We did that here, but below is a summary:

• Applied Opto loses some $45M in revenues from Amazon in Q3, or half its prospective total revenues for the quarter.
• Yet, it still manages to grow revenues at 25%+ with respect to last year's third quarter, and the shares are cheap.
• That is, not only has it fully replaced the lost revenues from Amazon, it has added another 25%.
• The Amazon loss is transitory, and even if it's not, revenue from it is just $8M-9M now.
• Even if it lost all of that, which is highly unlikely, it wouldn't mean much. More likely Amazon demand will come back.

Applied Optoelectronics: Some Stylized Facts To Keep Investors Grounded - Applied Optoelectronics, Inc. (NASDAQ:AAOI) | Seeking Alpha

Here is some stuff from Barrons:

Leading off the call, CFO Steffan Murray described the shortfall, and without naming names, seemed to indicate it was Amazon (AMZN), which has in recent quarters been Applied's biggest customer for fiber-optic transceivers.

"We indicated last quarter that we expected to see softer 40G demand," said Murray, "however, we saw a lower demand overall from one of our large customers."

He said the cut in spending was a result of a shift from transceivers for "40G," operating at 40 gigabits per second, to the newer parts, "100G":

Revenue from this customer in the quarter was approximately 10% of total revenue compared with 47% last quarter. As a reminder, we have a vendor-owned inventory management model that we employ to this customer, which can impact our revenue visibility. As previously discussed, this VOI program allows the customer to pull inventory from a hub that AAOI manages and revenue is recorded at the time the inventory is pulled. We continue to have ongoing discussions with this customer, and based on our conversations, we believe that the disruption in order flow is related to the ongoing transition from 40G to 100G and not specific to AAOI. We also do not expect an inventory stocked in our VOI hub to be impaired because forecasts indicate that this inventory will be consumed over time.

Murray said demand is still "good" from the company's "other top data center customers," with "particular strength in our CWDM full MSA spec 100G transceivers."

• We have a company who lost most of its business of its largest customer, yet still managed to grow overall revenues by 25%+
• Even if it lost all business of that customer it would do little additional harm
• There is little to suggest this is a permanent loss, AAOI's products keep getting market validation and management argues it's a transitionary problem with Amazon due to technology migration
• The shares are fairly cheap.

Apparently Applied Opto isn't the only optical networker affected by a slowdown in demand from Amazon:

Wall Street thinks Juniper's warning may mean capital spending slowdown from Amazon Amazon is hardly known for pinching pennies, but, like many of its customers, the Internet giant may actually be watching its bills heading into the holiday shopping season. At least that much was suggested by a warning from , which sells networking gear to companies like Amazon that operate huge cloud computing networks. Late Wednesday, Juniper

Lifting Amazon’s Cloud - MoneyBeat - WSJ

Applied Opto and Amazon have a vendor owned inventory warehouse where Amazon pulls at will.

• For all we know, that agreement and warehouse still stand, this should reduce the panick of a possible loss of Amazon as a customer.
• It also explains why it is so difficult for Applied Opto management to accurately guide sales, as it is basically up to Amazon, which used to be the customer responsible for half its sales.

Rest assured, the company is selling 100G, there is no reason to think that their competitive position is weaker in 100G compared to 40G:

In this quarter, 57% of our data center revenue was derived from our 40G data center products and 39% was from our 100G products which represents an increase of 62% in 100G sales from the prior quarter.

Applied Optoelectronics: Q2CC

Here is another useful assessment from jy263 at the end of an SA article:

I'm trying to understand reconcile the statements AAOI made in August with what they said this week in an attempt to understand potential Q4 guidance. Maybe I'm missing something, but I can't quite understand how all the facts fit together. In August, the stated problem was that AMZN switched over from 40G to 100G sooner than expected. That forced AAOI to retool their production lines from 40G to 100G in order to meet that new demand. That was expected to take 6 weeks. AAOI said they didn't know about this problem until after pre-release (mid-July). So if they started retooling in mid-July, it should have been complete by end of August. In any case, AAOI's low Q3 guidance only $107M to $115M) seemed to account for that.

This week, AAOI said that on-going transition resulted in AMZN-related revenue to be just 10% of total quarterly revenue. (Note: I'm understanding the CFO's use of "transition" to refer to AMZN's 40G to 100G transition.) Given that Q3 revenue was approximately $89M, the breakdown of AMZN/non-AMZN revenue was $9M/$80M. Revenue from AMZN was probably all from 100G parts. All AMZN revenue was probably in September only (after they finished retooling their some production lines to 100G). They also said that AMZN was a 47% customer in Q2 ($117.4*47%=$55.2M). Therefore, based on the above, it appears that AAOI was expecting AMZN revenue to be at least $27M ($107M - $80M). (Obviously, if non-AMZN guidance was lower than $80M, e.g., $70M, then the expected AMZN revenue would have been more than $27M, e.g., $37M. Likewise, if actual revenue exceeded its guidance, e.g., $120M, then AMZN guidance would probably have been higher than $27M.) (If AAOI was expecting AMZN revenue to be lower than $27M, then non-AMZN revenue--while it grew--did not grow nearly as much as forecast.)

