Details on Palantir’s Investor Day held September 9th are here.held September 9th are here.
The Economist was correct when it recently stated that Palantir is “more than a technological project, it is a philosophical, even political one.” Palantir has a mythical and esoteric reputation in the Bay Area. The name is well-known and what the company does has circulated for years, which in a nutshell, is data mining for the government.
But until now, a customer list and any level of transparency has gone against the core purpose of the company. Therefore, I was somewhat surprised at the leak in 2018 that Palantir was considering a public offering as it seemed odd the company would operate openly and transparently. In fact, about five years earlier, the CEO had said an IPO was unlikely as it would make “running a company like ours very difficult.”
Nonetheless, the company is wanting to attract more commercial accounts and going public should help facilitate this. The old adage, “you can’t sell a secret” may be hindering Palantir’s growth especially as artificial intelligence startups raise their first and second rounds. Now is a good time to make sure to penetrate commercial accounts before AI bring more direct competition.
Below, I go over some of the folklore that surrounds Palantir and then I discuss S-1 filing. We aren’t dealing with a company where one has to wonder if the company or product will be popular on the public markets. Rather, we need to drill down into valuation and decide how much we are willing to pay.
The Folklore
Palantir can neither confirm nor deny if the software was used to kill Osama bin Laden, but the CEO required a body guard as of 2013, and it was generally understood for about a decade that Palantir had only one customer: the CIA. The company then grew to have three customers: the CIA, the FBI and the NSA.
By 2015, a leaked document from TechCrunch dated in 2013 confirmed twelve government agencies were using Palantir, including the “CIA, DHS, NSA, FBI, the CDC, the Marine Corps, the Air Force, Special Operations Command, West Point, the Joint IED-defeat organization and Allies, the Recovery Accountability and Transparency Board and the National Center for Missing and Exploited Children.” Palantir’s leaked document was the first time the CIA and the FBI had databases linked rather than siloed.
Nearly twelve years after Palantir was founded in 2003, that leaked document was the only record that indicated who used the company’s software. Palantir can be a divisive company that draws strong opinions from supporters and critics. Regardless of how you feel about the work Palantir does, one thing is for certain: as the IPO approaches, the company will dominate headlines.
Often those headlines will get it wrong in an attempt to frame Palantir in various lights. For instance, I don’t think anyone in Silicon Valley batted an eye at Alex Karp’s letter when the company exited for Denver. He stated that engineers “may know more than most about building software but they do not know more about how society should be organized or what justice requires. Our company was founded in Silicon Valley. But we seem to share fewer and fewer of the technology sector’s values and commitments.” These sensational headlines and CEO-centric storylines are distracting (i.e. who, then, does know how society should be organized and what justice requires? This is fairly presumptive and one of the many dramatic sentences from the filing).
When looking at things rationally, it’s probably better that Palantir be in Denver as government is a major industry in Colorado and being centered in the country will position the Palantir closer to Washington D.C. Palantir’s investors are not traditional Silicon Valley VC-firms, either. The company was likely there to attract top talent.
Palantir’s venture firm, In-Q-Tel, is located in Virginia and is funded by the CIA. This group has funded many projects, including Google Maps, Gitlab, Pure Storage, MongoDB, Cloudera and FireEye – but Palantir is on a different level as the CIA was the primary customer for many years. For these other companies, the CIA was not a primary customer. In-Q-Tel does not typically disclose funding rounds, amounts or dates. However, according to CNBC, Palantir received a $2 million funding round in 2004. Other investors include Peter Thiel, Stanley Druckenmiller and Tiger Global.
