Exponential growth in data creates new demands for which traditional data management techniques like relational databases are not satisfactory. The pervasive nature of the internet and the so-called IoT has ensured that the exponential growth of data remains unchecked.
The point of processing this type of data is to be able to react - make decisions, automate tasks, refine actions and content. This is very different than the traditional methods of store, warehouse, process transactions and then update the "system of record." Bank accounts still work that way, but real-time streams of location data from a fleet of a million vehicles do not.
Companies like Google and Amazon have had to deal with this problem early and built their own data infrastructure to do so. Much of it incorporates and gives back to the community in the form of open source software. The first big building block put into open source was Hadoop, which became the standard and is very widely used. A host of additional Hadoop-based open source projects have built on and extended the capabilities of Hadoop into a real data platform.
The problem is that building, maintaining and using this software is hard, very hard. Because it's new there is also a shortage of skilled engineers and "data scientists" that can make it work and get real value out of the tools. The situation is similar to what we had with the Linux OS 20 years ago. Back then, Red Hat (NYSE: RHT) seized the opportunity to take Linux and make it business friendly and enterprise ready. Today Red Hat generates $2.5B in revenues and has a respectable market cap of $16B.
Those of you who have followed the software IPO market closely might be saying "wait a minute, wasn't this the positioning of Hortonworks (NASDAQ: HDP)?" which came public in late 2014? With Goldman as the lead bank, no less? You'd be right - this was precisely the positioning of HDP. From the HDP IPO Roadshow Slides - "Our mission is to establish Hadoop as the foundational technology of the modern enterprise."
Hortonworks didn't really succeed and their current share price is about 1/2 of what it was post-IPO. There are two main reasons - 1) they focused more on open source technology and architecture than enterprise needs and 2) Cloudera.
Cloudera was smart and lucky at the same time. We'll get to the smart part next, but the lucky part was a $740M investment from Intel in 2014 which gave Cloudera unmatchable resources to make the heavy investments required to satisfy large enterprise customers. Their level of financing allows them to hire and spend aggressively to satisfy customers and drive increased contract sizes.
Much has been made of the huge ($4.1B) valuation of that Intel-led round, but that's all misguided noise. Intel didn't make the investment for a financial return so the valuation isn't relevant. The "real" private valuation of Cloudera pre-IPO should be considered to be $1.8B, which was the valuation of the last financial round just before the Intel investment.
Cloudera as an Enterprise Data Platform
Large enterprises approach technology in unique and challenging ways. Their scale is much greater and they have an array of "must haves" that include fine-grained management, real-time support, reliability, recoverability, security, etc. Each one of these often requires substantial investments and together they are too daunting for most companies to address. That's why the exit strategy of so many "enterprise" software companies is to sell to a big player like Oracle, Microsoft, SAP or IBM rather than IPO.
Cloudera started with the "boring parts" like handling data collection and batch processing. These are the basic building blocks for the enterprise and serve as the foundation for what Cloudera delivered next, which was the ability to process unstructured data and enable real-time and interactive stream-based data processing. From this deeper base, Cloudera was there to bring their customers down the path to enable advanced analytics and machine learning.
In order to qualify as a real "platform" an infrastructure technology has to do more than work inside a single type of use case. Unstructured data, stream processing, and machine learning may be where all the sizzle is, but basics like data storage, management, and back-office processing must work too. Large enterprises are especially keen on platform technologies that bridge across multiple areas rather than having to adopt a new one for each. That's the key difference between what Cloudera has done versus other more niche-oriented "platform" companies like Hortonworks, MapR and Pivotal Software.
One big and illustrative product introduction from Cloudera was Impala. The lingua franca of the database world is SQL. Impala has many technical advantages over other approaches for analytics, but by using SQL it allows it to integrate with the existing array of tools and technologies already in use. For example, the mainstream business intelligence tools can incorporate analytics from Impala because it supports SQL.
Cloudera can be thought of as a software layer that sits on top of dozens of advanced data processing tools and makes them look like more traditional tools with extra services for managing security and performance.
