This term is taking ever more precedence in the context of today’s AI market. This doesn’t mean fake revenue or operating profit. Rather, these shenanigans are essentially a “gray area” where companies can use… methods to artificially boost their financials to make them look more appealing.

As I was perusing through Meta’s 10-K’s, I noticed some disconcerting things. Take a look:

This, ladies and gentlemen, is what we call a uncommenced lease commitment. Companies can use these commitments to spend massive amounts of money on datacenters and AI infrastructure – without having to include them on balance sheets.

The deeper you go, the worse it gets. AI companies’ circular financing web, depreciation, etc., etc.

One thing to keep in mind as the article is read is that I am not a permabear. I am not telling you a crash is due; my job is to inform you all. The rest is left up to your interpretation.

Section 1: A Depreciation Doozy

At the core of the AI boom lies the hardware. All AI companies use hardware at some point in their development process. For the sake of simplicity, let’s assume that we have a company that uses expensive AI servers.

In 5 years’ time, the AI server will likely still turn on, the machine is probably still working. But by then, Nvidia has released a substantially better version of the same AI server. The hardware runs faster and is more efficient. Suddenly, the old machine still turns on but is worth much less economically.

This is the difference between physical life and economic life.

A company can spend $6 billion on AI servers. If they depreciate over a three-year time period, that equates to $2 billion in annual depreciation. However, if the same company chooses to state depreciation over a six-year time period, the depreciation cost shrinks to $1 billion.

Many companies in the AI boom are using these shenanigans to overstate their earnings. They reduce depreciation and therefore boost net income and operating income.

Meta has recently increased their useful life assumptions on their AI infrastructure to boost operating income. Amazon moved in the opposite direction, choosing to lower useful life assumptions as they believed that AI/ML advancement was happening quickly.

Nothing about the server changed when the accounting estimate changed. No new customer appeared. No additional product was sold. Yet billions of dollars of reported earnings can move.

The interesting part isn’t just that these companies are changing useful life assumptions; it’s the contradiction between AI narratives and accounting assumptions.

Investors are constantly hearing the narrative that “AI innovation is unprecedented,” or “Models improve every few months.” Well, that’s fine and all, but if technological improvements are occurring at such and accelerated rate, then why should certain economical assets become useful for longer?

Most of these AI companies that are being mentioned (Meta, Google, Oracle, etc.) all have real profits. Real net income, real operating income, and real margins. This method of accounting shenanigans is a way of artificially reducing expenses; choosing instead to move current expenses to a future period, put simply.

Section 2: The AI Boomtown: Everyone is Selling to Everyone Else

Imagine a city where a veritable gold rush was taking place. As history shows, we would expect the rush to create its own economy. Someone sells picks, another person sells land, another person lends money to the miners, someone finances the railroad.

This is good and all, but is any of this economy making money from an independent source? Absolutely not. This gold rush example exposes a broader market flaw taking place in the AI market right now.

This is very interesting behavior on the AI companies’ part. Company A sells an asset or service to Company B. At the exact same time (or via a pre-arranged agreement), Company A agrees to buy back the exact same asset, or something of equal value, from Company B. This, in essence, lets both the companies inflate their top line numbers while hiding their money spent in Capex or choosing not to include them at all.

The estimated value of this “circular loop” is estimated to be more than $800 billion. We are seeing deals popping up everywhere, from Nvidia to Oracle.

To understand this problem more clearly, let’s assume that we have some hypothetical (but an example that is being seen in the real world) money flowing in through the AI market.

Nvidia invests in an AI company. The [hypothetical] AI company uses this money to buy compute from a cloud provider such as CoreWeave. CoreWeave needs more capacity to satisfy the contract, so they buy some more Nvidia GPUs. Nvidia records more hardware and records more cash. Some of that capital gets invested back into the AI ecosystem.

The cycle repeats again. The interesting part is that investment at one end of the chain can eventually return as revenue at the other.

Right now, one could convincingly argue that a large part of the AI ecosystem relies on companies exchanging money, no? And the larger problem in all of this is that the money being exchanged between the two companies can directly inflate the other’s top line numbers, creating an illusion that it’s organic revenue.

The danger stems from counting the same demand signal at every layer of the stack and mistaking repetition for diversification.

Section 3: AI Companies aren’t Asset-Light Anymore

For years and years, our “Mag 7” were incredible asset-light businesses. The appeal of these businesses was that Microsoft, Meta, and Google could scale revenue without needing a proportional increase in their physical assets. Microsoft could sell another software license, Meta serves another add, and Google could process another search.