On the positive side of things, Palantir is believed to have helped with counter-terrorism, human trafficking and disaster response. On the more questionable side, the company has helped to deport immigrants and allegedly track law-abiding United States citizens without consent (i.e. spy software).
The company embodies “taking the good with the bad.” A former Marine, Samuel Reading, was quoted saying “it’s the combination of every analytical tool you could ever dream of. You will know every single bad guy in your area.” That, of course, implies having to know every good guy in the area too.
In the past, the Board has included Condoleeza Rice and former CIA director George Tenet, who said “I wish we had a tool of its power” before 9/11. The software was also allegedly used to convict Bernie Madoff.
Just when you think Palantir couldn’t be steeped in any more controversy — there’s more. In 2016, the company sued the United States Army for unlawful procurement solicitation for the Army’s internal intelligence software suite. Palantir argued the Army should be stopped from developing a risk-prone software project that would cost more than using Palantir’s software. In the end, Palantir won and the Army signed a $800 million contract over the course of 10 years.
So, why is Palantir going public now? Well, for one, it will be easier to gain corporate clients when (not if) the company becomes a stock market darling. The stock market is becoming a phenomenal source of free press and Wall Street will glamorize the company if it produces solid returns. This, in turn, will help Palantir attract more commercial customers and perhaps bury any ethical opposition.
The markets came close to burying the ethical issues around Uber. Perhaps this time it will succeed with Palantir. I also personally believe Palantir’s wide lead and lack of direct competitors (moat) will erode with artificial intelligence and machine learning. Time is of the essence to go public as AI startups need another few years before they can compete on this level.
Product:
Founded in 2003, Palantir is described as a company specializing in big data analytics. Palantir’s specific expertise in government intelligence and its existing ties to national security and the intelligence community differentiate its offering from competition.
The company has two platforms: Gotham and Foundry. These platforms allow organizations to combine core data with critical tools into a single platform to help users obtain actionable insights from a unified data asset. What Palantir tackles is the issue of data being siloed and ineffective for problem solving. These problems may relate to manufacturing, product development or customer experience.
The data Palantir gets is from the customer themselves and their existing databases although Palantir can crawl and scrape data that is freely available. For instance, Palantir can easily scrape public social media profiles but probably does not have access to private profiles except when the FBI issues government requests to Facebook.
The traditional deployment involves hosting Palantir servers in a customer’s data center. There is a cloud-based offering, as well, so the company can work across a range of hosting environments.
The company differs from a business intelligence solution like Tableau, Alteryx or Cloudera by answering questions that a model cannot answer. An example might be “how do we service car loans to people least likely to default” or “how do we catch fraud before it happens.” With traditional BI, it’s assumed you have the complete data set. Palantir tackles situations where a company may not have the complete data set. This is a crucial difference.
Palantir Gotham was the company’s first platform, built for government operatives in defense and intelligence sectors. The platform enables users to identify patterns hidden deep within datasets using semantic, temporal, geospatial and full-text analysis. Here are some ways the platform is used:
Graph: This application allows data objects to be seen as nodes and edges. Users can visualize events, filter objects and plot characteristics in a logical manner.

