Palantir: AI-rebooted "Mysterious" Military Weapon

This year, while ChatGPT has gained popularity, Palantir Tech, a big data company, has also experienced a threefold increase in its stock price, despite only having official announcements and trials for its AI products and no established monetization model. Although Palantir has gained fame in the field of big data for its involvement in locating and capturing Osama bin Laden, this "mysterious" company with annual revenue of less than $2 billion has faced controversy in both society and the capital market:

Is their data collection and processing compliant? Can they sustain profitability in the future?

Nearly half of Palantir's total revenue comes from government clients, especially the military. However, the high conversion costs associated with government clients make it difficult to scale and monetize effectively. Therefore, after going public, Palantir focused on the capital story of "increasing penetration in the commercial market." However, in 2022, as economic expectations declined and companies reduced capital expenditures, IT spending decreased, making Palantir's commercial expansion in the second half of last year less smooth.

The "declining growth and uncertainty" caused Palantir's valuation bubble to burst rapidly, falling to as low as $12.5 billion, only half of its pre-IPO valuation. Dolphin Research believes that the "uncertainty in growth" stems from Palantir's heavy emphasis on customized products, which leads to slow customer acquisition. In reality, Palantir does not possess the "scalability" advantage of a SaaS company. Therefore, directly comparing its valuation to that of SaaS companies is not appropriate.

However, with the new wave of AI technology starting this year, Palantir, which happens to be in the spotlight, can it break through the penetration bottleneck in the commercial market with its hardcore AI product (AIP)? How much market expectation has already been priced in? Dolphin Research will delve into Palantir, focusing on these questions.

This article is the first part of the Palantir research, mainly outlining Palantir's development history, business model, and current investment logic. The second part will focus on a comparative analysis of operational capabilities (compared to competitors Oracle and Splunk) in the enterprise market, as well as reasonable growth expectations to evaluate Palantir's fair value.

Below is detailed analysis

1. How was the mysterious defense "blade" forged?

Palantir is a big data company that provides data integration, search, management, collaboration, and other big data functions based on knowledge graph technology. From the end-user experience, its product features include secure access, clear display of data relationship graphs, and a user-friendly UI interface.

Currently, Palantir's main products include four platforms: Gotham, Foundry, Apollo, and the newly launched AIP platform this year. The four platforms have clear positioning, targeting different end customers: Gotham (government customers) and Foundry (non-government customers), while Apollo is a support platform for automatic updates, repairs, and monitoring modules (with relatively low fees). Apart from AIP, which was recently launched and has not yet generated revenue, the main revenue-generating products are Gotham and Foundry, with a revenue ratio of 1:1.

Palantir went public at the end of 2020, during a period of global liquidity injection, and with the endorsement of influential figures, its stock price soared shortly after listing, quadrupling at one point. Therefore, it is also one of the hot stocks that many retail investors pay attention to in the capital market.

However, before going public, Palantir was not an unknown small company, especially in terms of technological strength. It was early recognized by the CIA, the US intelligence agency, and received investment from its subsidiary fund, In-Q-Tel.

1）Palantir is not a technology company starting from scratch. One of the "key figures" in the founding team is Peter Thiel, the co-founder of PayPal and an early investor in Facebook. The second venture of the PayPal co-founder is already eye-catching enough.

2）After its establishment, Palantir gained fame twice. One was in 2011 when it assisted the US government in successfully capturing Osama bin Laden (by identifying abnormal data representations of a particular villa, such as the absence of basic household waste, thus pinpointing bin Laden's location). The other was in 2009 when it discovered the Ponzi scheme of former NASDAQ chairman Bernard Madoff.

This magnificent operation also reflects Palantir's expertise in two application areas: counter-terrorism intelligence analysis and financial fraud detection. The former is the foundation of Palantir, while the latter has helped Palantir expand its business to commercial clients, with JPMorgan Chase and other banks being among the first customers.

However, compared to corporate clients, Palantir initially received more contracts from government and military sectors, which also provided Palantir with high technical endorsement (such as the cooperation with JPMorgan Chase recommended by the New York Police Department). So how does Palantir possess a competitive advantage in the government sector? Does this advantage have high barriers?

Dolphin Research believes that a combination of factors, including a focus on technology (talent), highly customized product services, and high lobbying conversion costs, have collectively built a moat for Palantir in the government business field.

1.1 Mysterious technological strength

Palantir has been providing services to the Central Intelligence Agency (CIA), as well as government and military sectors, for many years, so it is often seen as "mysterious". Due to the deep customization of its products, it is not suitable to directly compare its technological strength with other peers.