So what I am trying to reconcile is why AMZN revenue was only $9M, which was way below the the $27M, which already accounted for the transition (and was previously $55M). I didn't hear anything on the call that accounted for this. But I feel like understanding the discrepancy between $9M actual and $27M expected is the key for Q4 guidance. Some possibilities:

(1) Unexpectedly lower demand possibility #1: AAOI management overestimated how much AMZN wanted to buy / AMZN has paused or decreased buying while doing their 40G to 100G transition (as it ultimately doesn't know how much it will need overall, and from AAOI-specifically). The CFO's statements that the problem was not specific to AAOI and that it's due to an on-going transition support this possibility. Juniper's soft guidance also could support this possibility. If this continues into Q4, revenue will grow only as fast as non-AMZN grows. But if AMZN comes back fully in Q4, it could be a $150M quarter.

(2) Unexpectedly lower demand possibility #2: AMZN decided mid-Q3 to buy less than $27M from AAOI and instead more from competitors (FN, INTC, etc.). The CFO's not-specific-to-AAOI does not completely contradict this possibility as AMZN may still be buying from AAOI, but not just not nearly as much. So rather than $55M in revenue per full quarter that AAOI might get, AAOI may only get, e.g., $20M. But the CFO's on-going-transition statement appears to undermine this possibility. If this continues into Q4, revenue will grow only as fast as non-AMZN grows.

(3) Unexpectedly lower supply possibility: AMZN wanted to buy more, but AAOI didn't have enough 100G parts to sell them because retooling took longer than expected. The CFO's not-specific-to-AAOI and on-going-transition statements appear to undermine this possibility. If this continues into Q4, revenue will grow only as fast as AAOI can make parts.

(4) Timing/low inventory possibility: AMZN delayed deciding when to buy AAOI parts until very late in the quarter. At that time, AAOI only had enough time left in the quarter/inventory available to sell $9M to AMZN. The CFO's not-specific-AAOI and on-going-transition statements may support this possibility. Juniper's soft guidance also may support this possibility. If this continues into Q4, revenue will grow only as fast as AAOI can make parts.

(5) Some combination of the above possibilities There could be other possibilities that I haven't thought of. But the discrepancy between $9M and $27M really concerns me. Some random thoughts:

(1) P/E still attractive

(2) I didn't hear on the call that AAOI was selling everything they are making. CEO/CFO were not asked, but they also didn't volunteer. I will be listening for this point in upcoming earnings conference call.

(3) That said, AAOI seems to be heavily investing in new manufacturing capability. So they seem to believe their future revenues are significantly limited by existing manufacturing capability.

(4) While it may be difficult to make accurate forecasts when AAOI has a VOI arrangement for AMZN, management either has a poor grasp on expected AMZN-related business, AMZN is being particularly opaque with AAOI, or perhaps both.

(5) Non-AMZN revenue is up significantly (on a percentage basis) QoQ over the last couple quarters. It will be interesting to see what kind of on-going non-AMZN demand there is.

Applied Optoelectronics: Ugly Q3 Miss, Transitory Amazon Weakness - Buy The Dip - Applied Optoelectronics, Inc. (NASDAQ:AAOI) | Seeking Alpha

And another interesting perspective from Guttaperka as a comment to the same article:

Jay, thanks for the run-down of the competitive landscape, awesome the way you share your vast knowledge! I share your assessment that if AAOI's AMZN business has been affected by another vendor, the most likely candidate would seem to be INTC. I still find it unlikely that this has happened. Reasons:

1. AMZN is surely more of a CWDM buyer than PSM, based on the size of their datacenters.

2. INTC only announced volume shipments of CWDM 100G products with Azure as their sole customer about 2 months ago, about one year behind their PSM product. INTC must be seen as a risky vendor to put a lot of eggs into, especially since they launched their PSM product in 2015 with much ballyhoo, only to then postpone commercial availability for one year. Not likely that AMZN would take such a risk with their new product.

3. Even reading what Richard Shannon (Craig Hallum) is saying, he mainly referred to AMZNs “hard shift” from 40G to 100G as the AAOI head wind and also said “Intel is making inroads into 100G optics as well", he notes, "and that could hit revenue growth as soon as Q4.” This does not specifically refer to AMZN, and if you google other research notes from him, he has been an INTC SiPh bull for at least a year, always talks about them.