The marginal economics of this period of time were exceptional, even outstanding.

As AI took over our world in a flurry, these economics have changed drastically. Now, these same companies need land, data centers, hardware, cooling, etc., etc.

These companies are not software giants anymore; rather, they are starting to accumulate obligations that are making them look more like heavy industry.

The problem most investors are facing (without realizing it, subconsciously) when analyzing an AI company is asking the question of “how much debt does the company have?”

This simply isn’t the right question to ask anymore. Debt is only one form of a fixed commitment. A company can also have operating leases, purchase commitments, SPV-backed arrangements (we will get into this later), etc.

In today’s market, a company can look lightly leveraged on a conventional balance sheet screen while still having committed itself to enormous future cash payments.

When AI infrastructure commitments were a small part of cash generation in any given company, this problem would have mattered much less. When these “future commitments” are measured in hundreds of billions of dollars, they matter much more.

Alphabet and Meta are prime examples of this. As of the end of June 2026, Alphabet has already committed to a whopping… $85.2 billion in future lease commitments. Even worse, Alphabet has not recorded these statements on their balance sheet.

Meta is at an even more extreme level than Alphabet. It has disclosed $279 billion in lease commitments that have not commenced as of June.

This is not “hidden debt” in the fraudulent sense. The commitments are disclosed. The problem is that most investors never read far enough to see them.

And leases are only one way these obligations can sit outside the debt number investors usually focus on — special-purpose vehicles take that a step further.

A special purpose vehicle is essentially a separate legal entity created by a company for one specific project.

Suppose a company wants to build a data center. Rather than that company borrowing tens of billions of dollars, a separate entity can raise the money, own the project, and lease the data center to the original company.

There are many significant real-world examples of this, but one of the easiest to visualize is likely Meta and Hyperion.

Meta has a massive, even colossal data-center project in Louisiana. This data center campus is named Hyperion. The interesting part about Hyperion is where the financing lies. Meta partnered with an entity named Blue Owl Capital to create a joint-venture project for Project Hyperion.

Meta owns a 20% equity stake in this project while Blue Owl has the remaining 80% stake in the venture. The joint venture covers $27 billion of development costs plus the cost for cooling and connectivity.

Meta did not simply borrow the $27 billion themselves. Instead, Meta contributed outside land and development materials to Blue Owl, while Blue Owl contributed the outside capital to fund the project. Meta then signed leases to use the facility once it was completed.

The key accounting point to be gleaned from this example is that Meta does not directly own this project, so they do not have to record this $27 billion lumpsum on their balance sheet. This Hyperion project shows that AI infrastructure can be financed outside the parent company while the parent can still be directly involved in the project.

In today’s market, checking companies’ debt balance no longer tells the true story. The project can sit with an entirely different project vehicle, while the parent company can be the customer and pay using leases to make the project financeable.

AI, all things considered, is into an infrastructure business with years of fixed obligations attached. The question stops being only how fast revenue grows. It becomes how much capital the growth consumes.

Section 4: “Adjusted EBITDA”

For companies whose revenue is decided by GPU’s and servers, how much should investors be trusting a profit metric that removes the cost of those assets aging?

CoreWeave is possibly the best example of this problem. CoreWeave reported $1.51 billion in adjusted EBITDA in Q2 2026. Revenue was also up 112% YoY. These headline numbers look very good.

Look at the flipside. CoreWeave’s GAAP operating loss was -$49 million. It’s adjusted net loss was -$567 million. This is the same company in the same quarter.

Adjusted EBITDA is essentially removing several items that CoreWeave specifies, including stock compensation and acquisition-related costs. EBITDA can be useful. My problem isn’t the existence of the metric. My problem is what happens when the “D” being removed represents the very machines generating the revenue.

CoreWeave’s adjusted EBITDA reconciliation starts with a $626 million net loss. Then it adds back $1.393 billion in depreciation and amortization, $640 million in interest expenses, among smaller items.

Suddenly, a $626 million net loss goes to a positive $1.51 billion adjusted EBITDA. It gets worse, though. CoreWeave depreciation and amortization alone equaled roughly 54% of quarterly revenue.

Earlier, we saw how much changing a server’s useful life can change depreciation. Here, CoreWeave goes one step further: adjusted EBITDA removes depreciation altogether.

If AI hardware really remains economically productive for many years, EBITDA may be more informative. But if GPU generations become economically outdated quickly, depreciation represents something very real:

the gradual consumption of an asset that will eventually need replacing.

