To The Moon

$Palantir Technologies Inc.(PLTR)$  

Micron Technology (MU) Blowout Earnings signal a significant shift in Market Economics even for Military and Cybersecurity Players like Palantir, Tenable and others.

Investment Thesis For Palantir (PLTR)

Shares of Palantir Technologies Inc. (PLTR) are down almost 40% year to date on the back of a market that is beginning to question which companies can reasonably thrive in the AI era.

While Palantir, on the surface, appears to be a major winner in the AI era, exponential growth in large language models, or LLMs, from the AI giant OpenAI (OPENAI), coupled with advancements from Anthropic (ANTHRO), seems to show that (with the right infrastructure underneath) replacing complex software like Palantir may actually be possible. Over the last few months, with both of the major foundation model companies announcing new consulting companies (in partnership with major PE firms), the odds that foundation model companies go for the application/workflow layer just jumped.

Palantir is thought to have a strong moat compared to other software companies. Yet, as we dig deeper, their key pillars to their moat are shrinking.

I predict this will accelerate. Software itself is increasingly not becoming a moat. Agile teams and relationships are the moats. It turns out some traditional government contractors and private sector players can match this when partnering with the LLM companies directly.

With this, I continue to believe Palantir is a Hold.

Why I'm Doing Follow-Up Coverage

On the surface, Palantir appears to be a solid way to play AI adoption at both the government and the enterprise level. For most of the last 3 years, I actually agreed with this. I originally wrote on Palantir back when shares were at $17.84/piece. I downgraded to a Hold last summer after a 10x+ increase in the price (and because I thought shares were getting overheated).

Now, I believe the fundamental story underneath is starting to fall apart. Especially since I last wrote on the AI software giant, foundation model companies have started to play in the Palantir workflow space. I’ll talk about this more later, but in hindsight, this was kind of inevitable. I believe the market is beginning to price this in with their 41% drop in shares since I last wrote on them.

With this, I think Palantir is at a structural disadvantage. This is why I am doing follow-up coverage.

Accelerating Disruption Risk

Palantir’s core business is their AIP platform, which helps governments and corporations automate critical workflows. For the sake of this discussion, you can think of it as a dashboard that helps orchestrate (set up) repetitive, typically white-collar job workflows. AIP is certified with FedRAMP, which is a certification software providers need to have in order to do business with the federal government.

I’m simplifying it because I want us to take a step back: none of what I just said is a moat in the age of AI.

Setting up workflows? These can be transferred to another orchestration layer with agents. Proprietary data? This is owned by the customer, often the U.S. government or an F500 company, not Palantir.

Proprietary AI models? Palantir does not own any themselves.

To be clear, Palantir has done a fantastic job of capturing the first few innings of the AI era sweeping corporate America and the U.S. government. The company’s 84% YoY growth with the U.S. government and 104% total YoY growth in the most recent quarters are rock-solid examples of this.

One by one I expect their moats to collapse. And with it, I think growth can slow.

Complex AI orchestration layers are becoming the core of the development companies both Anthropic and OpenAI have launched. These companies have raised billions of dollars to closely study human workflows. I do not believe that they will have any substantial deficit in know-how compared to what Palantir knows after a couple of years.

On the FedRAMP side, this used to be a real moat (it used to take years). Now you can get your platform certified in months. I expect the development companies that OpenAI and Anthropic have announced will both be extremely competent and will get FedRAMP certified quickly. They are both well capitalized.

With this, I think Palantir’s biggest flaw could start to be more damaging: customers have long complained about price. The justification is that Palantir delivers exceptional, unique value. I think Palantir will continue to provide a ton of value. My concern is that they will no longer deliver unique value.

Any market that moves from one dominant player to one with multiple competitors sees prices drop. I expect the high compliance orchestration layer market Palantir plays in will be no different. I do not see how they can keep margins (84% gross profit) so elevated going forward.

