There are many articles on Palantir Technologies (NYSE:PLTR ) at present, therefore, in this report we aim to cover aspects untouched by the investment community to date. The stock is shrouded in controversy and currently losing hundreds of millions, therefore, some context of the driving philosophical beliefs of the owners and understanding the journey thus far, is required to overcome some investor reservations.
The first half of the report is more qualitative and covers the drivers in PLTR's journey to date and considers some parallels with the 1980s/90s Microsoft for the reader to develop an intrinsic knowledge of the company to aid in filtering out future noise. The first half also covers competitive advantages and PLTR's market opportunities.
The second half is more quantitative and covers the financials (revenue, operating metrics, and free cash flow) and the valuation analysis.
The extraordinary trade of short and intermediate-term gains for long-term superiority has put PLTR several years ahead of the competition. In fact, currently there are no direct competitors. At present, the biggest competition is the desire for organizations to continue developing data connectivity in-house or with disparate third-party custom apps. This moat has been built upon a deep-rooted and enduring focus on extracting the best elements from humans and machines, and pushing the boundaries of what a user interface can achieve is what has brought this to fruition. The recent standardization of PLTR’s software has begun to provide operating leverage and has enabled them to target over 6,000 large institution/enterprise customers across all verticals – although the scaling stage of the business is still in its infancy. The company has reached $1bn in sales with a low number of customers and an immature business operation. As the company continues refining the business model and pushing gross and contribution margins even higher, this will be a first-class business, in our opinion. Once they gain further operating leverages and scale the business, they have a great opportunity to develop ‘flywheel’ network effects in multidimensions – within the customer’s organization, within specific verticals, and within PLTR itself.
Headquarters: Palo Alto, California. Also known as the heart of Silicon Valley.
Products & Markets: Gotham for Government business; Foundry for Commercial business.
The initial concept of Palantir Technologies was conceived by Peter Thiel not long after the sale of PayPal – in which he was the CEO and co-founder - to eBay for $1.5bn in 2002. Thiel drove PayPal to success by observing that the optimal way to combat financial fraud in online payment transactions was to combine data analytic technologies with human decision-making – rather than do it solely with one or the other. The banks either deployed more fraud analysts or more technology but didn’t leverage the two in a cooperative manner very well, and ultimately hackers were able to defraud in substantial measures. Thiel envisioned the way to prevent the fraudsters and beat the banks would be to intelligently cooperate machine with human to balance quantitative power with intuition and experience. Lo and behold this gave PayPal the edge in online payments which led to the eBay buyout. Subsequent to the PayPal sale and influenced by the 9/11 terror attacks, Thiel wanted to apply this tech and human synergy to assist in counterterrorism which gave birth to the initial idea of PLTR in 2003.
Around the same time, fellow PLTR co-founder and President Stephen Cohen, had recently graduated in Computer Science from Stanford, and had been playing around with various tech to begin an entrepreneurial venture; however, to no avail. He decided to spread the word across his network that he was willing to work on any pioneering software project which eventually put him in touch with Thiel which is when he learned about the PLTR vision. Joe Lonsdale, a PLTR co-founder and another CS graduate from Stanford, had interned at PayPal just prior to the eBay sale and was brought in by Thiel to work with Cohen on the programming aspects. Another Thiel/PayPal connection, Nathan Gettings, was also brought in for the coding. Once these four were together as co-founders, Thiel put forward several candidates for the CEO role who would then be interviewed by the CS grads and Gettings. Reportedly, Alex Karp, a former Stanford classmate of Thiel, was hands down the best person for the role, despite him not having any computer science background. From this point, PLTR had the financial backing and vision from Thiel, the computer expertise from Cohen, Lonsdale, and Gettings, and an X factor from Karp that pulled it altogether and made it work – although very slowly to begin with.
To understand why the first four members of PLTR wanted Karp to lead PLTR it’s useful to learn a bit about the CEO. Back in the early 1990s, Karp studied law at Stanford where he first met Thiel. After he graduated in law, he studied a PhD in neoclassical social theory at Goethe-Universität in Frankfurt, Germany. His thesis was titled “Aggression in the life-world” which was an extension of the works by a famous sociologist named Talcott Parsons. Structural functionalism is a sociological theory that underpinned the majority of Parsons’ work and appears to have had a profound impact on Karp and PLTR’s journey to date. This theory outlines similarities between sociology and biology. It states that, in the same way the human body is dependent on the body’s organs to function properly, society is dependent on institutions to function properly. In other words, institutions are vital for a healthy society.
CEO & Cofounder Alex Karp's Economic Beliefs
Source: depositphotos.com, Convequity interpretation of structural functionalism
For instance, as a society we need the police to combat crime and protect, we need courts and prison systems to punish criminal behaviour, and we need hospitals to maintain our health – these institutions are the ‘organs’ of society. We don’t know the details of Karp’s thesis as it’s written in German, though due to the ties to structural functionalism, it appears that Karp holds the role of institutions in very high regard. Therefore, his views aligned with Thiel’s desire to develop a pioneering software to help U.S. government institutions to fight terrorism and provide better protection for societies.
There is a strange, contradictory dynamic between Thiel’s and Karp’s overall philosophical views, however. The PayPal founder advocates libertarianism, which in large part entails skepticism of authority and state power and believes government should play a limited role in economies. Furthermore, Karp has been captured on audio expressing his disdain for Trump whilst Thiel has donated to Trump’s Presidential campaigns. We’ll touch on how a Biden Administration may impact PLTR later on.