Source: Palantir.com

Map: This brings geospatial capabilities to track geo-located objects and events and to create heatmaps for the density of the objects.
Source: Palantir.com
Object Explorer: This feature is powered by the Horizon in-memory database, which competes with Apache Spark by letting users query billions of objects. The database provides further analysis for Map and Graph data.

Source: Palantir.com
Browser: This enables search queries for investigations and surfaces information, runs relevant searches, displays key data points and answers analytical questions.

Source: Palantir.com
Palantir Foundry is the commercial offering and has four layers of tooling: Foundry Core, Data Foundation, Ontology and Workflows. This four-step process does the following:
- brings volumes of data into one place,
- transforms the data into a format that analysts can work with and enables validation in any number of programming languages
- the “ontology layer” allows datasets to be turned into real-world concepts with the ability to accelerate on the company’s core ontology to reduce redundancy
- workflows is where it all comes together in an integrated environment for object exploration, point-and-click top down analysis, code authoring, time series analysis, data science and application development. When a user has a question, it answers it using all layers and tools available.
Palantir describes Gotham and Foundry as the “ability to construct a model of the real world from countless data points.” Unlike a SQL database, natural language is used to query data and return results in real-time rather than through strings. To some extent, Palantir resembles Elasticsearch in its ability to use a search stack to answer complex questions. For instance, Elastic is used to pair a passenger with an Uber driver or to process billions of log events from Sprint for outages or Fitbit to validate failures and for data discovery. Kibana can be used with Elastic to visualize the data. Another company where Palantir could potentially share the customer pool is Splunk or perhaps Sumo Logic. It’s not clear though if Palantir is price competitive with these other tools to be used in their place for analysis or if Palantir’s offerings are overkill for the analysis most companies require (to justify a higher price).
The truest, closest competitor for Palantir is Semantic AI, which supplies graph-based analytical platforms to the DoD and other government agencies. As stated, I think Palantir’s real competition is being developed as we speak as it will machines will answer questions from incomplete data sets once the AI/ML market is built out.
Some real-world uses for Palantir include Hershey’s using the software for global food distribution and to correlate weather patterns with snack consumption. Chase Bank and other financial firms use Palantir’s data analysis to catch fraud. Pharmaceutical companies to expedite the development of new drugs – this being a substantial use case this year and perhaps why Palantir’s revenue has accelerated.
Financials
The company grew revenue 25% year-over-year to $742 million in 2019. This accelerated to 49% year-over-year to $481 million for the six months ending June 30th. According to a Reuters article in June, the company is expecting $1.5 billion in 2021, which looks easily achievable. The company’s annual run rate based on the current quarter is about $1 billion.

Bloomberg reported from an unidentified source that Palantir’s revenue in the second half of the year is often larger during the fourth quarter due to government contracts being finalized. According to the article, Palantir books roughly 60% of revenue during the fourth quarter. For valuation purposes, we will run three instances between $1 billion and $1.5 billion. The higher number assumes Q4 is strong and the revenue acceleration we are currently seeing will continue. To simplify things, we will offer a scenario with $1.25 billion in revenue (see below).
Net losses for 2019 of $579 million were flat year-over-year compared to net losses in 2018 of $580 million. On an adjusted basis, net losses in 2019 were $337 million. The losses are shrinking with H1 2020 reporting a loss of $164.7 million compared to $280.5 million in the year-ago period.
On an adjusted basis, the company was profitable in the first six months of this year at $17.2 million compared to a loss of $167.6 million in H1 2019. This improvement in operating results was driven by increasing revenue and reducing the number of engineers needed to install and deploy software programs.
Gross margins for H1 2020 are at 73% and the company spent only 42% of revenue on sales and marketing.
The company has cash of $1.5 billion and debt of $297.6 million as of June 30th.
Contribution margin is a Non-GAAP key metric that represents the revenue the company generates relative to the costs incurred. It strips out variable costs related to deploying and operating the software and identifying new customers.
You can think of it as falling somewhere between gross margins and operating margins. For comparison purposes, Palantir’s gross margins are at 72.3%for the six-month ending June 30 and the company has negative operating margins of -48.5% and negative net margins of -78%.

