I am normally very tight lipped about my shorts since public theses often cause borrow costs to rise and volatility to increase, especially for sub $3bln tech companies but I hope this makes up for any poor marks in MX I may have caused.

Ambarella (AMBA) is a fabless developer of system-on-chip (SoC) products for the high-definition video-capture market, offering video/image processing on a single chip for the security surveillance market (40% of revenue), the action camera market (35% of revenue), the automotive market (10% of revenue), the broadcast/film infrastructure market (6% of revenue) and various additional smaller markets (6% of revenue total).

AMBA is overvalued because its end-markets are bifurcating; there is significant competition and little continued innovation at the low end, while growth at the high end decelerates. I believe the stock is worth about $18 in a base case scenario (18x $1.05 in FY18 EPS) or $12 (5x Price/Gross Profit) in a downside scenario where the company is acquired post margin decline.

Why is AMBA misvalued? This is a question I ask myself on every investment.

(1) First Mover/Fast Growing Markets

Bulls believe that AMBA should benefit from the continued transition of analog to IP in the security surveillance market as well as the continued growth of GoPro within the action camera market.

(a) GoPro

GoPro represents 35% of AMBA’s revenue. Bulls assumes GPRO should grow units at 25-30% yoy and maintain ASPs as they hold market share in the fast growing action camera market. They believe GPRO continues to sole source AMBA chips with limited SoC ASP declines.

(b) Security Camera Market

The security camera market represents 40% of AMBA’s revenue. The long thesis extrapolates IHS market estimates of digital security camera unit growth from 22M in 2014 to 73M in 2018, a 35% CAGR. They assume that AMBA maintains its market share and ASPs in order to grow revenue at similar levels.

(2) Earnings Beats/Unexpected New Markets

(a) Quad-Copters

In FQ2 2015 (quarter ending 7/31/2014), AMBA reported revenue of $47M vs. $45M consensus and EPS of $0.37 vs. $0.28 consensus. They guided to $60-64M for FQ3 vs. $52M consensus and $0.50-0.56 in EPS vs. $0.37 consensus. Never had AMBA (in conference calls or meetings) discussed Quad-Copters (drones) as a potential new end market, yet it comprised over $1M in the quarter (half of the beat). The stock rallied from the high 30s into the mid 40s on the results and sell-side exuberance. The company noted that while increased GoPro adoption and more 4K in IP security/growth in home security cameras caused the beat, the enthusiasm surrounding the stock was associated with new end markets that management could not even predict.

In FQ3 2015, AMBA reported revenue of $66M vs. consensus of $63M and EPS of $0.68 vs. consensus was $0.54. They guided to $57M-$60M for FQ4 vs. consensus of $51M and EPS of $0.45-$0.52 vs. consensus of $0.34. However management mentioned they expected FY16 sales growth to decelerate and margins to potentially degrade, resulting in somewhat of a stabilization of the stock.

In FQ4 2015, AMBA reported revenues of $64.7m vs. consensus at $60m on strength in GPRO and IP security cameras and guided FQ1 2016 to $64-68m vs. the street at $60m mostly due to the Xiaomi action cam at $64 using the A7LS chip. The market has further cheered at the expectation of another new, large customer who became the third largest smartphone manufacturer within less than two years.

(b) Cop-Cams

At CES in January 2015, AMBA debuted a new 1080p cop camera. Much like at CES in January 2014 when AMBA debuted Google Helpouts’ “around neck” cameras, the bulls heralded the new end market (the Google Helpouts product never progressed past beta). Cop-cam revenues are currently negligible but AMBA estimates the market size at 800k total police officers in the US and an additional 7.2M internationally. At a $399 end market ASP and a $10 SoC ASP to AMBA, management believes this could represent another 10% in annual revenue (at a four-year replacement cycle). While the market itself is small, the bulls believe it represents one of many currently unimaginable end markets for AMBA.

(3) Niche Stock

Less important to the overall bull case, AMBA is a niche stock. It is an interesting, innovative, small and fast growing semiconductor company, very different from the TXN’s, ONNN’s (analog semis), TSM’s, and UMC’s (digital semis) of the world, that grow at 8-10%. Many growth-oriented tech funds that must own semis rotate into growing fabless companies like AMBA, further amplifying the mispricing on the upside.

Short Merits:

Action cameras (mainly GPRO) and security cameras represent almost 80% of AMBA’s revenue and the bifurcation occurring in both segments represent the crux of the short case.