Right-wing libertarianism – the prominent form in the U.S. – appears to encapsulate Thiel’s views more narrowly when one considers the Trump affiliation. This form advocates for conservatism on economic issues and liberalism on personal freedoms. The latter ties into the need for national security that protects westernized nations from autocratic and/or communist forces. Therefore, the CIA becoming PLTR’s first customer followed by the FBI, the NSA, state police departments, and other defense/intelligence/law enforcement agencies, is the common ground for Karp and Thiel. This mutual agreement on the importance of defense, intelligence, and law enforcement institutions has provided a solid foundation for Thiel and Karp to work together to shape PLTR and to have a meaningful and gratifying way for them to positively impact the world.
The main driver that has pushed PLTR’s software to be so different and far ahead of the competition is Thiel’s preoccupation with both the user and machine working together and getting the best out of each other. Practically all other big data analytics companies are striving for AI/ML superiority to the point that their technology can make human-like decisions by itself and remove the need for as much manual labour. The possibilities of AI have been widely touted for over two decades and it has continually felt that the true breakthrough whereby machines can make qualitative decisions is just around the corner. The reality is that AI-enabled machines are very unlikely to be able to make qualitative decisions in our lifetimes or even in the 21st century. Machines are infinitely better than humans in quantitative decision-making, though, humans are infinitely better in decisions requiring intuition, subjectivity, and qualitative consideration. And this isn’t going to meaningfully change for a long time. PLTR intrinsically understood this right at its inception in 2003, and hence set on a mission to build software that could extract maximum benefits from both humans and machines.
Maximizing the Potential of Human and Machine
Source: depositphotos.com, Convequity modification
To achieve this, PLTR has heavily focused on the user interface and process design – more so than any other big data analytics company. As it is the user interface that enables the user to maximize the power of the machine. This may sound obvious, though what is also apparent is that the rest of the competition hasn’t taken this notion as deep as they should have and this is why they are several years behind.
A major obstacle for software companies in creating high-quality user interfaces is not truly obtaining the deeper and nuanced needs of the end users. Inquiries into the true needs are often made via surveys, though, usually users do not know what they need to make their work easier because it doesn’t currently exist, or they have a vague idea but can’t articulate it, or can’t be bothered to articulate it on the assumption the vendor won’t have the interest to make the changes. This is why PLTR sends its engineers on the front-line to spend long periods of time with software users – whether it be military operators in Afghanistan or workers at the plant of an automobile manufacturer – in order to gain deep understandings of what will make the users work more productive without the user being required to explicitly articulate it themselves. No doubt the main reason some other vendors haven’t followed suit to the same extent is the cost of this approach – such customization for each customer is extremely expensive and very low in operating leverage. PLTR has not made a profit yet and in the past several years have made annual losses to the tune of several USD hundred million. However, the years of customizing has built up incredible experience to enable PLTR to create two broad platforms – Gotham and Foundry – that can each serve most of the user needs encountered. And this is where the operating leverage and scale is coming from – more on this later.
Until around 2Q19, Thiel and Karp have resisted the temptation of implementing a sales force and instead have allocated capital to Forward Deployed Engineers (FDE). As previously explained these engineers work with users on the front-line to customize PLTR’s software for their nuanced needs. Some may argue it would have generated better ROI for investors if PLTR had deployed sales reps instead, however, such critics are misunderstanding a key drawback of the typical S&M and R&D functions. In the traditional software business model, sales reps secure the deal and then relay the required needs and specs to the engineers. This communication, no matter how clear it is intended, will lose a considerable amount of nuance and context that can only be captured by the engineer if they can see the problem in person. Hence, the role of the FDE - and as a result PLTR have the world-leading software in regards to sophistication and user intuitiveness.
This expensive and arduous approach is a manifestation of the founder’s often expressed mantra of trading short-term gains for long-term dominance. By 2008 and 2009, five and six years from the company’s inception, revenue was only $2m and $16m and costs would have been multiple times greater. It’s rare to see such patience, especially for software firms that are expected to scale in a much shorter horizon. Revenue growth sharply decelerated in 2017 and 2018, to 10.5% and 15%, respectively, due to a refocus on product over sales to further improve their solutions. The majority of other software firms at this age are predominantly geared toward sales growth to satisfy investors. The underlying driver for this approach by Thiel yielding huge upfront costs, minimal ROI in the early stages and excessive product innovation and refinement before sales with the goal of long-horizon market leadership, is his views pertaining to Mimetic Theory originating from academic works by French historian Rene Girard. It relates to how humans tend to seek value by imitating what others are doing, though, in the end this leads to the competing for scarce resources and dissatisfaction. Thiel has often explained how Mimetic Theory transfers to business and it has essentially molded his view on how to create a successful company. He believes imitation is bad for technological and economic progress, often citing ‘competition is for losers’, which has certainly shaped the conception of PLTR.
Thiel realized that to operate a business with no competition, one needs to create something fundamentally new to serve customer needs radically better, and invest heavily over a long duration to build enduring competitive advantages and high entry barriers.
In relation to the graphic below, we currently see PLTR as having recently entered the scaling phase. The scaling ability is derived from three layers; 1) the two extremely broad platforms that can serve a multitude of different use cases, 2) the enhanced data integration connections that has radically reduced the time and number of engineers needed to install and deploy the software platforms, and 3) Apollo, the SaaS delivery mechanism introduced c. 2018 that allows engineers to more efficiently manage upgrades – in 2Q20, 41k upgrades were installed versus 20k in 2Q19 – and reach air-gapped systems previously unavailable to SaaS delivery, like air traffic control.
Fortunately, with the ideological beliefs and capital of Thiel, and his proclivity for missions for the greater good over near-term wealth accumulation, PLTR has had a distinct advantage over other VC-backed software companies that are often pressured for immediate high growth or profits not too long after the investment.