The company states the addressable market is $119 billion across commercial and government sectors. The TAM in the government sector is $63 billion and the TAM in the commercial sector is $56 billion. Within the government TAM, domestic is $26 billion and international is $37 billion.
The commercial sector is the growth story. For example, Skywise is a solution that connects in-flight, engineering, and operations data to break down siloed systems around maintenance, flight management and aircraft monitoring and safety. Palantir is partnered with Airbus who offers this solution as “the leading data platform for the aviation industry.”
This example can extend to many industries, such as pharma for drug development data to better understand population dynamics and drug outcomes. This is for the pre-clinical and clinical stages, mapping treatment pathways, and automating reporting. Manufacturing can benefit from Palantir Foundry by managing inventory, saving on distribution costs and prevent delays while increasing sales.
There are also solutions for financial compliance, insurance, automotive and sales.
Valuation
Palantir is doing a direct public offering (DPO), which means there will be no new shares offered and no underwriters. The goal of a direct listing is not to raise money rather to allow existing investors to sell their shares. However, unlike Spotify and Slack who did DPOs, Palantir will have a lock-up period. I find a lock-up period to be more favorable for retailers Spotify took nearly two years to break out from its opening DPO price and Slack is taking about a year to break out beyond its opening price.
The company’s founders, Peter Thiel, Alex Karp and Stephen Cohen, own 30.2% of the company’s stock. Peter Thiel owns additional stock through various investment management funds that own stock, such as Founders Fund. Thiel has 28.4% corporate voting power, Karp has 8.9% and Cohen 3.1%.
There will be three classes of stock: Class A, Class B and Class F common stock – which is unusual to have three tiers. Class A will allow for one vote, Class B will allow for 10 votes and Class F will share 49.99% of the voting power for Palantir. Class F is for the founders who will retain just under 50% of the voting rights at all times. This is reminiscent of Facebook where insiders control about 70% and Zuckerberg controls 58%.
Palantir’s last valuation at $20 billion from 2015 is outdated as is the $26 billion valuation from last year. There were rumors in 2018 that Palantir was privately valued at $41 billion and this is probably closer to where it will trade on the public markets.
If we give Palantir a generous $1.5 billion in forward revenue, it’ll be trading at 20 price-to-sales at a $30 billion valuation and 27.3 price-to-sales at the $41 billion valuation. At the more reasonable $1.25 billion in current revenue, Palantir will be trading at 24 price-to-sales at the $30 billion and 32.8 price-to-sales at the $41 billion.

We can see below that trading higher than 32 forward EV/Revenues is very rare with most trading between 16 and 24.

However, IPOs have a way of pushing emotional buttons and there have been a few recent IPOs that have traded at exorbitant valuations. Zoom Video, Agora, Datadog and Lemonade have all hit the 50 EV/Revenues level.

Palantir is not profitable like Zoom Video and Datadog were at their IPOs. In fact, it’s a bit strange that Palantir has the losses it does with its vintage and guaranteed government contracts. I prefer to not pay over 40 P/S for any IPO as all of them have eventually settled under this number. I’m evaluating Snowflake next, which is a company I would value higher than Palantir due to fewer risks.
Point being, I’m a buyer in the $35 billion to $40 billion valuation on Palantir and then will respectfully wait on the sidelines. Beyond this valuation and I prefer to put my money to work elsewhere.
Risks
Palantir’s biggest risk is customer concentration with the top 20 customers accounting for 67% of revenue and the dependence on government contracts at 54% of revenue. The Army attempting to develop a more expensive in-house solution illustrates there is a risk that government agencies eventually move away from Palantir in the future.
Reputation and social acceptance is also a risk. Tech companies often see employees engage in protests when a company contracts with the government on AI-driven war missions and privacy issues that potentially threaten human rights. Palantir’s biggest obstacle today is the work it does with ICE which pits the company’s internal employees against the CEO on social issues.
For instance, Google ended a contract with the Pentagon when employees protested using AI for lethal purposes. Karp became controversial and challenged Google on this decision, saying it was a “loser” position. This can backfire as Palantir may not be able to attract top talent as AI companies begin to compete from a small pool of AI developers and engineers who have proven to protest and walk-out of company projects they feel are unethical. Amazon, also, banned facial recognition for law enforcement for one year following the George Floyd protests. Therefore, Karp’s personality could be considered a risk as the tech world begins to explore and support ethical AI development.
Despite government-backing, Palantir’s products are certainly not bulletproof. The company attempted to launch a platform called Metropolis to help hedge funds with trading, among other things. This platform did not succeed as hedge funds already possessed AI tools that were more of a complete package and the project was shut down. There are also rumors that the CIA has been cold towards the company since the CEO chose to be more in the public eye, especially around Osama bin Laden’s death. Palantir began linking to articles asserted their software was responsible for bin Laden.
To conclude, Palantir must be sensitive enough to win over commercial clients and top talent yet must not lose government contracts from being too overt. For valuation, I’ll cap it at $35 to $40 billion max.





























































