Alibaba at $116 When My Worst Case Says $150: The Full Bull, Bear, and Eight-Pillar Breakdown

TL;DR My assumptions put Alibaba's intrinsic value at $150 conservative, $290 base, and $545 optimistic. The stock trades at $116 — below even my most pessimistic case. The bull case rests on AI and cloud demand plus a 1.7x price-to-book. The bear case rests on unresolved legal exposure, restricted access to Nvidia's advanced chips, and aggressive capex.

Why this name, specifically

Ten stocks on my watchlist clear a 15% expected return. Alibaba sits mid-pack at 20%. But that ranking isn't why I'm pulling it apart here.

I picked Alibaba because the bull and bear cases are unusually well-matched. One side calls it the central pillar of China's AI infrastructure. The other calls it a collection of regulatory and geopolitical risks wearing a ticker. With names like that, you look at the numbers first and attach the story afterward — not the reverse.

The bull case, in three parts

First, the AI story is real. Alibaba has been building its own large language models. They've secured a partnership to serve as the AI backbone for Apple devices in China. And their cloud business is seeing a surge in enterprise demand for computing power. The key isn't any one of those — it's that all three feed the same engine.

Second, the valuation is plainly cheap. Price-to-book sits around 1.7x. Comparable tech companies trade at 4.5x to 7x. Most major analyst price targets sit well above where the stock trades today. The market is not giving this company credit for what it's currently doing.

Third, the risk is actually coming down. They settled with the Department of Justice for $600 million. That was a legal cloud sitting over the stock for years, and it's gone. When legal and regulatory overhang starts clearing on a stock that's already cheap, things get interesting.

The bear case, in three parts

If you don't hear the other side before you buy, you're not analyzing — you're cheerleading.

First, the legal issues aren't over. Law firms are running investigations into potential class actions over Alibaba's AI data practices. The European Union is examining possible fines. And even with the DOJ settlement closed, the Pentagon blacklisting and US-China tensions are very much alive.

Second, the AI hardware problem. China may only get limited access to Nvidia's advanced chips. That matters because if Alibaba can't secure the compute it needs to scale its AI models, the growth engine I just described starts decelerating. On top of that, they're restricting which AI tools their own employees can use internally. That is not a great sign.

Third, the money. Alibaba is spending aggressively on cloud and AI infrastructure right now. If revenue doesn't grow fast enough to justify that spend, margins get crushed. Goldman Sachs has already dialed back its conviction on the name. The smart money is turning cautious.

Opening up the numbers

Market cap is $280 billion; enterprise value is $344 billion. That $64 billion gap is net debt. Not a trivial figure.

Here's where the capex discussion shows up in the financials. Free cash flow went negative last year. But even including that negative year, the five-year average free cash flow is still $12.75 billion. And net income came in at $15.65 billion — above its own five-year average.

So I do here what I do with the other large-cap tech companies in a heavy spending cycle: I anchor on net income rather than free cash flow, because the cash flow damage is very likely a near-term artifact of the build-out.

Margins are starting to inch back up. The ten-year average net margin is 14%; the five-year average is 9.6%. Notice the ordering. A five-year average below the ten-year average means the recent half-decade was this company's slump. At one point it was running close to a 20% net margin.

That's where the math gets interesting. If margins recover toward their old level and profit doubles, the current 18x P/E becomes 9x — without the share price moving a single dollar.

Returns on capital are poor. I'm not going to defend that one. Depressed operating income did real damage there.

Revenue growth, however, is not what a dying company looks like: 5.6% annually over the last three years, 7.4% over five, 26% over ten. And critically, there have been essentially no acquisitions. That's organic growth, not growth purchased and bolted on.

Finally, a company with 40% gross margins is trading at 1.85x sales. That's close to a car-manufacturer multiple attached to a tech business with 40% margins.

The eight pillars: five checks, three X's

Not pretty, not awful. The X's come from declining net income, declining free cash flow, and low returns on capital.

But there's one thing here I genuinely like: the buyback timing.

This company did not buy back stock at the peak. They started buying after the stock fell hard. Most companies do exactly the opposite — they repurchase at the top when cash is abundant, then conserve cash after the price collapses. Buying into weakness is what management does when they actually believe their own stock is cheap. That's not an accounting line item; that's evidence about how this management team allocates capital.

Analyst estimates and my own assumptions

Analysts have earnings per share more than doubling over the next four years. That's a number that falls out naturally if net margins simply return to where they were five years ago. Their revenue growth path runs 3%, 10%, 12%, 12%, 5.5%, 11%, 12% — taking revenue from $150 billion to $312 billion over seven years, roughly 10% annually.

Here's what I plugged in.

Input	Conservative	Base	Optimistic
Revenue growth (annual)	6%	10%	14%
Net profit margin	13%	16%	19%
Exit P/E or P/FCF (year 10)	14x	18x	22x
Discount rate for intrinsic value	9%	9%	9%
Resulting per-share value	$150	$290	$545

Two things worth flagging.

First, even my optimistic 19% net margin is below what this company actually delivered five years ago. I did not use peak historical performance as my best case.

Second, I don't think an 18x exit multiple is egregious. The long-run market average multiple is 15-16x. The largest company in one of the fastest-growing economies on earth deserves some premium in my view. Returns on capital are low today, but I've assumed they recover as the AI business matures.

And the 9% discount rate is not my required return — it's the rate I use to find intrinsic value. That calculation answers what the company is worth, not what I should pay. If you want a larger margin of safety, you plug in your own desired return instead.

Where this calculation could be wrong

Before concluding, let me attack my own model.

Is 9% the right discount rate for Chinese risk? I applied the same rate I'd use for a US business to a company operating under structurally different political and regulatory conditions. If you think that deserves an added risk premium, that's a fair objection. Push the discount rate to just 11% and the base case value drops noticeably.

The margin recovery assumes a pre-regulation world. The era when this company earned close to 20% margins predates the platform crackdown. There's no guarantee that margin structure returns. That's why I used a 13-19% range rather than the historical peak — but even 13% could turn out to be optimistic.

An 18x exit multiple ignores the China discount. For several years now, Chinese large-cap tech has carried a structural valuation discount largely independent of fundamentals. Assuming that discount has disappeared ten years from now is a bet, not a calculation.

I still put this name in my top tier, and the reason is narrow: even after granting all three objections, the conservative case of $150 sits above the $116 price. That's precisely what margin of safety means — a price at which I can be wrong in several places and still not lose money.

For more context, my earlier work on Alibaba's AI and cloud business and on contrarian value names more broadly covers adjacent ground.

One last repetition: nothing here is a recommendation to buy. These are my assumptions, and you have no obligation to share them. Finding the specific input you disagree with and replacing it with your own number is the entire point of running the model.