Interestingly, but not coincidentally, PLTR’s focus on the bridge between user and machine also touches on the highly political topic of people losing their jobs to machines. In a CNBC interview in February 2018, Karp spoke about how his company’s software is preserving jobs unlike that of many Silicon Valley firms. And the software is achieving this by increasing the productivity and quality in decision-making of front-line workers, to the point where machine replacement is no longer an improvement. Within his discussion regarding job preservation Karp also said “… without jobs there would be no democracy”, which brings it back to the whole serving for western societies narrative.
Another USP of PLTR’s software is that it is designed with both the nontechnical and technical users in mind. Nontechnical users love the intuitiveness, visualization features, and the ease of navigation whilst the software also provides employees who can code to write new scripts and develop customized apps. In comparison, Splunk’s software appeals to particular groups of well-trained analysts and is highly domain-specific. Splunk’s software cannot empower a police officer to swiftly correlate various data points across different databases and siloed systems to find likely suspects in a complex fraud investigation and then make these findings presentable for his superior to quickly digest. When it may take a PLTR-powered nontechnical user 20 minutes to arrive at a fairly solid conclusion, it could take a user using other software all day. It is this nontechnical appeal that sets PLTR so far ahead, in fact, even developers reportedly prefer using PLTR’s software as it speeds up their process and makes them more productive.
PLTR’s existence thus far has been a lot more mission-driven over profit-driven and this has pushed them way ahead of the competition, if there is any at present.
In the early 1980s before Windows, Microsoft partnered with IBM to produce the disk operating system, MS-DOS. A fair degree of customization was required to make MS-DOS compatible with each manufacturer’s PC, drastically limiting the scalability of Microsoft’s business. Apple was the only firm to vertically integrate to produce its own hardware, OS, and programs for its PCs – Lisa and then Macintosh in 1984. Except for what Apple was doing, the industry was incredibly fragmented, both vertically and horizontally. Clearly, Bill Gates and Microsoft saw a huge opportunity to create a PC agnostic, comprehensive OS, all developed in-house, known as Windows. Microsoft partnered with manufacturers to have Windows pre-installed which lowered the price of many PCs. The competitive pricing led to widespread adoption of Microsoft’s Windows as the default OS. By the mid-1990s every buyer wanted a Windows-enabled PC for compatibility with the mass of users and Microsoft dominated the PC market without even being a manufacturer. Becoming the standard OS led to most developers designing applications specifically for Windows – creating one of the first Flywheel Effects. Apple was left with a market share multiple times smaller than Microsoft’s because of the inability to scale.
There is a similar narrative taking hold with PLTR and its software that can be overlaid atop existing OS – whether it be Windows, iOS, Android, Linux, Unix – and integrated with all types of databases and even isolated systems such that are found in submarines, airplanes, space exploration, air-gapped networks, and legacy tech. Gotham and Foundry and all the incorporated programs are completely infrastructure agnostic and can extract and manipulate structured and unstructured data from any system across an organization. Such versatility expands PLTR’s Total Addressable Market (TAM) to include every large institution and large enterprise operating in westernized nations. PLTR management’s estimate of the number of potential customers on the commercial side is 6,000 to approximate a TAM of $56bn. They have not estimated the number of potential government customers, though, instead via IMF data, they project a TAM of $63bn – and at a guess we suspect there may be a few hundred potential customers that make this up. As of 30th September 2020, PLTR had 132 customers including both commercial and government agencies, needless to say they have an opportunity to increase their customer base by over 50x. From a revenue perspective, given that consensus expected FY20 revenue is $1.07bn, the potential upside is 111x [($56bn + $63bn) / $1.07bn = 111.2].
The above calculations are based on PLTR completely dominating all their TAM, which is unrealistic, so we shall revise downwards. In 2006, when Microsoft was 31 years old, Windows had 97% of the OS market share, though due to the emergence of smartphones and tablets and open-source software market share has dropped substantially. Given PLTR is not in the consumer industry like Microsoft predominately was in the 1990s and therefore will not be assisted by such consumer trend tailwinds, future TAM percentage penetration will be less. Though, if we forecast PLTR can capture 35% of TAM by, say 2040, that would still produce an impressive CAGR of 20% for the next 20 years.
Before Windows was first introduced in 1985, MS-DOS was Microsoft’s disk operating system. MS-DOS had to be tweaked and customized for each individual make of computer. The deeper compatibility of Windows allowed Microsoft to scale the business far more efficiently. PLTR have taken a similar journey. Initially they made customized software for U.S. government agencies and after observing many data problems across all types of organizations are fundamentally the same, they have been able to standardize their products. Ultimately, this standardization has powered a huge degree of scalability to their business.
Asides of the PC-agnostic element, Microsoft beat Apple to PC dominance by building an interface that was appealing for different types of users. Windows had an intuitive graphical user interface (GUI) for casual users and also had a command line for users that could code and write scripts. Apple’s Macintosh only had a GUI thereby alienating much of the techie communities. Furthermore, the programs Microsoft added to Windows such as Word, Excel, and PowerPoint, had far greater appeal to the business world than what was included in the Macintosh. This broad user-type consideration powered Microsoft to outcompete Apple in the PC world.
PLTR’s software first and foremost is designed to empower nontechnical analysts and business managers to easily navigate and digest data to enable better decision-making. However, the software also provides platforms for users to write code and develop apps. In essence, PLTR’s software appeals to all users across the technical spectrum as well as to the needs of users according to their position in organizational hierarchy. This parallel with Windows could prove to be very compelling when many other big data analytics vendors mainly target the technical analysts of an enterprise.
Whilst Microsoft’s rise to fame was initially ascribed to Window’s appeal to the masses, subsequently they penetrated the SMB and the enterprise markets. We view PLTR as having similar TAM expansion opportunities in reverse. At present, PLTR exclusively serve the government and enterprise sectors, however, as experience of user/data problems increase they will probably be able to further standardize their products to meet the demands of a wide set of use cases in the SMB market.