(1) IP Security Surveillance Camera Market (Slowing of the High End and Technology Wall)

As recently as 12 years ago, the security surveillance market consisted of only analog cameras—that is, a single camera connected directly to a monitor and tape recorder. Several large manufacturers controlled the market through a network of integrators with strong relationships with end customers. Over the past 6-8 years, the market has transitioned from analog to IP-based cameras integrated with a single hard disk drive recording device and a video management system (VMS). Given the slow moving nature of the entrenched analog companies and the limited technological sophistication of the integrators, the incumbents were slow to change and a host of new IP camera manufacturers took share, many using AMBA’s chips. However, these companies primarily serve mid-to-high end customers, the precise market segment where the transition from analog to digital is already largely complete. According to AXIS, the high end of the market is 95% penetrated by IP-based solutions. Therefore, while the overall industry may still grow at a 15% annual rate, most of that growth will occur at the very low end, such as small gas stations and convenience stores, which are extremely price sensitive and generally only install 1-10 cameras (720p). My research indicates that the mid/high end of the market should grow in the low single digits vs. 30%+ at the low end; this is end market growth and does not factor into price competition at the SoC level for AMBA.

Further supporting the AMBA short thesis, competition from Asian manufacturers is increasing as industry innovation has hit a standstill at both chip manufacturers and camera/software providers. While AMBA has historically offered new products (720p, then 1080p, 4k with increasing frame rates per second) six months ahead of competitors, 4K at 60 frames per second is the last innovation in the industry. AMBA, multiple security camera manufacturers and integrators have told me that customers are unwilling to replace 1080p solutions for 4K, and as demand grows at the low end, those customers opt for 720p solutions where competitors sell $5 ASP SoCs (vs. $9 at AMBA). Over time, incremental growth (i.e. most of AMBA/competitor sales) should come at the low end at $5 ASPs and 29% gross margins (calculated using AMBA BOM cost of $3.50 and $5 competitor ASPs). Lower chip pricing by multiple new entrants has spurred a competitive race to the bottom, which has in turn lowered the costs for new IP camera players to enter the market. Camera manufacturers such as Samsung, Hikvision and Dahua have followed suit by cutting prices, resulting in continued pressure on their SoC suppliers. With Panasonic’s recent acquisition of Video Insight and Canon’s acquisition of Axis, slowing growth at the high end is clearly causing consolidation, further pressuring component supplier prices.

“What we think is happening now is that a commoditization of hardware, particularly IP and megapixel camera, is underway. There is an increasing influx of typically Asian-made, low-cost cameras hitting the market, which is to be expected and is not something we’re surprised by,” –Avigilon CEO

"The year 2014 has been one of the most interesting—and disruptive—in recent memory, for both the professional and consumer video surveillance industries," Jon Cropley, principal analyst for video surveillance and security services for IHS, said in statement. "A sharp decline in the cost of semiconductor components has ushered in a new era of price competition, and the competitive landscape has shifted, with merger-and-acquisition activity affecting some of the leading global product and software vendors."

While our data aren’t perfect given the fragmentation of the security surveillance market, we know that there are many very price sensitive/elastic customers that will convert from analog to digital at a comparable price. This includes convenience stores and gas stations as well as extremely low-traffic areas.

These elastic customers in commercial, retail, banking and manufacturing did not convert from analog to digital as the market grew from 2009-2014 and as IP (digital) grew from 15% of the overall market to an expected 51% in 2016. The bottom line is that the 50% of the market that will upgrade in the future will do so at a significantly lower price of both the overall camera and the SoC.

Conclusion: As (1) security camera sales at the high end slow, (2) technological innovation slows and (3) sales shift to Chinese camera vendors who favor homegrown/local SoC vendors, AMBA’s sales/margins should decline substantially.

(2) Action Cameras (More Demand at the Low End and Increased Competition)

The success of GPRO’s IPO and lofty consensus expectations partially drive the misvaluation in AMBA. With bulls estimating GPRO maintains its 75% market share but increases units from 4M today to 10M by 2018, AMBA consensus does not properly evaluate a similar bifurcation and increased competition in action cameras as security cameras, which will further impair AMBA’s earnings power.

At the low end, especially internationally, Novatek-enabled action cameras range from $70 (1080p30/720p60) to $175 (1080p60). Sony is also taking share at the high end with its AS1000V (4K30) with OIS and GPS.

If we assume consensus estimates for GPRO unit growth of 25-30% hold over the medium term, SoC prices should still decline as low-end cameras capture a disproportionate percentage of GPRO’s unit growth and these cameras contain a cheaper AMBA SoC.

(3) Manufacturing/Competition (Superior Technology is Meaningless Going Forward)

As previously mentioned, AMBA became the sole-sourced SoC supplier for GoPro (and the security camera market) because it was first to market with encode/decode software and an HD solution at the leading manufacturing node. While AMBA is still the market leader and unveiled the industry’s only 4K60 chip at 14nm in October 2014 (to begin manufacturing at the end of 2015, and in products by 2H 2016), at least nine months before competitors (except Sony), their lead is less valuable as the low- and high-end security and action camera markets bifurcate. Manufacturing at the leading edge provides AMBA with a power advantage over competitors. However, with a 12+ hour battery life at 28nm at 1080p60, the only market where 14nm provides any useful power benefits is at 4K, or the market for $499+ action cameras. This market will likely represent only about 20-30% of AMBA’s action camera segment (8-10% of AMBA’s overall revenue) as the low end (720p/1080p) represents the majority of incremental growth going forward. Also, given the increased R&D and manufacturing costs at 14nm (hardest node without EUV equipment which means a higher cost per transistor), GPRO will likely see some marginal gross margin pressure even here.

