Read a full thesis without an account.

“To the shareholders of Berkshire Hathaway: I've spent the weekend looking at SpaceX. The 10-K runs 137 pages of mission statements about the "light of consciousness" and artist renderings of Martian colonies. I prefer balance sheets. Start with the first question I always ask: **Is this a wonderful business?** No. Not yet. Possibly not ever — and that's the problem. A wonderful business produces owner earnings that grow predictably over time at high returns on tangible capital. SpaceX gives us only three years of financial history, and that history shows owner earnings of -$4,628 million in 2023, +$791 million in 2024, and -$4,937 million in 2025 [dp]. That is not a pattern — that's a scatterplot. A business that loses nearly $5 billion, earns $791 million, then loses nearly $5 billion again is not one whose economics I can forecast with any confidence. As I said in 1997, "If you think you know what the price of a stock should be today, but you don't think you have any idea what the stream of cash will be over the next twenty years, you've got cognitive dissonance." (WB 1997 Berkshire meeting) **The moat question.** SpaceX launches more than 80% of the world's mass to orbit [10K-BIZ-2026-p1], operates ~9,600 Starlink satellites across 164 countries, and has vertically integrated manufacturing. That sounds like a moat. But the financials don't back it up. Operating margins swung from -33.7% in 2023 to +3.3% in 2024 to -13.9% in 2025 [dp]. A genuine moat business shows margin stability or expansion through cycles. These numbers suggest a business where the economics are overwhelmed by something else — and that something else is capital spending. In 2025, SpaceX spent $20.7 billion on capex against $18.7 billion in revenue [dp]. That's 111% of revenue plowed back into fixed assets. Free cash flow was -$13,952 million [dp]. Charlie Munger and I have a simple rule about this, and I said it plainly in 2012: "Cash-consuming businesses, by their nature, are unattractive unless the cash they consume gets to earn a reasonable return." (WB 2012 Berkshire meeting) We have no ROIC data to test whether that $20.7 billion will earn anything at all. **The pricing-power test.** I look for businesses that can raise prices above inflation because the customer has no alternative. With only three years of margin data — and margins that are deeply negative in two of them — I cannot run this test. A business with real pricing power doesn't swing from -34% to +3% to -14% operating margins. The 2024 blip of profitability may reflect something real, but one good year surrounded by two terrible ones tells me the economics are not yet stable. Now, it's fair to ask: isn't SpaceX investing for growth? Of course it is. But that's exactly the problem. In 2015, I said, "The best businesses during inflation are usually the businesses you buy once and then you don't have to keep making capital investments subsequently." (WB 2015 Berkshire meeting) SpaceX is the opposite. Starship development, V3 satellites, orbital AI compute — every strategic ambition listed in the risk factors demands billions more in capex [10K-RISK-2026-p1]. The company explicitly warns that "failure or delay in the development of Starship at scale" could materially harm the business. That's not a moat talking. That's a business betting the ranch on unproven technology. **The xAI acquisition.** In early 2026, SpaceX acquired xAI [10K-BIZ-2026-p1]. The 10-K frames this as an "integral pillar" of the company. I see a space launch and satellite internet business absorbing an AI startup with no disclosed purchase price, no disclosed economics, and a product — Grok — whose competitive position against the half-dozen well-funded frontier model builders is entirely unclear. My rule on acquisitions is simple: I want to see the price, the funding, and the earnings power. I see none of these. This is a conglomerate being assembled in real time, and the historical record on conglomerates assembled by charismatic founders with grandiose mission statements is not good. **Management and capital allocation.** I have no credibility report to work from [cred], so I will stick to what the numbers show. The company carries zero debt and holds $24.7 billion in cash [dp]. That's admirable — it means the equity holders are funding the enterprise, not lenders who can force unpleasant outcomes. But the 2025 share count of 5.7 billion shares [dp] suggests substantial dilution to fund operations over time. Without a history of buybacks or dividend discipline to examine, I can't assess whether management treats shareholders as partners or as a financing source. **Price.** The current market price is reported at $0.00 with a $0 market cap. I assume this means the stock is not yet publicly trading or the data has not populated. Either way, without a market price and without reliable owner-earnings projections, I cannot calculate a margin of safety. As I've said many times, risk is not volatility — risk is not knowing what you're doing. And with three years of financials, massive negative free cash flow, a transformative acquisition with no disclosed terms, and a mission statement that reads more like science fiction than a business plan, I do not know what I'm doing here. **The tiebreaker.** The macro backdrop is irrelevant for this one. Even if 10-year Treasuries were at 2% and the CAPE were at 15, I would not buy a business that consumed $20.7 billion in capex to generate -$4.9 billion in owner earnings. This isn't a valuation call. It's a business-quality call. Charlie and I try to distinguish between businesses where you have to be smart once and businesses where you have to stay smart (WB 1995). SpaceX requires staying smart — continuously, across rockets, satellites, AI, and Mars colonization, against well-funded competitors and unforgiving physics. There are too many unknowns stacked on unknowns. The business may one day be wonderful. But today, on the evidence I have, I wouldn't touch it.”