PLTR has two main products: Gotham for government, and Foundry for commercial. They are implemented for each customer as central operating systems, and additional programs are incorporated for specific use cases. These may be existing programs or new programs tailor-made for the customer. The software has become extremely popular with end users for predominately three reasons: 1) the intuitiveness and ease of use, 2) the ability to mold to any underlying infrastructure, and 3) the ability to integrate with disparate data sources in order to extract, make connections, and present in easy-to-digest forms. These three USPs raises the speed and quality of human decision-making thereby having a multitude of positive ramifications on organizations’ missions and objectives. As we enter deeper into the era of hyperconnectivity with tailwinds such as IoT and 5G, PLTR’s competitive edge is only going to be amplified.
For investors it’s particularly intriguing how PLTR are balancing the trade-off between standardization and customization. Having Gotham/Foundry meet, for example, 60% to 90% of the data needs of a customer, and then to complete the solution through customization strikes a potentially very prosperous blend of scale and product fit.
Policer officers in the LAPD have shared the difficulty in connecting data from legacy siloed systems when investigating – Gotham effortlessly enables officers to sift through information and unify a mosaic of data of which to draw possible conclusions. In the wake of COVID-19, oil producers turned to PLTR to optimize extraction and exploration operations. The drop in travel forced airlines to incorporate Foundry to manage seat positioning and resource planning amid the unprecedented idle fleets. The U.S. and Colombia are two of numerous countries that are using PLTR’s customized software to map out COVID-19 outbreak areas, overlay this with local hospital capacity, and allow central and local governments to coordinate more efficiently. And the list of examples goes on and on.
We’ve already touched on how PLTR make their products so superior – by sending engineers, known as FDE, to witness at first-hand the nuanced needs of the users. From 2Q19 to 2Q20, PLTR reduced the average deployment time five-fold to 14 days. Typically, consultancy firms and/or system integrators take several months longer, often rendering the software out-of-date by the time it is usable. This advantage over the consultancy-based tech firms is derived from PLTR’s standardized Gotham/Foundry platforms. Whereas consultants like Accenture typically review the problem, scan the market for the best collection of solutions, and then integrate them into the customer’s IT infrastructure, PLTR speeds up the process by already having a platform that can serve the majority of needs and which can be easily tailored for specific use cases – reducing costs customer costs considerably. This unique blend of software vendor and consultant is a competitive edge materialized by having the FDE customize on-site alongside the user. Large scale remote software upgrades via Apollo are going to reduce the man hours needed to maintain Gotham and Foundry, thereby freeing up the company to focus more on new sales.
In our opinion, PLTR’s product quality and deployment time is only going to get better. As the company permeates through verticals, from one customer to the next, industry specific knowledge will refine the product and experience will speed up the deployments.
Lastly on products, Gotham and Foundry are implemented to enhance existing systems – not to replace them. This circumvents most stakeholder objections who are in some way viscerally connected to legacy IT implanted in their organization. Moreover, it makes implementation far easier.
PLTR estimate the commercial TAM is $56bn and consists of c. 6,000 large enterprises and the government TAM is $63bn, broken down into $26bn from the U.S. and $37bn from international sector.
The commercial TAM could be understated if we estimate from a different angle. The following snippet from Gartner’s research indicates PLTR could be operating within a larger TAM. Gartner forecasts spending on enterprise software and IT services will be $1.5tn in 2021 – two sectors that PLTR currently operates. This is complete guesswork, but say PLTR targets the largest 1% of companies that account for 20% of spend, then $300bn could be a more realistic TAM for the commercial side of business. This may seem excessive though remember this is the TAM not a likely future annual revenue amount.
Gartner Worldwide IT Spending Forecast
Currently PLTR are mainly known as the data specialists that can extract, unify, and present all an organization’s data irrespective of where it is located, however, management’s longer-term goal is to become the default central operating system for entire industries. The Skywise partnership with Airbus is a prime example of this. PLTR’s software boosted Airbus’s production of its A350 by 33%, and the success led to a broader initiative, named Skywise, to help the entire aviation industry. Skywise is on open data platform to aid in better decision-making for all stakeholders in the aviation industry – and couldn’t have come at a more needed time.
PLTR has the opportunity to become drastically more than just data aggregators; it’s within their reach to become centre to the operations in various industries. With that in mind, a commercial TAM somewhere between $56bn and $300bn is more probable, in our opinion.
So, if we assume the current price of PLTR reflects the management guided $56bn of commercial TAM, and the real TAM is revised upwards in light of the Gartner estimate approach, there is significant upside once the market can visualize the greater opportunity. Especially, when considering the lack of legitimate competition – which we’ll cover under the Competition section later.
In the government sector there is even less prospect of legitimate competition. The closest rival would have been Google, however, cultural and political beliefs of its workforce is likely to prevent expanding ties with the U.S. government. Google employees opposing the possibility that their AI technology could be used for lethal purposes by the Pentagon, caused the tech giant to pull out of the bidding for the $10bn contract in 2018. As a result of the conflict between its workforce and certain needs of the U.S. government, Google are unlikely to be a future direct competitor to PLTR.
PLTR’s importance within the U.S. government is only going to grow in the foreseeable future. One manifestation of this is PLTR’s legal win against the U.S. Army in 2018, that ordered the army to pay for PLTR’s services instead of attempting to build an alternative solution in-house. PLTR sued the U.S. Army for unlawful procurement in connection with violating a 1994 law – the Federal Acquisition Streamlining Act. The Act states that the government must conduct proper market research and assess whether there are any commercially viable items that can meet needs without modification. The Army was developing something more expensive and less capable than what PLTR can deliver, rendering the court decision in favour of PLTR. Reportedly, officers on the front-line in Afghanistan resorted back to using PLTR instead of the Army’s own Distributed Common Ground System (DCGS) due to the superior capabilities. Subsequently PLTR were assigned to a $876m ten-year contract with their court rival.