We already mentioned Hortonworks (NASDAQ: HDP) which is a competitor, especially in situations where a very open source version of Hadoop is wanted but without all the enterprise bells and whistles. An all open-source solution is unlikely to ever address the specific needs of the large enterprise user. This is due to the highly variable environments there and the specific features needed. They don't translate well into general purpose open source projects.
MapR has built their own commercial and enterprise-flavored version of Hadoop infrastructure. MapR has a reputation for extreme performance but requires that customers use their own MapR filesystem and related toolsets. It's a good solution for many environments focused on performance but not as enterprise-friendly and versatile as Cloudera, especially when it comes to traditional data processing environments. In some cases, projects that are exclusively Hadoop-focused that demand very high performance might opt for MapR.
IBM is in the space with a whole stack of products that address development, deployment, business intelligence and traditional data management. However, the IBM product suite does not work so well with tools from other companies, which makes it problematic as a platform choice since many of the best-related data management products don't come from IBM.
The big traditional guys, Oracle, SAS and Teradata, are forces to reckon with. They own the installed base and have extended their traditional data tools with higher performance and more flexible offerings. They do tend to be much more expensive and this difference becomes profound when data volumes get very large. Design issues can make scaling up to large volumes on Oracle nearly impossible.
For example, a few years ago AMD migrated over 200 TB of data from Oracle to Cloudera due to outages and long recovery times. In addition to overcoming limits imposed by Oracle software, the performance of their systems was vastly improved in terms of response times.
Oracle is kind of a "frenemy" since they also want their products, like the Oracle Big Data Appliance, to work well in a Cloudera environment. For example, Oracle is a partner with a quote from the SVP of Oracle Server Technologies: "As the leader in Apache Hadoop-based data platforms, Cloudera has the enterprise quality and expertise that make them the right choice to work with on Oracle Big Data Appliance."
Given the size and scale of companies like Oracle, IBM, SAS and Teradata, Cloudera goes after three categories of use cases that are the most difficult for those vendors to compete for - customer insight applications, creating newly connected products and related services (like IoT) and next-generation risk management systems. Once installed, the Cloudera team will be in a better position to compete for additional projects, even ones that can be done with traditional data infrastructure.
Business Fundamentals and Valuation
The good news is that Cloudera is producing very strong growth, and if they can attain the margins in their long-term target model the company will provide great returns for investors. However today, like most SaaS companies, they are spending $2 for every $1 of revenue. Of course, it's recurring revenue and fairly "sticky" but one has to acknowledge it because any slowdown in the growth rate at this level of spending can be alarming. The market also can become suddenly impatient with this whole class of stocks as we witnessed during the first two months of 2016. That turned out to be a great buying opportunity (by March they were in full recovery) but it was painful.
We like to look closely at the dynamics of sales and marketing (S&M) spend and revenue growth. In the case of CLDR, the level of revenue generation to sales and marketing spend has been consistent at 1.2-1.4x during the past four quarters. Growth in S&M spend has been very lumpy so quarterly measures are not very helpful. On average though, revenue growth to S&M growth is 2x which is very good. Although the revenue growth rate has declined in the past few quarters, the efficiency of that growth has improved.
We do note that expense areas other than just S&M have been lumpy on a quarterly basis. For example, last quarter G&A expenses were up more than usual, even acknowledging the typical "IPO bump." As a private company these step changes are fine, but once in the public markets Cloudera might find these abrupt shifts unwelcome. The first few reports and conference calls with management will be closely watched.
Turning to Intrinsic Valuation (IV), the shares look inexpensive at the high end of the range $14. If one believes Cloudera will achieve their target model in the next five years then the IV comes in at $55-60/share. This is unlikely though, so we also ran a model (included here) that uses a 15% operating margin. That is still a stretch, but investors appreciate the recurring nature and profitability of the revenue stream. Using the more conservative IV we still get to a $35 price objective.
Subscribe to get our latest content by email.