“## What I agree with The thesis correctly identifies the central truth: this is a genuine business with a broken stock price. The numbers are clear—operating margins have marched from –107% to +31.6%, and free cash flow hit $2.1 billion in FY2025. The DCF base case of $22.19 per share, against a $138 quote, yields a margin of safety of –522%, and the reverse-DCF model clips at +50%, meaning no reasonable assumptions can justify the current price. The treatment of stock-based compensation is blunt and honest: SBC has averaged 142% of reported free cash flow, and the share count has swollen by 135.8% over the period. Only in FY2025 did true FCF after SBC reach $1,417 million—a number that, for the first time, actually belongs to owners. The balance sheet fortress ($2.3 billion in cash, zero debt) is an unquestionable strength. The thesis gets all this right. ## What I’d push back on The moat discussion is heavy on narrative and light on quantitative proof. The claim that contract stickiness from the Ontology and government incumbency creates durable competitive advantage is plausible, but there is no historical ROIC consistency to support it—the data is missing, and the current-year ROIC of roughly 22% is a single observation from a boom year. A real moat should show margin durability through cycles, yet Palantir’s quarterly revenue just dropped 47% sequentially (Q3 2025 to Q1 2026) and Q1 2026 YoY revenue fell –19.9%, a 52.5-point gap to the 5-year CAGR. That volatility doesn’t look like a business with predictable, fortress-like repeat revenue; it looks like lumpy, concentrated contract timing. The working-capital deterioration—accounts receivable days doubling from 45 to 85, while accounts payable days collapsed from 81 to 4—contradicts the “indispensable partner” story. If customers can’t live without you, they pay faster, not slower. These signals suggest the moat is narrower than the thesis implies and hasn’t been tested in a downturn. Management credibility should be a C, not a B. The $1.78 billion cash evaporation in FY2023, against $697 million of reported FCF, went entirely unexplained in the MD&A. That is a significant withdrawal. The earlier prominent use of “adjusted income from operations excluding SBC”—while the share count was ballooning—was misleading, and dropping the metric without acknowledgment reads as a quiet retreat, not a correction. A management team that dilutes owners by 136% and never repurchases a single share doesn’t merit a B. The operational underpromise-and-overdeliver pattern is a plus, but capital allocation and transparency are at least a letter grade lower. I’d reverse my view if the business demonstrates sustained per-share free cash flow growth (after full SBC subtraction) for four straight quarters, and accounts receivable days stabilize below 60, indicating that the cash conversion is real and the moat is translating into better payment terms—not just headline growth. ## My bottom line PASS. Palantir is an excellent, capital-light software business with real operating leverage and an AI tailwind, but at $138 the stock is a speculation, not an investment. The per-share economics have been so thoroughly diluted that even a 30% margin doesn’t flow to shareholders in proportion; I’d only consider it if the price fell to a level where the reverse-DCF implied growth drops below 15%—which would be a fraction of today’s quote. For now, it stays on the watchlist.”