PLTR are sweetly positioned for either austerity or pro-growth initiatives by the Biden Administration. If budgets tighten, PLTR are the most cost-effective option versus internal efforts. If budgets expand, U.S. government agencies will likely want to spend more with PLTR to explore further capabilities. At some point in the not-too-distant future, when COVID-19 is behind us, governments around the world will be pressured to reverse the fiscal deficit accrued as a result of the pandemic, and we see this occurrence as the catalyst driving PLTR’s importance further. It won’t be a bigger revenue driver than a pro-growth fiscal scenario, though will increase the necessity of having Gotham, and in turn drive stronger growth on a longer horizon. In this narrative, the government TAM is conceivably higher – probably closer to $90bn – than the estimate by management.
Realistically, how much of the TAM can PLTR potentially capture, in say, 10 and 20 years from now? Reiterating from earlier, that complete dominance for a B2B market is extremely unlikely, future annual revenue amounts should be considerably lower than any estimate of TAM. On the other hand, PLTR don’t have legitimate competition. There are some possible competitors in the future as listed below, but each have a disadvantage against PLTR:
C3AI - a level or two below the sophistication of PLTR. Doesn’t really have advanced algos and is mainly a Hadoop end service provider and ML/AI application development tool.
Google - one firm that could probably give PLTR a run for their money but are not likely to go down the customization route and will not compete for defense-related government contracts.
Splunk - software is way too techy and domain-specific to compete with PLTR.
Raytheon & Lockheed Martin – not likely to compete with such defense contractors. In any possible head-to-head, PLTR’s higher-quality, more cost-effective, and more agile solutions will comfortably beat rigid legacy approaches. Collaboration much more likely.
Accenture – absent of any broad and reusable software platform. Historically has acquired startup software packages to compete in the consultancy game. As a result, such consultancy firms have a repertoire of disparate specific use-case software that often cannot be deployed and hence they need to build yet another custom application from the ground up or bring in a third-party.
The lack of rivalry increases the possible penetration of the TAM. In effect, the only real competition at present is PLTR’s potential customers and their desire to develop in-house.
Whatever TAM an investor assigns to PLTR’s opportunity, it could potentially be enlarged if PLTR were to ever serve the SMB market. By considering the level of talent and technological excellence, producing a low-touch, standardized data platform for smaller companies seems an easy thing for PLTR to do. Perhaps, when PLTR have penetrated their primary TAM and have allocated more expenditure to S&M functions, the SMB market will be a strategic move to make.
PLTR’s deep-rooted competitive edges have been established by the focus to produce software that maximizes the strengths of both human and machine for users all across the technical spectrum. The user interface is integral to this, therefore PLTR have taken the extremely long and onerous path of working side-by-side with users to tailor the most intuitive software for each customer. In 2016, the culmination of knowledge gathered via innumerable custom-made software solutions led to the development of Foundry, a platform that can be applied to a very wide set of use cases. Improvements in data integration connectivity has dropped down average deployment time from months to 14 days, and Apollo makes software maintenance far easier, liberating resources to focus on new customer acquisition.
In the big data space, there is simply no other company that can:
Not only do they have these USPs, PLTR are now beginning to scale their operations. Probably the most important competitive edge for the future is the last bullet point above. PLTR’s platforms can be laid on top of any infrastructure, and can accommodate any future application that the customer adds. It enhances existing systems and will evolve with the changing IT landscape. Martial Arts legend Bruce Lee’s water analogy really springs to mind,
“Be formless, shapeless, like water. You put water into a cup, it becomes the cup.”
If any of this is true then PLTR is far less likely to experience similar fates of current legacy IT players burdened with huge amounts of technical debt and the inability to be dynamic.
Another future competitive advantage, that will secure monopolization, are the possible network effects – both for the customers, the industries, and PLTR.
Once Gotham/Foundry are installed and interwoven to the organization’s IT fabric, adding more users and more use cases generates further benefits for the customer. As PLTR’s software facilitates and harnesses user feedback, more users strengthen Gotham and Foundry. Moreover, the combination of more business users and developers also strengthen Gotham/Foundry – users can easily describe their data problem in which developers can swiftly respond to with bespoke app solutions. Generally, in due course PLTR has reported that customers become self-sufficient in using and enhancing the performance of Gotham and Foundry. This ecosystem evolution between users and developers will eventually make PLTR extremely ingrained into the core of customers’ operations – and vital to sustain their competitive advantages.
PLTR can deliver industry projects in a similar vein to the Skywise project, whereby initially helping Airbus transform the efficiency of their supply chain, PLTR created an open-source platform for the benefit of 100 airlines and 15 suppliers. The more industry stakeholders added to such a system provides additional benefits for all and solidifies PLTR’s position. The S-1 indicates there may be industry partnerships coming soon in healthcare and financial services.
The data feedback loop is incredibly powerful for PLTR to continue improving their solutions. More industries, more customers, and more users, will create more data for PLTR to digest and develop even deeper intrinsic knowledge of organization’s data problems. Being able to identify even more patterns and more prevalent underlying problems to solve will only fortify the power of Gotham and Foundry. And this is a stretch, but if Gotham/Foundry reaches a critical global mass, they may become crucial for general IT compatibility purposes just like Windows did. This is an absolute best-case scenario, however.