Depreciation itself is growing incredibly fast, even faster than revenue. Q2 2025 D&A was $559 million; in Q2 2026, D&A is $1.393 billion. That’s roughly a 150% increase in one year. Revenue increased 112% over the same period. CoreWeave D&A is currently growing even faster than revenue.

CoreWeave’s $1.5 billion adjusted EBITDA isn’t fake. But neither is the $1.4 billion of depreciation which is right underneath the EBITDA. For a company like CoreWeave, where the core business model is rapidly evolving ang growing, the aging of these machines can’t just be an afterthought.

Section 5: Are the Numbers Fake?

The revenue is real, the growth is real, the data centers are real. There is no secret conspiracy going on amid the AI boom.

The real problem that we’re looking at isn’t really whether the AI bubble is going to pop. The problem is that strong growth doesn’t tell us how much capital was required, how long the assets really stay valuable, or the economics of leases and commitments.

To successfully identify the state of the AI economy right now, we need to figure out what it takes for this exponential growth to happen, and how much profitability survives after the costs necessary for this growth to happen.

If we look back to the late-1990s, there is a perfect example capturing effectively what is happening right now: Telecom/fiber. Fiber demand has proved to be enormous. But during the late-1990s buildout, too much capital was chasing the same infrastructure, leading many investors to lose money.

The lesson to be gleaned from this: A technology can be transformative and still attract more capital than the economics can support. AI can keep growing rapidly while certain companies still suffer.

My concern isn’t that AI companies fabricated their revenue. It’s that investors may be using accounting metrics built for one kind of business to value an industry becoming something very different.

So I’m not asking whether AI is real anymore.

I’m asking something harder: after the depreciation, financing, leases and reinvestment are accounted for, who actually earns an attractive return on all this capital?

Spending time in Asia completely shattered my belief about Uber’s global moat. I realized that Uber’s competitive advantage has gone down substantially. Frankly speaking, Uber might never have had a competitive advantage in the first place. Uber’s entire business model is easy to replicate; something that many corporations can copy.

That is why I decided to write this thesis on Uber, arguing that their business model is in a time of stress, and this could cause Uber’s share price to decline considerably.

1: The Competitive Advantage Shrink

One of the core problems that Uber is facing as of right now, being one of the main focuses of the article, is a significant decline in Uber’s competitive advantage. Uber is trading at a premium valuation because the majority of Wall Street is treating Uber with a “winner takes all” mentality. The truth, however, is much starker than this. The data that we will cover suggests that aggressive regional competitors are actively compressing Uber’s core unit economics.

Lyft, Rapido, Ola, Bolt, inDrive, not to mention many others (including Waymo, which we’ll cover later). These names should sound familiar to you, because these platforms are directly competing with Uber.

Since the software required is cheap to deploy, local operators can launch a competing platform in a city within weeks, even days. They bypass global corporate overhead and can offer lower commissions to drivers and cheaper rates to riders.

This problem can be seen in Uber’s Mobility economics. Gross Bookings climbed from $21.2 billion in Q1 2025 to $26.4 billion in Q1 2026, but Mobility revenue barely moved, rising from $6.5 billion to $6.8 billion. As a result, Uber’s Mobility revenue margin, or take rate, fell from 30.7% to 25.8% over the same period. Uber’s still processing far more ride volume, but it is capturing a smaller percentage of that value as revenue. This does suggest that competitive pricing and regional pressure may limit how much of each booking Uber can ultimately keep.

2: Financials

Wall Street just loves a good growth story, doesn’t it. In order to help to Uber “perma-bulls” with their thesis, Uber’s recent Q1 financials gave bounties (in the plentiful) of vanity numbers to celebrate. Gross Bookings up 25% year-over-year to $53.7 billion, and trips also up 20% to 3.6 billion.

If we dig deeper, however, some problems are starting to take precedence.

The first problem that needs to be addressed is the growing divergence between ride volume and money hitting the top line. Even as Gross Bookings grew by 25%, reported GAAP Revenue only rose by 14% to $13.2 billion. Even worse, when you adjust for currency shifts and look at core constant-currency organic revenue growth, it decelerated down to just 10% YoY.

They are processing more trips and racking up more miles, but they are retaining a smaller slice of the gross transaction pie.

The second problem in Uber’s financials is as follows: For Uber to defend its market share against aggressive side platforms and local rivals, it has to constantly burn cash to acquire volume

This reality is showcased by their massive $14.8 billion acquisition of Delivery Hero’s portfolios. Uber finished the quarter with just $6.1 billion in unrestricted cash and short-term liquid investments. They are taking on massive debt and utilizing equity to buy low-margin food delivery volume in contested foreign territories where they have zero pricing power.