“## What I agree with The thesis gets the central fact of the case exactly right: the numbers have turned decisively against this business. Operating margin has contracted from 16.8% to 4.6% over three years, and ROIC has collapsed from 36.6% to 4.8% — now barely above the 4.57% risk‑free rate. When a capital‑intensive manufacturer begins earning returns that fail to clear its cost of capital, the economic engine is broken. The thesis correctly identifies that the FCF surge in 2025 ($6,220M) was a mirage — achieved by slashing capex 24.8% — not a sign of improving quality. I also agree that stock‑based compensation, which has averaged 50.2% of reported FCF, makes the headline free cash flow deeply misleading. True owner earnings after dilution are a fraction of the reported figure. The management credibility grade of D is well‑supported by the breakdown of 7 broken promises versus 7 kept, with the kept ones clustered around product launches and the broken ones squarely in the core metrics of profitability and volume growth. The pattern of setting aggressive near‑term targets — Austin safety‑driver removal, Cybercab production start — missing them without acknowledgment, and pivoting to new visions, is exactly the kind of evasive habit that destroys shareholder trust. The thesis’s final warning that the $23.7 billion CEO interim award will inflict a sudden earnings charge when the performance condition becomes probable is an important and under‑appreciated risk. ## What I’d push back on The thesis declares a verdict of “Pass” with moderate conviction and a quality score of 5/10. Given its own analysis, that conclusion is too generous. The intrinsic value range is $46‑$60, and the stock is at $429. Even the most optimistic sensitivity — 10% growth and a 4‑point margin lift — yields only $53 per share. The margin of safety is –670%. A business whose returns on tangible capital have fallen below the risk‑free rate, whose revenue is shrinking, and whose management systematically over‑promises and dilutes shareholders, is not a “5/10” quality business in my book. It is a sub‑par business with a speculative stock price. The gap between price and value here is so vast that no rational investor would consider it “on the watchlist” at current levels. If the thesis truly believes the business is deteriorating and the moat is eroding, the only logical verdict is “Avoid.” The “Pass” may be unintentional charity toward the narrative of optionality in AI and robotics, but the thesis itself correctly labels those as outside the circle of competence and unsupported by disclosed revenue. When you have a business whose current economics are weak and whose forward promises are unverifiable, at a price that implies 50%+ growth, you don’t pass—you walk away. Additionally, the thesis frames the moat as having “existed for roughly two years.” I’d argue that even that is misleading. The two years of high ROIC (2021 and 2022) coincided with a unique period: pandemic‑era supply constraints limited competition, and Tesla enjoyed a pricing umbrella that inflated margins. The median 10‑year ROIC is –1.84%. A two‑year spike amidst a decade of capital‑destroying returns is not a moat; it’s a temporary windfall. The business never developed durable pricing power or competitive insulation. The so‑called “moat erosion” is really just the unwinding of a favorable but unsustainable market anomaly. I’d reverse my view if Tesla achieves a trailing twelve‑month operating margin above 10% and an ROIC above 15% over two consecutive fiscal years, with no decline in unit deliveries, and with management explicitly tying forward guidance to audited automotive profitability rather than to AI or robotics narratives. That would be the first evidence of a genuine, defensible franchise. ## My bottom line AVOID. The current price embeds a fantasy of uninterrupted growth that the financial statements actively contradict. The core automotive operation is generating returns below its cost of capital, management has a track record of broken promises and shareholder dilution, and the balance‑sheet strength is being squandered on massive equity awards rather than returned to owners. At a price‑to‑owner‑earnings multiple that is off the scale relative to the business’s deteriorating fundamentals, this is not a stock I’d buy at any price near $429.”