At first glance PLTR’s bottom line looks worrisome – $853 net loss in 3Q20 versus $140m net loss in 3Q19. However, in anticipation of the Direct Public Offer (DPO) proceeds (which ended up accruing to $943m) the company ramped up spend multifold on S&M, R&D, and G&A. And they are likely to be unprofitable for a few years to come as they attempt to scale and dominate their market. Therefore, in this section we shall focus on a few performance indicators.
In 3Q20 government sector revenue grew 68% and commercial grew 35%. On the face of it the discrepancy may seem alarming and question Foundry’s product-market fit. However, it must be noted that Foundry was created in 2016 and the initial business strategy was to move all existing commercial customers onto the platform rather than seek new customers – as evident by the sluggish growth in FY17 and FY18. By FY19, all existing commercial customers were fully integrated and therefore the business could focus on new customer acquisition. So, when taking into account that Foundry has only been targeted onto new commercial customers since FY19 and sales teams were only assembled in FY20 – representing only 3% of the workforce – 35% is impressive indeed. And given the earliest stages of TAM penetration, this growth rate should increase in our opinion. In regards to the growth trajectory of government revenue, from the high base of 68% growth and $420m of revenue in 9 months through FY20, one would assume a sharp deceleration, however, there is $1.3bn worth of government bookings to turn into future revenue and an additional $2.6bn subject to funding approval. In our opinion, government revenue growth will decline but at a slower pace than what the consensus believes.
To model PLTR’s future revenue growth we’ve created peer groups based on age of company and a $1bn revenue level that serve as anchors for our revenue forecasts. Each peer group consists of stocks that are older than PLTR and therefore may offer insight into future growth trajectories. Below is the age-based peer group.
Bearing in mind PLTR is 17 years old, we charted the growth rates of this peer group when they were between 15 and 20 years old, as shown below.
Distribution of Revenue Growth Rates from 15 to 20 Years Old Tech Companies
Source: Convequity research & analysis
This sample is far from exhaustive though 20 stocks can still offer some insight. Typically, tech companies that are between 15 and 20 years old are growing revenues between 10% and 30%. PLTR was in this bracket when it was 15 and 16 years old, though at 17 years old in 2020, YoY 9-month revenue growth (to 3Q20) is 50% and based on management’s full year guidance of $1.07bn, FY20 growth will be 44%. The following chart shows PLTR’s historical growth versus the peer group at the same ages along with our forecasts.
Age-Based Peer Group Average Revenue Growth vs PLTR
There is pervasive skepticism that PLTR can sustain anywhere near the level of growth achieved during the first 9 months of FY20. This is reflected in the consensus 29% CAGR in the next 24 months (through to 3Q22). And this doubt is in large based on the above charts, however, typically when a company is circa 17 years old and have reached $1bn in annual revenues, operations are scaling and customer numbers are much higher than PLTR’s 132. Essentially, PLTR have reached the $1bn revenue mark with an immature business operation – a sales force only introduced in FY20, low customer numbers, just entered the scaling phase, and recently listed on the stock market. It is unusually immature for a company with $1bn in annual revenues indeed. And as early signs show a fair degree of operating leverage and management appear very focused on scaling with efficiencies, it is completely plausible that PLTR can sustain the growth levels indicated by the green line in the above chart. Also bear in mind that many of this particular peer group didn’t have the cloud to scale when they were of similar age.
Such peer group analysis is incomplete without considering the absolute revenue levels. Therefore, the same peer group has been modified to only include companies that have surpassed $1bn in annual revenue. We’ve also added in Workday. It is a company younger than PLTR but given it has surpassed the $1bn mark we decided to include it.
Aged-Based Peer Group with Annual Revenues Greater than $1 Billion
Below is the above information in chart forms. From this age and revenue-based peer group, we honed in on the SaaS stocks of Salesforce, Workday, and ServiceNow as one subgroup given PLTR’s cloud service ability, and the tech giants Microsoft, Oracle, Cisco, VMware, and Salesforce as a subgroup to envision long duration possibilities (Salesforce is in both groups). Together these two subgroups provide an anchor for our forecasts.
As can be seen in the above table and charts below we project that PLTR’s growth will be toward the upper levels of the distribution – though is this realistic? Well, let’s take our projected 40.6% average growth for PLTR in the next 5 years and compare to Symantec, Juniper Networks, Workday, and Splunk – 4 companies in operating in the enterprise sector. These firms’ 5-year average growth post $1bn in sales were 37.7%, 41.1%, 35.8%, and 35.5%, respectively. In our opinion, PLTR’s sophistication and moat is orders of magnitude greater than that of these 4 companies when they had just reached the $1bn mark. In fact, a growth trajectory closer to Microsoft’s is more likely given the parallels with we outlined earlier.
Revenue Growth X Years After Surpassing $1bn in Revenue
Unsurprisingly consensus revenue estimates through FY22E are in line with the mean. We’ve also used the mean but as an anchor rather than to dictate our forecasts.
Our forecasts actually go out to FY44 using the tech giant subgroup as an anchor due their age. A 10-year DCF valuation doesn’t make sense unless sustainable growth is likely to decelerate to the risk-free rate by year 10. Given the opportunity ahead this is unlikely therefore we’ve extended out the explicit forecast period to 24 years.
The following chart shows how the above projected long-term growth rates translate into annual revenue amounts. Based on the discussion pertaining to moat building throughout this report, we feel a 37% TAM penetration by 2044 is conceivable and probably slightly underestimates.
Annual Revenue Based on Tech Giants Actual Growth Rates and PLTR's Projected Growth Rates
The scaling of PLTR’s business is evident in the gross and contribution margins. As shown below, there has been a rapid increase in the latter as a result of reduced average deployment times from months down to 14 days – a few hours in some cases. This has been enabled in part via many years of experience and also thanks to their advancements in data integration connectors that make life a lot easier for FDEs. Subsequent to the initial installation, Apollo also contributes to lower customer ‘success’ costs by delivering updates etc. via the cloud. Note that margins exclude stock-based compensation.