However, since its establishment, an increasing number of government departments have chosen Palantir as their big data analytics software supplier. In addition, the CIA, which has high requirements for data processing (a large amount of unstructured and fragmented data), had its venture capital fund, In-Q-Tel, invest in Palantir before deciding to build its own internal system. All these factors have provided strong endorsement for Palantir's technological strength.

So how did Palantir cultivate such a technological advantage? Perhaps we can gain some insight from its development history:

a. Source of technology: Founder Peter Thiel brought the fraud detection technology from PayPal to Palantir, laying the technological foundation for Palantir. In the early 2000s, PayPal's data processing and analysis technology was at the forefront globally. Therefore, by using PayPal's anti-fraud technology as its technological foundation, Palantir was already standing on the shoulders of giants.

b. Market entry: To be more specific, what PayPal really brought to Palantir was the ability to process and analyze data from multiple sources, i.e., how to find correlations from a large amount of fragmented data and provide valuable information through effective analysis. This technology is highly suitable for intelligence analysis work. At that time, the 9/11 attacks had just occurred, and the United States and the world were in the highest state of anti-terrorism sentiment. Therefore, in 2003, Palantir officially entered this vertical market.

c. Talent Incentives: In addition to creating a sense of mystery (intelligence analysis), Palantir primarily attracts top global technical talent through unlimited stock incentives. Palantir did not raise new funds through an IPO, but instead went public through a direct listing (DPO), which means there is no lock-up period for shareholders after the listing. The shares that are traded in the capital market come from employees selling a portion of their stock after the company goes public.

Furthermore, the company dilutes its equity by approximately 5-10% each year for executive and employee stock incentives, as well as for additional financing. However, this compensation method is not friendly to small and medium shareholders. A positive sign, though, is that the company's stock-based compensation (SBC) expenses are decreasing each quarter. With the growth of revenue, the proportion of SBC to total revenue has rapidly declined to around 20%.

Under the influence of the factors mentioned above (1)-(3), Palantir has gradually penetrated into customer groups such as the US military, the Department of Homeland Security, the Federal Bureau of Investigation, and police departments, relying on its technological strength. As of the first half of 2020, the military alone contributed $7,880 in revenue, accounting for 16% of the total revenue.

1.2 Highly Customized but Expensive Product Services

Compared to other big data analytics software, one major characteristic of Palantir is its high cost. This is not only because its products are highly customized, but also because Palantir's contract quotes to clients include maintenance and training fees (which are essential for customized products), making the overall package price higher. Only governments and leading companies in various industries can easily meet this threshold.

The following image shows Palantir's early pricing structure, which includes software licensing fees, maintenance fees, training fees, and some professional service fees.

For example, clients can choose to purchase a perpetual license (on-premises deployment) for a price of $140,000 per Gotham server core (the smallest server installation option, typically with a CPU ranging from a single core to 48 or 64 cores). Alternatively, they can choose to subscribe to Gotham's cloud service license at a monthly price of over $7,000 per core. In addition, there are annual maintenance fees, training fees, and professional service fees ranging from hundreds to thousands of dollars.

Compared to other software platforms that provide similar big data analysis services in the market, the average annual payment per customer for Palantir reached $6.3 million in 2022, with government agencies paying over $10 million per year.

Although the price may not be attractive, the high-quality products and comprehensive services provided by Palantir are crucial for intelligence agencies and other defense departments that deal with large amounts of complex information on a daily basis. Since Palantir won its first contract with the CIA in 2005, the US defense department has relied on Palantir's products and services.

An interesting anecdote demonstrates how Palantir's product and service excellence. In 2015, the US Army wanted to develop its own intelligence analysis software and intentionally excluded Palantir from the bidding process. Palantir protested, but when the protest was ineffective, they sued the US Army in 2016, claiming a violation of the 1994 Federal Acquisition Streamlining Act (which requires US government agencies to prioritize existing commercial software over custom development). The federal court ruled in favor of Palantir that year. Therefore, when the US Army re-bid in 2018, Palantir, as expected, defeated its competitor Raytheon (a large US defense contractor) and regained the Army's contract.

As a result, the government's reliance on Palantir's "nanny-like" customized product and service will only grow stronger, and the cost of switching will be higher. This has established a high barrier for Palantir in government business, making it difficult for competitors to replace it in the short term.