“## What I agree with The thesis correctly dismantles the valuation. A base-case intrinsic value of $96.83 against a $214.22 price, a reverse-DCF implying 36.5% sustained growth, and a margin of safety that’s deeply negative – the math is unequivocal. The earnings-quality flags are well-supported: the Sloan accruals ratio at +10.9%, four consecutive quarters of net income outpacing operating cash flow (cumulative $34.1B gap), and the cash conversion cycle stretching to 133 days. The single-metric revenue compensation plan is a genuine red flag, as is the $17.5B venture splurge with disclosed circular-financing risk – these destroy the capital-allocation discipline that compounding requires. The management credibility grading is fair: product execution is superb, but demand-forecasting optimism has produced two multibillion-dollar write-downs, and that pattern matters at a $5 trillion price tag. ## What I’d push back on The thesis underweights the actual durability of the moat through a cycle. FY2023 is cited as proof of fragility – ROIC fell to 11.2%, operating margin to 15.7% – but those figures still exceed any reasonable cost of capital and reflect an inventory-clearing event on essentially flat revenue. A business that never loses money in its worst recent year, while maintaining a positive return on tangible capital, has a real competitive advantage, not a temporary demand lift. The same CUDA lock-in and systems integration that drove 60%+ margins at the peak also kept the bottom from falling out. The stress-test scenarios (2008-replay IV of $30) imply the kind of bankruptcy risk that doesn’t fit a net-cash, fabless firm whose chips are essential to modern computing. I’m not dismissing the price risk – at $214 the downside is enormous – but I’d frame the problem differently: the business is a potential long-term compounder, and the valuation is its only fatal flaw. The thesis’s conclusion that “extraordinary business at an impossible price” is right, but the pushback sections sometimes blur the distinction between “this stock is dangerously overpriced” and “this business is too cyclical to own.” At a sufficiently low price, I’d be a buyer. I’d reverse my view on the business’s competitive durability if operating margin drops below 30% for two consecutive quarters, excluding one-time regulatory or inventory charges, indicating the pricing-power moat has been structurally breached.”

“## What I agree with The thesis correctly identifies that Uber has engineered a genuine operational transformation. The numbers leave no doubt: operating margin climbed from -66.1% in 2019 to +10.7% in 2025, and free cash flow swung from negative territory to $9,763 million. That’s not accounting gimmickry—it’s a business that finally found its leverage point. The ROIC trajectory to 14.0% reinforces the inflection. The thesis also gets the narrow-moat assessment right: the 10-K candidly admits low switching costs and the company’s own history of cutting fares to stay relevant, which is the opposite of pricing power. I applaud the brutal honesty of the EPV calculation. At $13.15 per share with zero growth, it exposes that $62 of the current $75 price is a bet on future expansion—a fact most investors would rather not see. The callout on SBC dilution is equally sharp: averaging 141% of reported FCF over the period and a cumulative 20.7% share-count increase is the single most important drag on per-share value, and management’s refusal to discuss it in the MD&A is a credibility stain. Finally, the macro backdrop—CAPE at 41.7, 10-year at 4.47%—reminds us that this is not an environment to pay generously for growth stories, no matter how improved the operations. ## What I’d push back on The base-case intrinsic value of $118.11 and the resulting 36.4% margin of safety look too optimistic. That DCF assumes roughly 12% revenue growth for five years settling into 4% terminal growth, with margins expanding to ~13%. But revenue is already decelerating: the 3-year CAGR is 17.7%, and the last quarterly year-over-year gain was 14.5%. And the terminal value rests on the assumption that Uber survives autonomous disruption and regulatory headwinds for decades. The EPV tells you that if any of that goes wrong, the floor is $13.15. The market’s implied growth of +4.0% from the reverse-DCF might not be undervaluation—it might be a rational appraisal that the future faces real hazards. I’d also downgrade management credibility from B to B- or C+. The systematic omission of SBC dilution from the narrative, burying a $9.1 billion net loss in 2022 behind cash-flow positivity, and the three-CFOs-in-three-years pattern signal a team I can’t trust to execute a $27 billion buyback with shareholder discipline. That buyback authorization is potential, not proof; the share count grew only 0.6% in FY2025, but we don’t know how much was genuinely retired versus offsetting new grants. The margin of safety disappears quickly if buybacks don’t materialize or if a recession cuts FCF, as the stress tests show. I’d reverse my view if the share count shrinks by at least 2% annually for two consecutive years and ROIC stays above 15% for three years, demonstrating that per-share value is actually compounding and the moat has widened to durable levels. ## My bottom line WATCHLIST. Uber has crossed the chasm from cash-burner to cash-generator, and I admire the liability-management overhaul. But at $75, almost all the value rests on growth that I can’t predict with confidence given a shallow moat, a management team that won’t talk straight about dilution, and a frothy market. I’d be a buyer if the shares fell to the mid-$50s—roughly a 25% discount to the bear-case IV adjusted for sustained SBC—or if concrete buyback execution and a ROIC north of 15% prove the business can compound per-share value through a cycle. Until then, it’s a promising story, not a purchase.”