3: Autonomous Vehicle Problems

As autonomous vehicles are starting to take precedence in today’s world, Uber’s bull case narrative relies on one equation: When human drivers are removed, labor costs drop to near zero, which causes Uber’s margins to expand significantly.

However, as we shift towards automation, Uber’s “asset light” business model shifts into a low-yield value trap, while the asset holders have the true pricing power over Uber. Bulls are automatically assuming that AV (autonomous vehicle) developers will rely on Uber indefinitely. This just cannot be the case.

Alphabet’s Waymo and Zoox are two of the first autonomous vehicle developers; the first of many, if you may. In their own deployment hubs, Waymo handles everything in its app, from bookings to routings to customer acquisition.

When these autonomous vehicle developers (eventually) achieve scale, why would they pay Uber a massive middleman fee?

The image above is directly from Waymo. The user interface (UI) is very similar to Uber, proving that when these giants get to scale, ousting Uber may not be as difficult as it seems.

I suppose bulls can make the argument here that Uber has scale, connecting the AV developers (Waymo and others) to over 200 million users almost instantly. Well, sure, but there’s a major problem with this thesis.

Even if an AV developer does partner with Uber, they would hold all the cards. Industry models suggest that the fleet operator could capture the majority of autonomous-ride economics, while the marketplace layer may receive a much smaller share.

The AV developers will take the larger part of the per-mile fare to make up for the billions spent in R&D, leaving Uber fighting for an ever-shrinking slice of the pie.

4: Regulatory Quagmires

I don’t want to get too much into legalities here, but it can present a problem for Uber.

The first big legal/regulatory problem lies in the transition from human drivers to autonomous vehicles. If a self-driving system experiences some sort of a glitch or failure, the legal claim is automatically classified as a systemic product liability suit. Uber can no longer shield its balance sheet behind independent contractor classification; it faces direct corporate exposure for every hardware and software glitch.

That does not mean Uber automatically becomes responsible for every autonomous-vehicle failure—AV developers such as Waymo can remain responsible for defects in their own technology—but Uber’s expanding role in managing and dispatching autonomous fleets could expose it to forms of direct corporate liability that are less prominent in its traditional marketplace model

Also, federal safety investigators/investigations have historically cited an “inadequate safety culture” when analyzing autonomous networks, citing a human operator’s tendency to lose focus when operating autonomous loops, exposing Uber to damage if public deployments fail to perfectly police human safety-backup alert levels.

There are other regulatory problems here, but I will end it off at this point to not confuse you with endless legality talks.

5: Uber and Marchetti’s Constant

Marchetti’s constant states that humans spend an average of one hour commuting every day. Whether a society relies on walking, horses, trains, or automobiles, human behavior treats travel time as a fixed daily budget.

This law can, interestingly enough, apply itself to Uber. Uber’s entire valuation multiple rests on the tech-bull assumption that it can infinitely expand its Total Addressable Market (TAM) by turning ride-hailing into an omnipresent everyday utility. Marchetti’s Constant proves this assumption is structurally impossible.

6: Where Uber Becomes a Buy

I will not be giving you a price at which to buy Uber. I will, however, be giving you a series of events that must occur for Uber to become worth another look at.

First, Uber must prove that it has driver loyalty. By this, I mean that Uber has to prove that their drivers actually stay and don’t switch-a-roo to the platform offering the best fare. For Uber to get driver loyalty, they could (notice I said “could” and not “will”) shift from a 30% variable commission take-rate to a flat-rate monthly software subscription fee for drivers. This removes the incentive for driver platform-promiscuity and stops the multi-billion-dollar marketing bleed required to replace churning supply.

Next, Uber must cement itself as a player in the AV space, not just as a commodity that “connects by scale”. This one is a more “variable” problem to solve, as there are still many unknowns in the equation; however, this is a must for Uber.

The aforementioned checklist, along with all the items on it, is essential for Uber to truly thrive in the coming years and era of autonomous vehicles.

7: Final Thoughts and Conclusion

Today, we analyzed Uber piece by piece. We broke down the competitive advantage shrink, the financials, the autonomous vehicle problem, the regulatory quagmires, and Marchetti’s constant. Then, we finished it up with the point at which Uber becomes a buy.

My closing thoughts are as such: Uber’s growth is hard-capped because of population density and Marchetti’s constant. The same dynamics that caused Uber’s market share to decrease in Asia could come back to strike them harder in other parts of the world.