Valuation

My main issue with Palantir is the valuation. While most of this article covers their own disruption risks, I’d still be a shareholder if the valuation made sense. I don’t think the business has complete knockout risk over the next 5 years. I just think growth could be a lot slower due to a lot more competition.

Right now, Palantir trades at 84.14 times forward GAAP earnings (a 157.69% premium to the sector median). This growth assumes 96% YoY EPS growth and 40-70% EPS growth on a CAGR basis from next year through 2030.

Palantir, while in a good place to capture a ton of the GovCon contracts over the next few years, faces a new risk. Other players can now enter the space faster with quicker FedRAMP programs coupled with powerful LLMs that can build interfaces and workflows quicker.

Palantir’s long-time moats were their ability to comply with Federal regulations along with proprietary workflows. Both of these moats are shrinking. I think this means shares should not trade at such a valuation premium. In fact, Palantir’s forward price-to-sales ratio of roughly 33 is now higher than Anthropic’s rumored forward price-to-sales ratio of roughly 25 ($1 trillion valuation and $40 billion forward run rate on revenue). The company that owns the source IP (the models) trades at a lower valuation than the wrapper around the model (Palantir). This does not make sense.

What should they trade at? I’m not sure anymore. We need to see how quickly foundation model companies coupled with datalayers (like Databricks) can disintermediate them.

Bull Thesis

I think the biggest bull thesis may be looking at the same coin from the other side. On the one hand, being dependent on foundational LLMs to run your workflows means you put inherent supply chain risk into your system. On the other hand, if you can truly make your workflows model agnostic, then you get the best of any one model without the deep integrations that come with becoming exclusive with any one model provider (this is what the implementation companies Anthropic and OpenAI have launched will have to deal with).

On the surface, this is where the industry is going. Recent product announcements from OpenRouter show that some of the best models are synthetic models, or series of LLMs prompted about the same question in unison, and then using a judging model in between to pull the best parts of each response for the final output.

The problem for Palantir is that I believe the market is about to bifurcate. On one side, commercial, small business deployments will increasingly use open source models or OpenRouter to get high-quality LLM performance for a fraction of the cost of using a closed-source LLM (think OpenAI, Anthropic, Google, etc.).

On the other side will be the Mythos/Fable level models. These models will require KYC (Know Your Customer) account verification to use and may even be restricted by your citizenship (right now only U.S. citizens can even theoretically access Claude Fable even though Claude has restricted it to the public).

The problem for Palantir is that many of their customers will demand the latter models. Government customers will demand the best forms of intelligence even if these models require KYC or even security clearances to access. The TAM for this market is small, and Palantir services many of these workflows right now.

The risk is that, going forward, OpenAI and Anthropic move to capture more value. In other words, they copy the Palantir workflows and restrict these critical models to only their Development companies.

Logically, why would Anthropic or OpenAI expose these top-line models to anyone but their development companies going forward? The application layer is where a ton of value is captured. And, if the U.S. government is restricting your TAM, you have to capture as much value (make as much money) in the GovCon space as possible. It does not make logical sense for Palantir to command any of the value-capture here in the long run. That’s the risk.

Takeaway

While the Palantir story on the surface seems like an AI no-brainer, I think many of the reasons Palantir is well-positioned to disrupt traditional government contractors are the same reasons competition is intensifying for the government software giant.

Palantir’s moat is shrinking, and I think this is a massive narrative violation for a company that trades at 108 times forward earnings. As someone who originally called the stock a strong buy back in the fall of 2023 at roughly $18/share, I think the risks now outweigh the rewards (and this imbalance is growing).

Palantir’s biggest risks are traditional industry incumbents and in-house teams building competing solutions. Management has explicitly called out the latter one multiple times.

In other words, CEO Alex Karp and his team are right: internal competition risk is rising. Even as earnings came in strong, the company is too expensive on a forward multiple basis to price this in.

With this, PLTR shares continue to be a Hold in my opinion.