1.3 High customer conversion costs

So how did Palantir initially establish its connection with the government? This naturally involved the lobbying costs incurred during the channel establishment process. Although Palantir did not have a dedicated sales team in the early days, founder Alex believed that their product was good enough and did not rely on salespeople. However, winning government clients requires the involvement of intermediaries to convey messages, and the cost here is no less than that of building a sales team.

Since Palantir's products have always been more customized, even when penetrating commercial clients, they still require high marketing expenses. Currently, the sales expense ratio remains above 35%.

2. Doubts about growth, few optimistic voices

Despite being technologically advanced and having strong barriers to entry, Palantir's valuation was mostly cautious among institutions when it went public. Most of them gave a "neutral/hold" rating, and the target price did not have much premium compared to the pre-IPO valuation of $26 billion.

But the company painted a big picture for itself. In the prospectus, Palantir estimated a total addressable market (TAM) of over $100 billion.

The government market is estimated to be $63 billion based on 5% of government expenditures.

The enterprise market is estimated to be $56 billion based on the number of 6,000 enterprises with an average price of $9.3 million.

The total market size is estimated to be $119 billion. Considering Palantir's total revenue of $1.9 billion in 2022, its market share is only 1.6%, indicating that there is still a wide room for growth.

However, Dolphin Research believes that Palantir's growth path may not be as smooth as it thinks. The company faces different challenges in different markets. Specifically:

2.1 Government Market: Uncertain overall growth

Although Palantir has a wide moat in defense IT software investment, government demand also comes with inherent uncertainties:

1) Changes in a country's government budget are often linked to political factors, such as international situations and the preferences of ruling parties, making it difficult to predict and plan for long-term market demand.

Taking the example of U.S. defense spending, the scale has remained in the range of $600-800 billion over the past 30 years, with overall changes reflecting clear partisan preferences, except for special international disturbances. For instance, during the Obama administration (2009-2017) led by the Democratic Party, defense budget decreased year by year, while during the Trump administration (2017-2021) led by the Republican Party, the defense budget had an average annual growth rate of 5%.

Although the Biden administration is currently led by the Democratic Party, due to the impact of the Russia-Ukraine war and the confrontation between China and the United States, the Biden government has explicitly stated that defense budget will not be reduced in the coming years, deviating from the usual trend of annual decreases.

2) It is difficult to expand internationally when dealing with sensitive information

When it comes to the international government market, there are even more challenges in expanding the customer base. Due to the sensitive nature of data in the national security field, Palantir has limited opportunities to convert government clients outside of the United States. Currently, its main clients are the UK government, which is closely tied to the interests of the US government, and some non-defense government departments or organizations in countries such as France and Germany.

However, this year, the French Directorate-General for Internal Security (DGSI) chose to replace Palantir with Cybergem Chapsvision and Blueway after the four-year contract signed in 2019 expired. These two companies have been working closely with the French government. It is reported that the replacement was not due to issues with Palantir's products, but rather due to political security concerns. The French government has always wanted sensitive data related to national security to be processed on domestic software, even if it means higher costs (Palantir charges 10 million euros, while the French companies that replaced them quoted 40 million euros).

Therefore, whether it is the United States with its advantage of local relationships or the international government market, although the contract amounts for each department may be more substantial than those from enterprises, the actual uncertainty of revenue growth for Palantir is still high.

2.2 Enterprise market: Difficult to increase market share

Although Palantir has established a strong presence in the government sector, the capital market is more concerned about whether Palantir can achieve the same success in the enterprise sector. In the commercial market, the requirements for handling big data by enterprises are definitely lower than those of government intelligence agencies. However, on the basis of meeting the same technical requirements, enterprises will certainly consider cost-effectiveness and compatibility more than government institutions.

In order to enter the commercial market, Palantir launched a product specifically for enterprise customers called Foundry in 2016. Foundry is more customized and has the ability to integrate third-party software. However, the drawback is that the base price is higher than Gotham (the government version), which means that Palantir does not have a significant advantage in the competitive enterprise market.

(1) Customization: Compared to other competitors, Palantir's ability to handle heterogeneous data from multiple sources allows Foundry to delve into various aspects of enterprise data, such as production and management. After processing and analysis, it can then develop upper-level applications flexibly according to the needs of the enterprise. （2）Compatibility: The company can integrate other data analysis products from different fields as modules into the Foundry platform, such as ERP and SAP. In addition, although the client's platform is mostly customized, the Apollo platform can achieve unified updates for functional changes without the need for individual debugging for each client.

