August 02, 2017 - 11:43am EST by
2017 2018
Price: 96.00 EPS 5.17 5.58
Shares Out. (in M): 20 P/E 19 17
Market Cap (in $M): 1,900 P/FCF 30 22
Net Debt (in $M): -30 EBIT 137 154
TEV ($): 1,870 TEV/EBIT 14 12.5
Borrow Cost: Tight 15-50% cost

Sign up for free guest access to view investment idea with a 45 days delay.



AAOI is a short with approximately 80% downside to our $20 price target, which we believe could be achieved within 6 months time. The catalyst is a major loss of market share at its top three datacenter customers, AMZN, MSFT and FB (55% / 15% / 8% of total sales, respectively). This share loss will occur as all three customers transition networks to 100GBe, where AAOI's share is significantly lower than their incumbency in 40GBe.

Risk reward is asymmetric currently due to AAOI's extremely strong financial results during Q4 16 - Q2 17, whereby consensus 2017 EPS was revised from <$1.50 to above $6 at present. Shareholders and sell-side analysts are extrapolating recent AAOI's recent successes into the indefinite future, which we believe to be a massive miscalculation.

For context, AAOI makes fiber optic transceivers that convert light into electrical signals and vice versa, enabling the basic functioning of a fiber optic network. The transceivers consist of lasers and electric circuitry which are combined into modules. AAOI sells modules and manufactures its own lasers in house, making it a capital intensive, vertically integrated supplier of optical components. To support our argument that AAOI is capital intensive, we direct attention to AAOI's 2016 10-K which notes capex of $45 million, $62 million and $41.5 million in 2014, 2015 and 2016, compared to cash from operations of $8.5 million, -$15 million and $57 million, respectively. Therefore, across the trailing three years which includes the recent uptick in business , AAOI has actually burned nearly $100 million of free cash flow.

Based on conversations we have held with industry experts, we believe that the recent spike in AAOI's business, while impressive, has been driven by short-term factors. AAOI is the incumbent supplier of 40GBe fiber optic intradatacenter builds at AMZN, MSFT and FB. Clearly, these datacenter operators have been expanding at a breakneck pace and we make no argument that this is about to change. However, all three companies are in the process of transitioning their networks to 100GBe, where AAOI is not incumbent. In fact, AAOI is strategically disadvantage in a major way at 100GBe due to a vast array of competitors with greater financial resources, lower cost structures, and stronger technology.

The case that AAOI is a 40GBe driven story is not conjecture. Management discloses 40G vs. 100G revenue on its quarterly earnings calls, and in Q1 2017 revenue from 40G was 62% of datacenter revenue vs. 30% from 100G. 100G did grow dramatically q/q in Q1 '17, from $13 million to $23 million (+70%) and likely more in Q2 '17, but 40G is the lion's share of the business. Additionally, we believe a fair amount of the Q1 and Q2'17 volumes in 100G are at risk of being cannibalized imminently as new supply from competitors comes online, which will make 40G even more critical for AAOI.

The competitive landscape in intra-datacenter 40GBe is narrow, with AAOI and Innolight (private company owned partly by Google) dominating the market. The narrow set of competitors is due to the fact that 40GBe never became a widely endorsed networking standard, with 10G having been dominant and 100G the next frontier for some time. Accordingly, competitors such as but not limited to Intel, Luxtera, Mellanox, Finisar, Source Photonics, Kaiam, Oclaro, Lumentum and others chose to focus their investments on 100G which was sure to be the larger market. However, the major cloud CSPs, who are on very rapid datacenter build cadences, did opt to move to 40G optical as a "band aid" until 100G ramped, creating the significant opportunity for AAOI.

The primary gating factor for deployment of 100G optical in the datacenter has been simple availability of supply, as referenced many times by companies with good visibility such as Broadcom, Arista and Mellanox. This shortage was elongated by a major network upgrade cycle in China during 2015 and 2016 that had the effect of siphoning off EMS supply at Fabrinet and vertically integrated supply at Finisar and others that could have otherwise been dedicated to ramping 100G intradatacenter transceivers. Fast forward to today and the China Network upgrade cycle has widely reported to be in a digestion phase, with Oclaro, Fabrinet, Neophotonics, Finisar and Lumentum all revising revenue outlooks substantially lower in the past 6 months while simultaneously adding extreme amounts of capacity. This excess capacity is being repurposed, per all vendors, from "CFP" capacity allocated to the telecom NEMs, and into "QSFP" capacity with fungible application for telecom and intra-datacenter.

In addition to these legacy optical vendors (traditional AAOI "comps"), an entire gang of suppliers are present in 100G that did not play in 10 or 40, namely the silicon photonics vendors. The best known vendors of silicon photonics optical transceivers for the datacenter are Luxtera, Intel, Mellanox and Kaiam, all of whom are selling in volumes approaching or exceeding $10s of millions a quarter at present. Silicon photonics suppliers benefit from the lower cost of the silicon CMOS process as compared to the indium phosphide manufacturing process of most optical vendors. This is allowing the above vendors to quote prices far below the norm as they seek to (and succeed in) winning business from vendors like AAOI in the datacenter.
In addition, there is an emerging threat from the "whiteboxing" of transceivers. In Q1 2017, MA/COM and Fabrinet announced a collaboration to produce transceivers for Amazon, disintermediating the module maker (AAOI). MA/COM has stated that at their Lowell, MA facility, they are beginning to ramp production of 25G lasers (4x25 for 100G transceiver) for this project in August with the aim of being able to supply "the entire 100G industry" with 25G lasers by December. MA/COM is a troubled company that we are also short, and we believe they will sell these lasers at very low margins in order to backfill revenue shortages in other product areas. In addition, we believe that other contract manufacturers besides Fabrinet and other 25G laser suppliers besides MA/COM are talking to leading hyperscale customers about whitebox transceivers.

As a result of the divergent competitive set in 100G vs. 40G, where AAOI was massively privileged to be one of the only 40GB transceiver suppliers not partly owned by Google, a major competitor to AAOI's customers, we see AAOI's market share dropping precipitously in 100G. At the same time, we see the 40G intra-datacenter market plummeting in the second half of 2017 as the major "cloud titans" move to 100G.

The confluence of these factors leads us to forecast revenue and earnings for AAOI well below street consensus for 2H 2017. For example, we believe revenues are likely to decline to around $95 million in Q3 and $80 million in Q4 of 2017, compared to consensus at 122m / 130m (and whisper much higher) while earnings should approach $0 in Q4, compared to consensus of $1.33. In the event our view, or anything approximating it becomes true, AAOI is headed much lower in the coming weeks.


  Back to top

I do not hold a position with the issuer such as employment, directorship, or consultancy.
I and/or others I advise hold a material investment in the issuer's securities.


negative revision to outlook on earnings call 8/3
Market share loss at leading datacenter customers becomes manifest (this has occurred, but is not yet acknowledged by the sell side)

4       show   sort by    
      Back to top