HuaRenCa
Back to Forum
Community

Strongly Bearish on AAOI

郑建滨
郑建滨

1 week ago

AAOI will be worth at most $10 a share in a few years. I have no idea why so many people on X are hyping it up, even adding leverage to buy the dip every time it falls. Anyone thinking of going all in now and waiting for it to double automatically in a few years will lose everything they have.

The story sounds great: 800G optical modules are ramping now, capacity is fully booked, 1.6T will follow, and CPO is coming later. They're building factories in the US, vertical integration means no supply chain risk, they make their own lasers and don't depend on outside suppliers, and how much they can earn now doesn't depend on orders but on how much they can produce.

But all the problems stem from these so-called advantages, and analyzing each one individually reveals a massive pitfall.

The optical module industry is fundamentally a cyclical industry, just like memory. Whatever cycles memory goes through, optical modules will go through too. Manufacturers buy the quantities they need years in advance before building data centers, then install them gradually as the data centers are built out—stockpiling extras and buying more when they run short. The problem now is that manufacturers themselves can't accurately predict how many optical modules they'll need, and capacity is very tight, so at this stage they naturally buy as much as they can, creating the illusion of a long-term bull market with persistent supply shortages.

First, CPO. Anyone who believes CPO will dominate data centers in the future can take a break for now. In ANET's earnings call this quarter, the CEO explicitly said that pluggable transceivers will be mainstream for years to come (I suggest you look up what ANET does—the CEO's view is absolutely worth referencing). They don't see any possibility of CPO being procured at scale in the next few years. This wave of procurement is concentrated on 800G + 1.6T; 3.2T isn't even a hot spot in this round.

Since 800G and 1.6T are the main volume products in this cycle, and from this quarter's earnings reports of COHR, LITE, and AAOI we can conclude that 800G has already peaked, next year 1.6T volume will directly take over, and these optical module manufacturers are all building factories to expand capacity while new players are emerging (CRDO), it's entirely predictable that next year's 1.6T shipments will be 2-3x this year's (a very conservative estimate) or more.

I strongly believe this wave of hyperscaler demand for 3.2T/CPO is not large—just small-scale pilots—so the oversupply of optical modules will definitely peak next year, and the year after that supply and demand will balance or even tip into shortage. These optical module manufacturers will have to absorb price cuts from intensifying competition while also bearing cost increases from capacity constraints. Long term, they cannot become cash flow machines.

Another issue is so-called "vertical integration." This isn't anything new—AAOI has been talking about vertical integration for over a decade. The benefit is obviously a significant reduction in supply chain risk, but the downside is clear: when you make everything in the US, factory and labor costs are high, one module problem halts the entire product line, and when supply exceeds demand, idle capacity means you bear all the costs yourself.

Look at FLNC, a company that assembles energy storage batteries. Because the final assembly factory had problems, the entire business line stalled, and fixing factory problems in the US is extremely troublesome—poor grid infrastructure, supply chains nowhere near as convenient as in China, and skilled workers hard to find. Management promised a year ago that the problem would be resolved quickly, but after all this time it's still the same. Partly it's the CEO's execution problem (the CEO was fired over this), and partly it's the inherent problem of building factories in the US—a problem management simply can't solve easily.

AAOI's new factory in Sugar Land faces exactly all these problems. Months ago, when X was celebrating AAOI building a new factory and significantly expanding capacity, I firmly believed this factory was a hidden risk—their capacity ramp in 2027 would not go smoothly, and even 2028 would have problems. By the time they sort out the factory issues, the optical module cycle will have passed, and the factory's idle capacity will keep draining their cash, causing the company to lose more and more money in the future.

Then there's profitability. AAOI did make money in the latest quarter—Q3 GAAP net margin was about $6.3 million—but the accounts also included $8.16 million in government subsidies, so this company still can't make money even in such a good market!

And margins: last year Q4 was 31.2%, this year Q2 dropped to 27.7%, Q3 only recovered to 28.2%. Revenue doubled, yet margins went down! This shows that running existing factories at full capacity doesn't bring much operating leverage; instead, the vertical integration positioning by management eats up most of their margin. This will remain a major risk even after the new factory is built. By comparison, LITE and COHR have margins of 51.0% and 39.1% respectively.

Now look at historical optical module cycles. In Q3 2016, the new Sugar Land factory went into full production, and data center revenue hit a sixth consecutive quarterly record. On the Q4 earnings call, management said the new equipment and staff would be fully utilized as soon as capacity came online; demand from major customers was very clear; 40G modules would gradually decline and 100G would immediately take over. Sound familiar? Demand was real, AAOI really made money, and the optical module iteration was real—but at the moment when AAOI management was most bullish, the stock had already peaked, then went downhill all the way from around $100 at the top to $2 at the end.

Why? Because later, major customers' 100G purchases fell short of expectations—and it came as a sudden announcement! In fact, AAOI management had already signaled that major customers' 40G purchases were gradually declining and that future 100G purchases would replace 40G as the main profit source. But nobody knew whether major customers would delay purchases because their data centers were mostly built out.

This is exactly the same as the current situation: 800G ramping, 1.6T taking over, and 3.2T and CPO in the future. But nobody knows whether next year hyperscalers will delay 1.6T optical module purchases because they've already stockpiled enough, or because data center construction is paused due to Democratic policies. Once delayed, the blow to optical module manufacturers like AAOI would be unparalleled, because not only would revenue drop directly, but they'd also bear the huge costs of building new factories and idle capacity.

Strongly Bearish on AAOI

6
1.4K

Comments (6)

Your avatar
Sign in to comment
华夏星月
华夏星月1 week ago

Could we reconfirm the timeline for the new factory's production ramp-up? If the 2027 milestone slips, how will the cost pressure change?

海角天涯
海角天涯1 week ago

Government subsidies are definitely something to keep a close eye on. Some companies use one-time subsidies to make their books look good, and once the subsidies disappear next quarter, the problems may show. I'd suggest looking at the specific terms and sustainability of the subsidies in the 10-Q.

曾阿牛太太牛
曾阿牛太太牛1 week ago

The core issue is determining where we are in the cycle, not whether a company is good or bad.

34206
342061 week ago

The most convincing point is that 6.3 million includes 8.16 million in subsidies—with the market this strong and still no real profit, that's genuinely something to watch out for.

Omar221
Omar2211 week ago

I'm not entirely sure about the CPO assessment. ANET's CEO was talking about their own procurement plans, but other hyperscalers like Google and Microsoft may not follow the same roadmap. Also, CPO could be adopted first in specific use cases rather than fully replacing existing solutions across the board.

库页岛巡山人
库页岛巡山人1 week ago

Beyond waiting for earnings reports, what tracking and verification can ordinary investors do right now?