Tesla's most profitable asset is not a Tesla

By Cash Flow University · 2026-07-25 · 8 min read

Unrealized equity gains, not car sales, carried much of this quarter. Only one of these companies paired the marks with real cash.

Tesla's most profitable asset is not a Tesla

Last updated: July 26, 2026

The one question worth asking about any company: is it writing the cheques, or cashing them? I started this quarter the way most people do, reading net income and headline EPS. Then I opened the cash flow statements, which is where we spend our time at Cash Flow University, and three big quarters stopped looking like three big quarters. Strip out one accounting line and only one of these companies is still standing.

The line is identical in all three filings: gains on stock these companies own in other companies. Marked up on paper, not sold. Tesla's operating income was $398 million. Net income was $1.114 billion. $763 million of that gap was an after-tax gain on SpaceX shares. Alphabet reported $9.11 in EPS and $6.26 of it came from equity gains. Nvidia booked $15.9 billion of equity gains against $58.3 billion of net income.

Same season. Same line. One of them did not need it.

Writing the cheques

Health shows up on the cash flow statement, not the net income line. The income statement tells you what the quarter looked like. The cash flow statement tells you what it cost. Open the second one and this stops being an accounting story. It becomes a story about where a company sits.

Net income vs. free cash flow, quickly

If this is new to you: net income is the number on your pay stub, bonus included, some of which has not hit your account yet. Free cash flow is what is left in the bank after rent, groceries and the car payment. You want the companies with money left over. Cash buys options. Everything else is a claim.

Tesla posted a record and a cash drain in the same three months. Revenue rose 26% to $28.2 billion, and gross margin gave up more than four points getting there. What made it to the operating line was $398 million. Meanwhile capital spending jumped 142% to $5.8 billion. Operating cash flow was a healthy $4.7 billion and still came up short. Free cash flow: negative $1.09 billion. Management is not hiding where this goes. Capex is guided above $25 billion this year, with up to $30 billion of new debt flagged.

Alphabet is the harder one, because the business itself is humming. Cloud grew 82% at a 35.6% operating margin. Search, the thing three years of AI commentary said would be gutted, grew 17%. The cloud backlog hit $514 billion. Then capex came in at $44.9 billion, roughly double last year, and free cash flow went to negative $5.9 billion. I own Alphabet. I am not selling it over a quarter like this. Paying cash now for revenue that shows up later is what you do when you think the position is worth holding. I am not calling the spending a mistake. I am pointing out that it is spending, and somebody at the other end is receiving it.

Cashing the cheques

Follow the money out the door and you land on the suppliers. The cash leaving Tesla and Alphabet does not evaporate. Same season, opposite sign, and not because one set of executives is smarter. It is where they sit on the chain.

Count the substitutes

Cash follows pricing power, and pricing power is just the inverse of how many other calls the customer can make. Want an electric car? There are dozens of people to call. Want hyperscale cloud? Three or four. Want AI accelerators? Nvidia, AMD, and a growing bench of custom silicon. Leading-edge foundry at AI volume is one supplier at scale. EUV lithography is one.

TSMC gives you a clean way to measure this. For years the accepted wisdom was that making chips in Arizona costs far more than making them in Taiwan. TechInsights built a cost model and put the gap under 10%. Equipment is more than two thirds of wafer cost, and a machine costs the same in Phoenix as it does in Hsinchu. Labour, where US wages run triple Taiwan's, is under 2% of the total. TSMC charges a 20–30% premium on US-made N4 and N5 wafers anyway. Customers pay it. Arizona is sold out into late 2027.

What settled it for me was what TSMC did next. Asked about memory makers earning 86% gross margins against his own 67.7%, C.C. Wei said he was jealous, then said TSMC would not suddenly raise wafer prices because it is a trustable company with its customers and earns enough to keep expanding. That is not modesty. Restraint is only available to someone who has the option.

The loop

TSMC's customers cannot walk. Hold that thought and look at who else is stuck. In June, Alphabet agreed to pay SpaceX $920 million a month for AI compute. That payment is part of the capex that pushed Alphabet's free cash flow to negative $5.9 billion. In the same quarter, Alphabet disclosed it holds $94.1 billion of SpaceX stock, and gains across its equity portfolio delivered $6.26 of its $9.11 in EPS.

Turning this into trades

None of this is academic. Knowing who is spending cash and who is collecting it changes how you structure a trade. CBOE research puts roughly 80% of options expiring worthless, which is the whole reason we sell premium rather than buy it, and we sell it on companies we would be happy to own.

Sell premium on the cheque cashers

Nvidia and TSMC are cash machines with real pricing power.
What to do with that: wait for a pullback and sell a cash-secured put. You get paid to agree to buy lower. Stays above your strike, you keep the premium. Drops through it, you own a good company at a discount. Covered calls do the same job from the other side if you already hold shares.

Use spreads where the volatility is

Tesla and Alphabet are spending hard on growth, and that spending brings volatility.
What to do with that: stop paying up for outright options. A bull call spread caps your risk and, with implied volatility elevated in these names, the short leg is worth enough to cut your cost meaningfully.

"Be the house. Sell premium on the companies that are printing money and use risk-defined strategies on the ones that are spending it. That is how you turn financial statements into cash flow in your own account." Stephen Manning, Founder, Cash Flow University

Where to push back on this

A toll booth is not immune to the cycle. It is levered to it. TSMC and ASML sell into somebody else's capital budget. Shrink the budget and the order book shrinks with it. There is no world where the hyperscalers pull back and the suppliers hold their numbers. The near-term read is already mixed: TSMC guided next quarter's gross margin lower as 2nm ramps, and ASML carries roughly 20% China exposure and no longer publishes quarterly bookings, so you are inferring demand from commentary. Position on the chain is not permanent either. Those custom silicon programmes inside Nvidia's biggest customers exist for exactly one reason: to turn Nvidia into a payer. Nvidia reports again in late August. That is the print to read closely.

Strip the marks

Back to the opening question. Take the equity gains out of all three quarters and see what is left.

The marks were everywhere this season. One company had operations underneath them.

Is it writing the cheque, or cashing it? And when its customer finally decides to walk, where do they go?

I do not know when this capex cycle turns, and neither does anyone else sending you a newsletter. What I have decided is that the timing is not the part I need to get right. When it turns, nobody will care who told the best AI story. The question will be who was getting paid the whole way through.

What to do this week

1. Sort your holdings: cheque writers or cheque cashers?
2. Find the entries: take the strongest cheque cashers and wait for a pullback to sell premium. Tastytrade looked at 52,000+ trades and found that closing winners at 50% of max profit lifts your probability of success and frees the capital up sooner.
3. Build the watchlist: the key suppliers, their OEM customers, earnings dates, and where volatility sits. That is your entry and exit calendar.

Stephen Manning

Founder, Cash Flow University

← Back to Blog