（3）High Price: As mentioned earlier, Gotham's base price for a permanent licensed server core is $150,000, or $7,000 per month. If 48 cores are needed, it means an annual cost of over $4 million. However, Foundry's base price is $100,000 per set per month, and for each additional set, the cost increases by $50,000 per month. Just purchasing three sets internally would cost $2.4 million, which means that an enterprise customer of Palantir would spend nearly $3 million per year on internal data analysis management.

The high price and limited versatility have slowed down Palantir's penetration in the commercial market. Although the company claims to have cooperation cases in more than a dozen industries, only top companies often have the purchasing power. For small and medium-sized companies, Palantir is simply too expensive!

Three, New Perspectives on Saving Growth Expectations: AI

The product is groundbreaking but lacks commercial scalability, which is the core reason why institutions felt that "the valuation is too high given the uncertain growth" when Palantir went public. Palantir's management has gradually realized this problem:

On the one hand, they are continuously trying to change the pricing model and introduce more versatile and compatible products.

On the other hand, in order to strengthen communication between the company and its customers (1. high price, 2. due to data security issues, Palantir has always had negative news), Palantir has also been investing in expanding its sales team in the past two years (rapidly expanding the sales staff since the first half of 2021).

However, Dolphin Research believes that deep customization is precisely the advantage of Palantir's existing products. If the focus is solely on adjusting the business model and changing the unique characteristics of the product, it may not be worth the effort to start over and create a "standardized" product.

Palantir's strength lies in its technology. Therefore, if there is a technological-driven industry transformation, Palantir is more likely to stand out. The current AI wave presents such an opportunity. Ultimately, AI large models are fundamentally a data problem, and Palantir's strength lies in the clarity, processing, and analysis of data.

In late April 2023, Palantir officially launched its newly developed AI product, the Artificial Intelligence Platform (AIP). However, the core selling point of the AIP platform is not Palantir's own development of large models like LLM, but rather its focus on AI applications:

a. Managing sensitive data for large model training: Leveraging Palantir's existing technological advantage in handling sensitive (military) data, AIP provides a secure application platform for governments, military forces, and commercial enterprises to use (train, infer) mainstream AI large models. The core feature of the product is the effective control of the data scope for large model training and operation (refining data granularity, defining API interfaces to expose data to large models). Due to Palantir's experience in handling highly detailed military sensitive data, it excels at setting up isolated protection for sensitive data, providing accurate and compliant decision analysis, and avoiding legal, regulatory, and ethical issues.

As AI applications continue to be implemented globally, users are increasingly concerned about AI compliance, ethics, privacy, and supervision. The launch of AIP aligns with this trend of demand.

b. AI assistant for the platform: Like other companies that provide software services to clients, AIP provides an AI dialogue entry function for the two client-facing platforms, Gotham and Foundry. Clients can use a question-and-answer approach to access the platform's modules and functions, reducing the learning curve of the platform.

Overall, Palantir's AIP aims to provide a secure and compliant AI application platform while also enhancing the user experience through AI dialogue capabilities. This strategic move aligns with the increasing demand for AI compliance and ease of use. As of the end of the second quarter, the company claims that since the launch of AIP 10 weeks ago, more than 100 institutions and over 5,000 individuals are using AIP, with a month-on-month growth rate of 50%. In addition, they are also in collaboration with over 300 institutions. Currently, AIP has not been commercialized, and Palantir's current goal is to involve more customers in the trial phase.

As for when to start charging and commercializing, it can only be said that there is currently no timeline, and it needs to be adjusted dynamically as the product improves. Palantir's management believes that they need to continue understanding customer needs in order to provide more in-depth services. At that time, monetizing the product will come naturally, and it is believed that this will not take too long.

The fact that they have not provided a timeline for commercialization indicates that the current surge since the unveiling of the AIP platform in May is not based on a solid foundation, but rather on excessive optimism.

Although Palantir's short-term valuation still cannot rule out the presence of a bubble, at least in the AI wave, Palantir has a definite ticket to enter. Therefore, Dolphin Research believes that Palantir still has value for long-term tracking and research, especially in the direction of data-driven business. Researching Palantir will greatly help the industry's development process.

In the next Dolphin Research report, we will evaluate Palantir's operational capabilities and growth potential by comparing it with two publicly listed big data industry competitors (in the commercial market field) - Oracle and Splunk, which have detailed financial and operational data. We will also provide a reasonable valuation judgment.

Risk disclosure and statement for this article: Dolphin Research Disclaimer and General Disclosure