Source: Company filings, Convequity presentation
The following chart shows how PLTR’s customer acquisition costs (CAC) has been trending lower during the past few quarters. CAC can be calculated by revenue * (gross margin less contribution margin). The rise in 3Q20 will likely continue, however, as DPO proceeds are ploughed into customer acquisition.
Customer Acquisition Costs & Implied Net New ARR
Source: Company filings, Convequity calculations and presentation
In the S-1 document, it was stated that the first quarter was usually a poor quarter on a sequential basis due to the ramp up in business in the third and fourth quarters as decision-makers hasten the purchasing process before year-end budget amendments.
As a result of PLTR’s net new ARR, the gross margin, and the CAC during the past 6 quarters excluding 1Q20, the average CAC payback period was 56 months. In the most recent 3Q20, the CAC payback period was 25 months. CAC payback period is calculated by dividing the previous quarter’s CAC by the current quarter’s net new ARR multiplied against gross margin. This tells you how long it will take to recoup the CAC from revenue specifically generated from the customers acquired in the previous quarter – it’s not perfect but is a fair indication.
The CAC payback period is one of the most important metrics for SaaS-related companies because extraordinary expenses are usually needed to acquire new customers. Meritech Capital do great research on SaaS stocks, and the chart below is from their website. As PLTR’s background is part-consultancy, they are usually excluded from such group SaaS stock analysis, however, increasingly more of their business derives from prepackaged software delivered from the cloud and proportionally less is from consultancy. Therefore, comparing some metrics with SaaS groups of stocks is insightful. This is why we’ve edited Meritech Capital’s charts to show where PLTR is positioned against other SaaS vendors.
SaaS Stocks CAC Payback Period
Source: Meritech Capital, Convequity editing
For anyone who follows SaaS-related stocks it is abundantly apparent from the above chart that the payback metric is correlated with multiples. For instance, most stocks with a payback of 12 months or less have an EV/NTM Revenue of >25x, and most with a payback of 36 months or more have a multiple of <10x. Thanks to Meritech Capital this relationship is presented in the following chart, again taken from their website. And we’ve edited in PLTR’s data.
EV/NTM Revenue vs CAC Payback Period
Source: Meritech Capital, Convequity editing
At the time of writing, PLTR’s EV/NTM Revenue is 42x. If the last 6 quarter (exc. 1Q20) is representative of PLTR’s true CAC payback period going forward then the above chart suggests they are grossly overvalued relative to this SaaS group. However, if the 3Q20 CAC payback period of 25 months is more of an accurate reflection going into 2021, then PLTR appears (just about) fairly valued relatively speaking.
If PLTR can continue the upward trend in contribution margin (as shown a few charts back) then the payback period will carry on shortening. In the S-1 filing the contribution margin was presented in three phases relating to the customer relationship – Acquire, Expand, and Scale. The general takeaway is that as customers move through the phases the sales costs as a proportion of revenue declines. Therefore, the contribution margin pertaining to a single customer will increase as they are moved onto Expand and then Scale. Currently, due to the low customer base, the highly negative contribution margin in the Acquire phase has a significant impact on the overall contribution margin. As the customer base grows, new customers will naturally represent a smaller proportion of all customers, and hence the Acquire phase contribution margin will have less impact. So as time goes on the overall contribution margin will be closer to the Scale phase than the others. And further improvements in the length of time for customer deployments will expedite this. Below are extracts from PLTR’s S-1 (includes through 1H 2020) to help digest the above.
S-1 Extract: Contribution Margin Across Customer Relationship Phases
Through 1H20 the contribution margin in the Scale phase was 68%, however, for some customers it can rise 20% higher as shown in the second extract from the S-1 below.
S-1 Extract: Scale Phase Contribution Margin of Top Quartile
In the S-1 it’s not stated whether customer size is relevant to the contribution margin. Therefore, the customer size mix of the top quartile above is unknown. All of PLTR’s customers are considered large, though within the base it would be interesting to know how the largest, smallest, and average sized customers impact the margin. Due to the lack of disclosure, we can only speculate. In the following chart we’ve surmised, on a qualitative scale, contribution margin impacts across the three phases from the largest, medium-sized, and smallest of PLTR’s customer base.
Contribution Margin by PLTR Customer Size Speculation
In the Acquire phase we suspect the biggest headwind on contribution margin comes from the largest of PLTR’s customer base. However, a degree of cost of revenue is fixed (primarily hosting), therefore, these can be spread out further with a larger customer and drive-up gross and contribution margin to very high levels in the Scale phase. Moreover, overall contribution margin is likely to be higher with the larger customers thanks to the greater land and expand opportunities. Conversely, PLTR are likely to spend less on CAC to win a contract with smaller customers, putting less strain on the Acquire phase margin, though less fixed cost leverage and less land and expand opportunity on the table will likely render overall contribution margin to be relatively less.
So, for future reference, any disclosure regarding customer sizes may offer additional insight into the evolution of the contribution margin and what a mature level looks like.
Based on the aforementioned opinions regarding contribution margin, we forecast the following.
FCFF margin (exc. Stock-based compensation) for 9 months through FY20 is -37%. With the scaling effects aforementioned we suspect this might improve slightly to -35% for the full year. Modelling PLTR’s FCFF margin is tricky because it is currently very negative compared to SaaS stocks at the same stage of annual revenue.