Driven by the frenzy surrounding AI, Palantir’s share price has completely detached from the stock’s underlying fundamentals, reaching a sky-high valuation of more than 130x P/E. Combined with other factors discussed below, this suggests Palantir’s current valuation is unsustainable. This also leads me to the conclusion that a multiple compression for Palantir is coming soon.

An important thing to note is that I am not saying Palantir will experience a stock drop in the near future; I am saying that a multiple compression (P/E, P/S, and EV/S ratios will come to more reasonable levels, either through a correction or an acceleration in top-line and bottom-line profits) is mathematically probable given the situation Palantir is in right now.

1: Financials unsustainability

On the surface, Palantir’s financials look great, and for the most part, they are. Q1 2026 earnings came in at $1.633 billion, up 85% year-over-year and 16% sequentially. Palantir also raised guidance for full-year 2026 revenue to $7.650–$7.666 billion, up approximately 71% year-over-year.

However, when we dig deeper into Palantir’s financials, we can see numerous signs that point to Palantir experiencing a multiple compression in the coming years. First, we can see that Palantir is trading at an astronomical 38.4x EV/sales multiple (people are paying $38 for every $1 of revenue Palantir brings in). In comparison, Microsoft, a trillion-dollar company, is trading at 8.6x–9.4x EV/sales.

Additionally, Palantir’s current 1.77 forward PEG ratio relies on an institutional consensus projection of 47.5% annualized EPS growth over five years. However, in Q1 2026, Palantir recorded an adjusted operating margin of 60% and a GAAP net income margin of 53% ($871 million). Margins are already operating at theoretical software ceilings; further bottom-line EPS acceleration cannot mathematically outpace slowing revenue growth over a multi-year horizon.

The image above shows a chart of P/S ratio (price to sales ratio) in Palantir. One thing we can clearly grasp from this graph is that Palantir is being priced for near perfection. Currently, Palantir is being priced at close to 45x revenue/sales, leaving almost no margin for error or operational hiccups.

If the market’s enthusiasm for AI cools down or enterprise CapEx shifts, a standard reversion to a premium software multiple of 30x sales would instantly trigger a 33% drop to $89. Furthermore, if Palantir’s metrics compress down to traditional high-growth tech industry averages of 15x sales, the stock would experience a devastating 66% collapse down to $45—all while meeting management’s exact financial targets. The chart clearly outlines that a long position at current prices is not a bet on Palantir’s business capability, but rather a speculative gamble on the sustainability of an inflated valuation multiple.

Bulls may argue that Palantir’s financials are accelerating (sure, they are); however, they cannot keep accelerating at this rate indefinitely. To quote Ben Graham, “A $1-billion company can double its sales fairly easily; but where can a $50-billion company turn to find another $50 billion in business?”

Similarly, Palantir is growing at an exceptional rate, but sooner rather than later, it will inevitably go through a phase of multiple compression, where P/E and P/S ratios will decline to more reasonable levels.

2: Flaws in the Core Business Model

To understand the flaws in Palantir’s core business model, we first have to understand what Palantir does as a company. Palantir builds software that helps large organizations connect, analyze, and use messy data from many different computer systems. Instead of selling or mining personal information, it sells powerful data-integration and artificial intelligence platforms, mostly to government defense agencies and major corporations

The first big problem worth pointing out in its business model is that Palantir has failed to build a globally diversified business. This is proven by two main points:

First, a whopping 79% of Palantir’s total revenue comes directly from the United States. True enterprise software giants scale to other parts of the world, such as Europe, Latin America, and Asia. Palantir, however, cannot replicate this. Palantir has deep ties with the United States that make relations with other countries a “non-starter”. Due to this problem, Palantir is facing a severe shortage in its TAM (total addressable market).

Second, Palantir still remains largely reliant on public sector defense funds, which largely place a cap on predictable SaaS growth. Palantir deals with what is called a “monopsony”, meaning it is largely reliant on one dominant buyer. If Congress enforces federal defense budget caps, or if the Pentagon shifts its AI strategy toward open-source, multi-vendor software, Palantir faces an immediate revenue cliff that it cannot quickly replace in the private sector.

Along with these two problems, which limit Palantir’s TAM, one other issue also takes precedence in this bear argument. Palantir cannot just sell a license and let an organization or customer use its software. Instead, Palantir uses aggressive deployment of expensive “Forward Deployed Engineers (FEDs)” to configure unique systems for its clients.

