Alibaba's 64% Drawdown: The Spending Cycle That Makes It Cheap

Alibaba has fallen from its 2020 peak of roughly $319 a share to around $114 today — a 64% decline that has left the stock at a forward P/E of 11.8x, a multiple you'd expect from a mid-tier industrial, not a tech giant reinventing itself.

The discount is real. But it doesn't exist because investors have forgotten what Alibaba does.

It exists because the company is spending money at a scale that changes the math.

Alibaba announced plans to invest over $55 billion in AI data centers and cloud infrastructure over the next three fiscal years. That's more than the company spent on capital expenditure across the previous decade combined. The effect on the financials has already arrived. Operating margins have collapsed from 14% a year ago to 4.9% in the most recent quarter. Free cash flow, once a strength, is now negative $6.9 billion on a trailing-twelve-month basis. Return on invested capital has fallen from 8% to 1.5%. The company has missed earnings estimates for four consecutive quarters, with the fifth miss arriving on August 20.

You can call this undervaluation if you believe the spending will pay off. You can call it a justified discount if you don't. The numbers themselves force the question, and the answer determines whether this stock is a patient investor's opportunity or a structural trap dressed up as a bargain.

Here's what's happening inside those numbers. Alibaba is running two companies simultaneously. The e-commerce core — Taobao and Tmall in China — is growing at roughly 6-8% and remains the primary profit engine, generating the bulk of the cash flow. It's a mature, highly competitive business defending against PDD, JD.com, and live-streaming commerce, but it's not dying. The second company is Alibaba Cloud, which is growing at 45% year-over-year, the fastest pace in 22 quarters. AI-related products now account for 35% of cloud revenue and have delivered triple-digit growth for eleven consecutive quarters. Management targets AI revenue crossing 50% of cloud within a year and annualized AI recurring revenue of RMB 30 billion (roughly $4 billion) by year-end.

The growth story in cloud is credible. But the path from cloud growth to profitability runs through an enormous capital expenditure bridge. Capital spending has risen from $5 billion annually in 2023 to nearly $18.4 billion on a trailing basis today. The $55 billion three-year plan accelerates that trajectory. Alibaba expects the spending to support proprietary GPU chips called Zhenwu PPUs, more than 100,000 of which are already deployed on its cloud platform, plus modular data centers designed to scale AI capacity in 100 days at 10% lower cost than traditional builds.

This is where the investment case crystallizes into a specific question: can Alibaba turn $55 billion of capex into returns that exceed the cost of capital at scale? Right now, the answer is no — return on invested capital sits at 1.5%. But capex spending precedes revenue by 12 to 24 months, and the company is still in the build phase. The margin question won't resolve until the infrastructure is deployed and the AI services begin generating high-margin recurring revenue.

There's a structural advantage working in Alibaba's favor that doesn't show up on an earnings surprise table. The company is building its own AI chip supply chain through T-Head Semiconductor, with Zhenwu processors that are now running more than 60% of external compute workloads. If you're investing in Chinese AI infrastructure, this independence from Nvidia — and from U.S. export controls — is a real competitive moat, not a marketing claim. The company has also developed the Qwen AI model family, which it's integrating directly into its e-commerce platform and offering to enterprise customers, creating a full-stack capability from chip to application.

The bear case isn't about whether Alibaba is building something valuable. It's about the capital intensity of the build and the competitive pressure on the other side of it. Tencent Cloud, Huawei, Baidu, and even Amazon Web Services and Azure are all racing to embed AI into their service stacks. The margin environment in cloud AI is still forming, and Alibaba is choosing to invest aggressively now rather than wait for pricing clarity. That discipline — or recklessness — is what separates a contrarian buy from a value trap.

Let's also address what the valuation is actually telling us. A forward P/E of 11.8x is cheap by any measure. But forward earnings estimates have been revised down 45% over the past year, and they're built on assumptions about when this capex cycle ends and margins recover. The trailing P/E sits at 26x, which is not cheap. The disconnect between trailing and forward multiples tells you exactly what the market is doing: it's pricing in a multi-year earnings trough caused by the spending cycle, then assuming recovery. If recovery takes longer than the market expects, or if peak margins never return to pre-spending levels, that forward P/E is accurate, not pessimistic.

The recent earnings miss on August 20 is worth a closer look because it contains both the bull and bear narratives in the same report. Cloud revenue jumped 45%, a 22-quarter high, with AI now representing 35% of cloud revenue. MaaS annualized recurring revenue exceeded RMB 16 billion, putting the company on pace to hit its RMB 30 billion target. Revenue grew 9% year-over-year at RMB 269 billion. That's the growth engine intact. On the other side, adjusted EPS missed again, operating margins compressed further, and the capex trajectory continued upward without a visible inflection point. The stock fell roughly 4.5% on the report.

Neither the bull nor the bear can claim this quarter proved their case. The growth is real but unprofitable. The profitability is depressed but for a disclosed, intentional reason. The question isn't hidden in the numbers — it's right there in the margin trend.

So where does this leave the investment case? Not at a simple buy or sell, but at a specific condition.

The forward P/E of 11.8x makes Alibaba cheap only if the $55 billion capex cycle generates returns that justify the scale. That's a management execution question, not a sentiment question. The company has the technical stack — chips, models, cloud platform, and e-commerce distribution to monetize at scale. It also has a $55 billion spending commitment that will suppress margins and free cash flow for the rest of this decade, at minimum.

What changes the calculus isn't another cloud growth quarter — we already know it's growing. What matters is whether adjusted operating margins stabilize, whether free cash flow turns positive before the spending cycle peaks, and whether AI services begin generating pricing power rather than requiring subsidy to grow. The next two earnings reports will start to reveal the answer, because by then the modular data center builds and Zhenwu chip deployments should be producing revenue, not just consuming capital.

For investors, the position sizing question comes down to how you weight execution risk against valuation discount. A 64% drawdown with a forward P/E of 11.8x is the kind of setup that looks attractive in hindsight — but only if the business emerges from this spending cycle stronger than it entered. The market has arguably priced in the trough. What it hasn't priced in is whether the recovery is real or whether this is simply a more capital-intensive version of the same business, growing faster but earning less.

That's the gap you need to decide about. The numbers tell you both sides are possible.