4. There are actually no concrete rumours of INTC entering AMZN with SiPh, just speculation on these conversation forums. I’m sure that an important event like this would not go unnoticed by analysts. Moving over to my “theory” that AMZN simply ran into disruption by a sudden management decision that turned out more disruptive than expected, this is how I support it:

JNPR is obviously Amazon Web Services (AWS) other main core network vendor, (switches, gigarouters). AWS is (was?) JNPRs sole +10% customer. The exact same thing seems to have happened to them as to AAOI: "The (third-quarter) miss could suggest lumpiness at AWS, which we believe was Juniper's first 10% customer in several years," Steven Milunovich, a UBS analyst in a report. "Our industry contacts tell us AWS is changing its data center architecture; we believe this change could be creating softness in AWS spend on Juniper." Here is the article: http://bit.ly/2zloJFX

This would be the timing of the “sudden management decision” sequence. Around July 13: Right around the time for AAOIs awesome q2 update they get informed by AWS that they will be moving hard style from 40G to 100G. AAOI can’t tell anyone about it until their earnings call August 3 due to that they are in a “quiet period”. The info shocks the market on August 3. In the meanwhile AWS are scrambling with a new network design where they have to relocate 40G stuff to peripheral parts of the network to replace it with faster 100G in the central parts. This turns out to be very disruptive and time consuming slowing down AWS buildout severely. (Perhaps the network revamp also contains other complexities that we don’t know about). Around September 3: Stefan Murry is interviewed by Barron’s (published Sept 11th) right after hurricane Harvey has hit AAOI HQ. At that time he is probably worried because AWS has hardly picked up any stuff, but their contacts say it will resume soon, so he puts up a brave and optimistic face in the interview, where the topic is hardly touched upon.

Stefan says that 40G demand is going down, but that is “hardly a surprise” as it is being overtaken by 100G. October 11: JNPR announce soft guidance October 12: AAOI announce soft guidance. When AMZN data center core traffic handling buildout will pick up is anyone’s guess, but for sure they will be in deep sh-t unless they get rolling soon. Of course this theory of mine may not be the explanation, but to me, it is the most plausible turn of events to match what I’ve seen lately. I do believe that AMZN will bring in other vendors as we go along, but always in a smaller initial scale (like Macom-FN). I don’t think anyone else than AAOI has the credibility to take on AMZNs main massive 100G CWDM business at this time, and if that happened, we would know about it. Good luck to all!

Applied Optoelectronics: Ugly Q3 Miss, Transitory Amazon Weakness - Buy The Dip - Applied Optoelectronics, Inc. (NASDAQ:AAOI) | Seeking Alpha

Juniper Networks' turnaround hit a rough patch Thursday as the company pre-announced third-quarter earnings below consensus estimates, a surprise that one analyst attributed to weakness at customer Amazon.com. A rival of Cisco Systems, Juniper Networks has been seeking a turnaround under its new chief executive, Rami Rahim. It has aimed to sell more networking gear to cloud-computing providers such as Amazon, Microsoft and Oracle. Its biggest customers have been telecom companies, such as AT&T.

Juniper, in a release late Wednesday, blamed the disappointing outlook on "lower than expected revenue in our cloud vertical." Amazon Web Services is the cloud unit of e-commerce giant Amazon. "The (third-quarter) miss could suggest lumpiness at AWS, which we believe was Juniper's first 10% customer in several years," Steven Milunovich, a UBS analyst in a report. "Our industry contacts tell us AWS is changing its data center architecture; we believe this change could be creating softness in AWS spend on Juniper."

Juniper Turnaround Gets Stormy In The Cloud, Amazon To Blame? | Stock News & Stock Market Analysis - IBD

Applied Opto isn't the only optical networker getting headwinds from Amazon lately, this suggests it's more an Amazon problem than a problem specifically related to Applied Opto..

Applied Opto wasn't the only company to receive headwinds from Amazon's datacenter business in Q3:

Cowen & Co. analyst Paul Silverstein said in research note that cloud customers have driven Juniper's results for the last 18 months. Silverstein said Amazon likely cut spending with Juniper given the size of the revenue miss. Amazon was Juniper's largest cloud customer followed by Facebook, said Silverstein. If other cloud customers like Apple and Google cut spending the outlook may be worse for Juniper in the future. "we suspect that the issue is not purely 'transitory,'" he said. William Blair analyst Dmitry Netis said that Juniper was benefiting because cloud providers were building out networks to bypass carriers. By building out their own internet backbone, cloud players can deliver content and services to consumers and enterprises more efficiently. "Our recent industry surveys reflect material slowdown (and potential digestion) of IP backbone build out with a couple of these cloud operators this quarter," said Netis, who noted that the spending pause may cover multiple cloud customers such as Google, Facebook, Apple, and Amazon.

Cloud giants hit pause on Internet backbone spending, Juniper's Q3 suffers | ZDNet