This business model undermines the real financial advantage of any traditional SaaS business. To onboard and retain clients, Palantir must continuously recruit and pay for top-tier engineering talent. This creates a human capital bottleneck: customer acquisition costs scale alongside revenue, creating long-term headwinds for operating margin expansion.

Consequently, these deep structural flaws expose Palantir as a high-risk contractor rather than a scalable software giant, leaving its hyper-inflated valuation entirely at the mercy of single-point political and operational failures.

3: Stock-Based Compensation and Dilution Problems

For most stocks, the problem of stock-based compensation (SBC) and dilution does not exist at all. However, in Palantir’s case, SBC is a real and notable issue.

Palantir is diluting its shareholders on an institutional scale. In the full fiscal year 2025, the company issued $684 million in stock-based compensation. This equity-printing machine has accelerated, with trailing twelve-month SBC reaching an astronomical $1.689 billion by early 2026. Every single dollar represents permanent dilution that reduces the equity percentage and future earnings power owned by public retail shareholders.

Management’s ‘Adjusted’ metrics mask the true operational costs of the business. In Q1 2026 alone, Palantir recorded $201.59 million in stock-based compensation, which made up 23% of its entire $878.58 million operating expense budget. By adding back this $201.59 million expense to boast about a record-high 60% adjusted operating margin, Palantir hides the fact that it is substituting billions in standard corporate cash wages with aggressive share printing.

4: Valuation Problems and the S-Curve Trap

The main argument that bulls use to defend Palantir’s current valuation relies largely on the notion that its massive valuation multiples can be justified by its growing revenue trajectory. However, this argument fails to account for fundamental mathematical limits and the inevitability of an aggressive macro deceleration curve.

Palantir trades at an astronomical P/E ratio of more than 130x and an Enterprise-Value-to-Sales multiple exceeding 38x its guided FY2026 revenue. For a company to sustain a 38x sales multiple, it must achieve operating efficiency and scale unmatched by any infrastructure tech company in market history. Even assuming it hits its maximum $7.66 billion FY2026 revenue target, the premium remains totally astray from sector fundamentals.

The S-Curve Illusion illustrates what Palantir is going through right now. Each high-growth software and AI business goes through this “curve”, so to speak. After their explosive early-stage deployment, these businesses hit an S-Curve growth ceiling, where growth eventually normalizes to a more reasonable level.

In Palantir’s case, compounding a multi-billion-dollar business at 70%+ year-over-year becomes mathematically improbable when calculating over a multi-year timeframe. As soon as this localized sprint naturally decelerates toward a standard 20% to 25% maturity rate, the market will aggressively compress its multiple, causing a massive structural re-rating.

5: Where Palantir Becomes a Buy

This thesis is not just a bear argument against Palantir, though, quite frankly, it may appear to be one for the most part. Rather, I am also looking at the point at which Palantir becomes a buy. For Palantir to convincingly become a buy, the following must happen:

First, Palantir must shift away from being largely reliant on the United States for its operations. Additionally, Palantir must find clients other than the United States government. If this happens, then Palantir will prove that it is not just reliant on one dominant buyer to sustain its operations (monopsony).

Second, let’s look at the price-to-earnings ratio. A 138x P/E leaves zero room for operational error. When growth inevitably hits its mathematical S-curve deceleration and settles into a mature 20% to 25% annualized pace, the multiple will compress. Purchasing PLTR at a compressed 40x P/E ensures you are buying a premier defense contractor at standard sector multiples, providing significant downside protection against broader market pullbacks.

And finally, Palantir becomes a viable buy when its forward EV/Sales multiple compresses to the 10x–12x range. Assuming Palantir successfully tracks toward its guided $7.66 billion FY2026 revenue baseline, an adjusted 12x sales multiple translates to a target enterprise value of approximately $92 billion. At this normalized tech valuation, the stock stops trading on speculative AI hype and starts trading on tangible, realizable enterprise software cash flows.

Conclusion:

Throughout the course of this article, we have taken a deep dive into Palantir’s business. We looked at four main problems that can cause a multiple compression in the coming months or years: financial issues, core business flaws, SBC, and valuation concerns. We also examined what it would take for Palantir to eventually become a viable buy (or at least worth consideration). This whole thesis is not meant to tell anyone that a correction is nigh for Palantir; rather, it is informing the reader that a multiple compression is likely to happen given the situation Palantir is in right now.

What I am saying, however, is that the growth that is coming for Tesla in the foreseeable future is likely to already be priced in. At a 357x P/E ratio and a forward P/E ranging from 170-190x, investors are assuming flawless execution with no road bumps across the way. This type of growth is just unrealistic, as no company can provide investors with flawless execution. This leads me to believe that TSLA is trading at an unrealistic price to sustain going into the future. With earnings on the brink of being released as this article is being written, this article serves as a testament to the warning signs that are being seen in TSLA.