In contrast to AMBA’s pricing power at 4K (at 14nm), the rest of their business is massively commoditizing given the competitors listed above/below. With pricing in the security camera market dropping from $1,000 for high-end US/European manufacturers to $500 for Asian products at 1080p and $250 at 720p, few price conscious buyers will pay a premium and few manufacturers will use a premium chip given $5-6 SoC ASPs at the low end. Moreover, as the incremental action camera enthusiast opts for one of the many low end products, action camera manufacturers will also very likely use a low end $5-6 SoC.

Current Major Competitors:

Hisilicon

Hisilicon is a large Chinese SoC maker that recently launched an H.265 SoC (3516A and 3536) that matches the compression functionality of AMBA’s S3 SoC. Hisilicon has 80% market share in the NVR market in China, through which they can easily grow market share in IP camera SoCs. Hisilicon has historically taken share at the low end but the introduction of their 3536 chip changes that.

Dahua

Many Chinese security camera vendors are adopting a new technology that combines analog with coaxial cable, calling it HD over Coax 2.0. Dahua’s DH5000 chip combines an image signal processor chip with a transmit chip. They say the cameras provide clear and crisp images at 1080p and 720p with good night vision.

Axis

Axis, a security camera vendor, utilizes its own SoC, the ARTPEC 5. Axis is the largest security camera company, known as the leader at the high end. If the barriers to SoC development were so great, they would outsource to AMBA.

Novatek

Novatek supplies SoCs to action camera competitors of GoPro’s. Novatek’s 96650 is used in the SJ4000 1080p camera, popular in Asia (but available internationally) which sells for $70, 45% lowers than the low-end GoPro Hero. As previously mentioned, Novatek ASPs sell for $5.

Sony

Sony sells its own SoCs in its action cameras. The new Sony AS1000V debuted at CES offers 4k at 30 frames per second, as well as optical image stabilization and GPS functionality.

While these are the major competitors, all the other competitors in the list above sell SoCs at significant ASP/margin discounts to AMBA.

Forecasts

·         Action Camera Revenue

o   Assume AMBA sells units with a 3-5 month lead time to GPRO

o   Assume consensus unit estimates for GPRO through FY16 and then 15% GPRO unit growth in FY17/FY18

o   Assume GPRO ASPs decline based on an increased mix of $8 ASP SoCs and a decreasing mix of $12 ASP SoCs (from 90/10 to 50/50 by the end of FY16) and 15% annual ASP degradations thereafter

o   Assume “other” action camera revenue growth of 46%, declining to 10% in FY17/FY18

·         Security Camera Revenue

o   Assume 25% unit growth yoy at the beginning of FY2016 which declines to 15% by the end of FY16 and stays at 15% through FY18, along the lines of IHS unit estimates

o   Assume 10-15% ASP declines, as ASPs drop from the current $9 to $5.80 by the end of FY18

·         Assume flat infrastructure revenue, 13% annual auto revenue growth and 45% annual other camera revenue growth due to potential new end markets (drones/cop cams)

o   Assume 50% market share in drones since the largest market share manufacturer (Parrot) uses its own SoCs

·         Assume 4% annual BOM cost declines along the lines of lower manufacturing costs, hence cost per chip declines to $3.25 by FY2018 vs. $3.75 today

·         Assume 15-18% R&D growth in FY16 to reflect higher 14nm tapeout costs as discussed by management and 3% growth thereafter

·         Assume flat S&M expense, tax rate and share count

Free Options

Should GPRO dual source, which they told me they will within the next year and I believe they will do before the CY2015 holiday selling season (the October 2015 quarter for AMBA), the loss to AMBA’s earnings power is as follows. Moreover, should GPRO lose share as discussed above, the loss to AMBA is also listed below.

Above are losses to earnings.

Price Target

Assuming AMBA does $1.59 in EPS in FY17 and $1.05 in FY18 vs. consensus at $2.71 for FY17 and $3 for FY18, the stock should easily fall to $18 (18x $1.05).

If they lose 20% of GPRO’s business to dual sourcing and GPRO loses another 20% market share, $1.05 in EPS easily becomes $0.80, hence a likely high single digit stock (tacking on share loss adds more downside).

Moreover, should gross margins fall below my estimated 49.5% at the end of FY18 to a more realistic 40% (similar margins to FIO and STEC, two other single product semiconductor companies in markets where competition increased), we can value the business on a takeout price to gross profit multiple. Given that all the sales force and some of R&D can be cut, this would be similar to a cash flow multiple. Both FIO and STEC were acquired at 5x P/gross profit, which would yield a $12 AMBA stock price. This does not include GPRO share loss or dual sourcing.