecycle for us to have absolute certainty that it can earn these excess returns in the long run.

Risks include competition, data breaches

Snowflake runs the risk that other cloud-neutral software will enter its market or that a public cloud company opens up its data warehouse and sharing offering to be interoperable outside of its respective cloud. While we think it’s unlikely that either AWS or Azure will open up their ecosystems to compete with Snowflake, respective owners Amazon and Microsoft have vastly greater resources to compete in this space if they so choose.

Furthermore, Snowflake is at risk of compromising the data on its platform either through data breaches or inability for compliance tools to do their job. For example, Snowflake offers a number of features for ensuring data is compliant with regulations, such as GDPR. However, if any of such compliance tools—like the ability to delete data compliant with the right to be forgotten—were to fail, Snowflake’s brand could suffer significantly and possibly lead to diminished future business.

Finally, Snowflake’s share price has reached an astronomical valuation at times. We believe that extremely lofty growth expectations are baked into the shares, and investors run the risk of potentially seeing the company grow extremely well but still fall short of such expectations.

We believe Snowflake is financially stable, given the early stages of the company, as we are confident it will generate positive free cash flow in the long term. It had cash and cash equivalents of US$434 million at the end of fiscal 2020 with zero debt on its balance sheet. Undergoing its IPO in 2020, Snowflake raised over US$3 billion from the offering. We think the cash generated from its IPO will act as ample buffer for Snowflake to keep its cash and cash equivalents positive without taking on debt over the next 10 years. We forecast that Snowflake will become free cash flow positive in 2025, after which we believe it will continue to reinvest heavily in its business rather than distributing dividends or completing major repurchases of its stock. We do not model acquisitions in our explicit 10-year forecast, as we think Snowflake will focus primarily on in-house research and development. Nonetheless, Snowflake has made several small acquisitions in the past, so we would not be surprised if it did make minor acquisitions over this period.

Prem Icon

Prem Icon See also Morningstar Guide to International Investing