Problem 1: The Financials

There is a massive gap between TSLA’s financials and the price of stock that we can clearly see when looking at almost any financial metric. The image above shows us the net income and free cash flow from 2019 to 2025. The gap above is clear: Tesla’s valuation skyrocketed while, during the same time, earning power saw a sharp decline from its highs of 2022 and 2023. Net income and free cash flow are both significantly lower than the highs given in that period.

In addition to this, revenue growth has stayed largely stagnant from 2023 to 2026 sitting at around 95 billion in this time period, which is largely driven by rising competition in the EV space, along with declining demand for Tesla vehicles.

We can see an even clearer picture of this exact problem in the image above. This image shows The GAAP diluted EPS growth (in this case, decline) relative to the share price growth. The disconnect clearly shows through in this image. While Tesla’s share price saw a 53% growth from 2023 to 2026, in the same period, GAAP diluted EPS fell 70% (from 4.30 in 2023 to 1.08 in FY 2025) in the same period.

Each chart may show different numbers, but they all highlight the same story: Tesla, in the last 3 years, has seen absolutely no growth on the fundamental level, yet in the same period, the share price has been driven up to an unrealistic level.

Even as the earnings season arrives, Tesla needs nothing short of a financial miracle to prove to investors that the story is changing. Even a modest beat on revenue, an aggressive new product timeline, or a few hundred basis points of margin improvement would not be enough. Tesla must show that automotive demand is accelerating, margins are recovering, free cash flow is expanding, and its autonomy, robotics, and energy businesses are becoming real profit engines. Anything less would leave the same uncomfortable conclusion: the share price is moving on promises that the financial statements still cannot support.

Problem 2: The Competitive Advantage Shrink

Tesla once had the EV market largely to itself. Its vehicles offered better range, stronger software, faster charging, and a brand that competitors struggled to replicate. That gap has narrowed rapidly. Traditional automakers have improved, Chinese manufacturers have moved faster, and BYD has shown that electric vehicles can be produced at scale across a wider range of prices.

The image above compares Tesla and BYD’s annual battery-electric vehicle deliveries (in millions). Tesla was clearly the dominant supplier up until 2023. But from 2023 till date, BYD has taken the crown and has dominated Tesla by over 600,000 more deliveries in full year 2025. This clearly highlights a weakening competitive advantage on the part of Tesla.

To add on, we can see that Tesla has not launched a new vehicle in over 3 years, choosing to instead rebrand and sell their same cars repeatedly. In full year 2025, almost 97% of Tesla’s automotive sales came from Model 3 and Model Y, showcasing Tesla’s overreliance on their narrow range of products. Meanwhile, Chinese EV manufacturers release new models faster, giving buyers more choices and price range to buy EV’s more comfortably.

Tesla has increasingly relied on price cuts and financing incentives to defend their share and bolster their deliveries. In January 2023, the company cut the base Model Y by $13,000, from $65,990 to $52,990. By 2025, it was introducing stripped-down versions of Model 3 and Model Y priced as much as $5,500 below their predecessors. These moves may have protected volume, but they also suggest Tesla increasingly must compete on price rather than product superiority alone.

Their reliance on price cuts is further shown by the following: Tesla’s deliveries from 2022 to 2025 increased by 25%, while their gross margin declined almost 11 points due to their price cuts affecting Tesla’s profitability.

Tesla’s US electric vehicle share has also seen a significant decline in the past 5 years, from their being the dominant US EV maker, their share of the market has been cut in half, and their share of the EV market in the States has become a mere 45%, compared to their highs of 2021. This is not just a United States problem that Tesla is facing. Tesla’s worldwide deliveries fell 8.6% in 2025 to 1.64 million vehicles, marking a second consecutive annual decline.

China has showcased pressure as well. Tesla’s domestic sales fell 35.8% year over year in October 2025, reaching their lowest level in three years, while its share of China’s EV market dropped to just 3.2%.

Global demand for electric vehicles has not disappeared; in several major markets, it continues to grow. Tesla is simply capturing less of that demand as buyers gain access to cheaper vehicles, newer models and stronger alternatives from BYD, Xiaomi, Volkswagen and other competitors.