SaaS Peer Group FCFF Margin at $1bn Revenue Mark
Additionally, as FCFF margins are more volatile than revenue forecasting it the same way as we did revenue is likely futile. Therefore, we decided to simply estimate a maximum that PLTR may reach in the long-term. As we project PLTR will become a future enterprise tech giant we referred back to the Enterprise Tech Giant peer group we used earlier.
Except for Salesforce, each of these tech giant stocks have a mix of on-prem and cloud business. At present PLTR also has a mix and will likely do so in the future, therefore, using these software giants to benchmark a possible ending FCFF margin is logical. For our valuation model, we chose the mean of the mean highlighted in red in the table above, as PLTR’s FY35 FCFF margin target. In the model there are accelerative improvements in FCFF margin in the short-term followed by more gradual gains thereafter. In FY35 and beyond the FCFF margin remains at 29%.
PLTR Long-Term FCFF Margin Forecasts
We have applied a DCF in two ways to estimate the intrinsic value of PLTR’s stock. The first method is the conventional Enterprise DCF approach, and as an alternative we valued PLTR by forecasting contribution margin.
Estimating PLTR’s beta requires a good dose of judgement because the stock has only been trading for 3 months. Because of the limitations of mathematically calculating the beta we referred to a SaaS group to estimate PLTR’s beta for the beginning of the forecast period. For FY35, when we project PLTR will be in a mature stage of business, we’ve used the Enterprise Tech Giants current betas as a guide.
Starting & Mature Stage Beta Guides
Source: Investing.com, Convequity presentation
We’ve decided to use a beta of 1.6 for the beginning of the forecast period to reflect the high revenue concentration risks in comparison to the SaaS stocks listed above. Some analysts may argue it should be even higher, though the long-duration revenue visibility – an average contract duration of 3.5 years – will counter a degree of the concentration risk. For FY35E, we’ve decided to use the current mean of the Enterprise Tech Giants of 0.91. Despite the COVID-19 induced dislocation between cloud/tech and the broader market, we feel 0.91 is a fair long-term target beta. We checked Microsoft’s and Oracle’s beta all the way back to 2005 and they have been consistently lower than 1.0 by a large margin.
Below we present the inputs for discounting PLTR’s future FCFFs calculated as forecasted revenue multiplied by forecasted FCFF margin for each year of the forecast period (through to 2044). The path from ‘present’ to FY35 is linear for the dynamic variables. The Equity Risk Premium (ERP) will, of course, change but for simplicity we’re assuming a constant of 4.9%.
In regards to the risk-free rate, we’re not macro economists but we feel it’s appropriate to assume a degree of rate normalization throughout the forecast period. And as the 10-year risk-free rate is probably the best proxy for the sustainable growth rate, we’ve also used 2.5% to estimate PLTR’s value in 2044 and beyond. As a result of the forecasts and the discounting parameters we generated the following sensitivity tables.
Intrinsic Value Per Share, P/S NTM, & FY35 Beta Sensitivity Tables
If the FY35 beta estimate is too low, and say, 1.1 is more realistic, then based on the price at the time of writing ($25.00) one could argue it’s at least fairly valued. The purple highlighted value per shares in the third sensitivity table shows what we think are possible true intrinsic values based on estimation error in the FY35 beta and the ERP. It certainly feels like there is a lot more upside than downside based on this valuation analysis. Even if the mature stage (FY35) beta should be 1.30, and we assume the current implied ERP is accurate for the forecast period, then the intrinsic value equals the price at time of writing.
Below is the Enterprise DCF value breakdown by time and shows the reconciliation from EV to equity value.
Enterprise DCF Breakdown by Time
As can be seen, through FY25 there is actually no value being added according to our forecasts. This is typical for many tech companies willing to spend huge upfront costs in the intermediate term in order to dominate the market in the long-term. Thiel takes this concept further than most, however.
For this valuation approach we use exactly the same WACC and associated discount input parameters. We used our forecasts for contribution margin (shown earlier) and customer number forecasts to break down the value into all existing customers – we estimate to be 142 by FY20 year-end – and all new customers accrued in FY21 and beyond. This approach also required us to calculate the cost of indirect corporate expenses to subtract this from the value derived from existing and new customers. The intrinsic value per share we arrive at is $49.45.
We estimate that the average value per existing customer, and per new customer, is $180m and $140m, respectively. This indicates that there is substantial revenue growth opportunity within the existing customer base as well as in future customers.
We see the main risks to our valuation as the following:
PLTR have a great opportunity ahead of them to develop ‘flywheel’ network effects to build atop their software that is already several years ahead of the competition. In regards to big data analytics there is no other company that comes close – they have, in effect, created their own market. The reality, however, is that they will become much more than a ‘big data’ vendor – their platforms have the capability to become central operating systems for organizations. The software is infrastructure-agnostic; it can be laid on top of any group of networks and systems and enhance the performance of everything it connects with. The high-entry competitive advantages give them a great chance of grabbing a huge chunk of the management guided $119bn of TAM. We suspect the TAM is quite a bit higher, however. Our research has brought us to really consider that PLTR could be a future tech giant.
This article was written by
Long-time tech investors with special interests in cybersecurity and cloud-related stocks. Recently we decided to turn our passion into an equity research business called Convequity. We combine quantitative and qualitative methods to gain a deep understanding of a company's business, products, and markets, and the stock's intrinsic valuation. Our process aids us to identify companies in-process of developing wide and sustainable moats with the promise of exceptional long-term returns.
Disclosure: I am/we are long PLTR. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.
Additional disclosure: Dear Editor, I apologize in advance for the length of this article. As there is a lot of noise surrounding PLTR at present I believe it's crucial to present the deep intrinsic knowledge to investors before they dismiss or take the risk. Future articles will not be this long as I see this stock report as a special case. However, I completely understand if you require me take parts out - hopefully you don't.