Problem 3: Robotaxi expectations

Keeping in mind the valuation that Tesla currently has, it is noticeable that Robotaxi expectations (alongside other expectations that we will discuss later) are already largely priced into the current valuation of Tesla shares. Tesla launched Robotaxi in June 2025 whilst saying that they are expanding their service, but have not disclosed revenue, weekly paid rides, or profit. As of May 2026, Tesla held a California permit for testing with a safety driver but was not listed for driverless testing or deployment. Waymo, on the other hand, reports more than 500,000 fully autonomous electric trips per week, giving it a much deeper operating history.

Scaling into Robotaxi requires much more than just autonomy software; it requires charging, cleaning, repairs, insurance, teleoperations, permits, and fleet uptime. As of right now, Tesla has not reported utilization, profit per ride, or the capital required to support a network of this scale.

Look, I am not telling you that Robotaxi is not possible. I would be a fool if I were to tell you all that. What I am saying is that Robotaxi has many unknowns still in the mix. Additionally, a product of this scale and size takes time to build. It’s time to think hard about the question: How much of Robotaxi’s growth is already reflected in Tesla’s stock price?

Problem 4) Optimus is valuation without revenue

Tesla describes Optimus as a general-purpose autonomous humanoid in development. Yet, we do not have a clue as to how demand for this product will develop alongside the commercialization timeline.

Just like Robotaxi, Tesla has not separately disclosed Optimus revenue, gross profit, or unit economics. Preparing for large scale production is not the same as proving repeatable manufacturing at an attractive cost.

Internal factories and demonstrations can show technical progress, but they cannot facilitate external customer demands. Tesla must prove useful tasks, uptime, efficiency, and economics relative to human labor or competing robots.

As of right now, Tesla has been assigned value before even proving the commercial product. We are giving Tesla credit for something that has not even happened yet, as seen in the 350x P/E ratio and the 150-170x forward P/E ratio.

Problem 5: The Musk Problem

First, let’s get things straight. Elon Musk is an absolute legend. I am not trying to undermine what he has accomplished, which is something no one has come close to. Now, to the point:

Tesla’s own 2025 Form 10-K states that the company is highly dependent on Elon Musk, needing him to facilitate new product launches and bolster existing revenue. The same filing states that Musk does not devote his full attention to Tesla.

Not only does Musk largely run Tesla, but he also holds management positions in SpaceX, xAI, Neuralink, and The Boring Company. In addition to this, Tesla is trying to go through a major transition phase, attempting to bolster their current vehicle growth, but also trying to launch Robotaxi and Optimus.

Tesla’s 10-K warns that product goals could direct resources towards technologies that fail to match market preferences. How can a company successfully go through an unusually broad strategy into several different product lines when its most important executive directs his attention elsewhere?

Tesla’s strategy is broadening while leadership attention is being directed elsewhere.

Problem 6: Earnings create more downside than upside

Tesla reports earnings on Wednesday, or 7/22. Tesla does not need to report disastrous quarterly earnings for the valuation to remain incredible difficult to justify. Even if Tesla manages to report a solid beat in EPS and gives decent guidance for all its coming product lines, these businesses are still years away from being profitable.

A normal earnings beat would show one quarter of good results, while the stock still reflects success in business that hasn’t been proven yet. The most important near-term metrics that investors should keep a watch for are automotive demand, free cash flow, capital expenditures, and the quality of guidance.

For Tesla to meaningfully justify its current and future valuations, all things above must occur. If any of these goes awry, then Tesla will not be able to justify its current or future valuation effectively. With a forward P/E ratio of 160-200x, this shows that investors are already pricing Tesla largely for future growth. For context, this is roughly 14 to 15 times higher than the median P/E of the broader automotive and vehicle industry. This clearly highlights the scale at which Tesla must perform to prove itself and its valuation.

Conclusion:

Over the course of this article, we have gone through six problems that suggest that the future growth expectations for Tesla have largely been priced in: The financials, the competitive advantage shrink, robotaxi expectations, Optimus growth, Elon Musk’s diverted attention, and the earnings problem.

Something I am not saying is that Tesla is out of room to grow; let’s make that clear. What I am saying, and something all current or potential investors in Tesla should think about long and hard, is whether future growth expectations for Tesla is something that is already priced in. If your answer is yes, then you (as a prospecting investor) should refrain from buying into shares of Tesla, as it exhibits many of the signs of a classic value trap. If your answer is no, and you believe that Tesla does have room to meaningfully grow, then you can buy Tesla, while keeping this article in mind.

I hope this article provided value to you in any sort of way. If you would like to showcase your support for this publication, don’t hesitate to subscribe. If you have any questions, please let me know in the comments below. Till next time…