Last week, I wrote an article entitled CEVA: A Good Time To Buy. I was in Barcelona this week at the Mobile World Congress, and took a deeper dive into CEVA (CEVA), spending time at CEVA's booth, talking with competitors and customers, and generally making a deeper analysis of CEVA. This article is a supplement to CEVA's investment thesis laid out in last week's article; I highly encourage readers to review that article and comments first, before reading the analysis below.
I would also like to address one reader comment by Richard Morrison. While I do not agree with some of his input, I do see his point that the article did not fully address CEVA's risks, which will be addressed below. Similar to Mr. Morrison, our Quan Technology Fund built-up a CEVA position and then a few years later scaled-out by January 2012 as CEVA faced some transition issues described in last week's article. But unlike Mr. Morrison, we did not envision that CEVA would land Qualcomm (QCOM), nor was that a large part of our initial investment rationale. Qualcomm has some core DSP technology that many other semiconductor companies do not have. Mr. Morrison laments that CEVA is not in the iPhone, but fails to mention that CEVA is in Samsung (OTC:SSNLF) smart and feature mobile phones, including all but the top-line Galaxy 4G model (for now), which is where most of global volume is today. I agree with Mr. Morrison, if CEVA lands Qualcomm, then yes, CEVA stock "could be over $100." But even if CEVA does not land Qualcomm or a stronger-mobile "Intel (INTC) with CEVA inside," CEVA stock can still potentially double from current and future business with Samsung, Intel, Broadcom (BRCM), Spreadtrum (SPRD), MediaTek (2454.TW), and a host of other companies.
Fund managers are inundated with information and stock ideas. We also have an urge to refresh the portfolio with new names, especially after riding a stock for a large gain like CEVA. Thus, most fund managers do not take the time to fully revisit their past winners, which subsequently tanked, and incorrectly categorize them as permanent broken stories. I often made that mistake, and in fact, in the current fund, CEVA is only the second time that we have gone back into a stock after fully exiting our position. This is a shame, because the built-up knowledge and perspective are wasted, especially when the company's outlook has truly changed for the better.
I would like to provide an insight on deep-dive analyses with a rough analogy. Most Seeking Alpha readers may know 1% - 5% about all there is to know about CEVA, a professional fund manager after doing a deep-dive may know 20%, but that still leaves 80% that is unknown. Even management/board members may know 40%, leaving 60% as unknown. The deeper you dive, for every one new answer, there are two more questions. Some questions nobody can answer as they are tied to variables beyond their control, and other data is just not reasonably available. Thus, readers should not take recommendations made in any Seeking Alpha article as gospel. Instead, readers should view at least my articles as one source of information that should increase their chances of a successful investment outcome, but with no guarantees. With those caveats, I will now proceed with the deep dive addressing CEVA's risks that were not brought out/emphasized enough in last week's article, as well as significant positive points that were missing in that article.
Diversifying from lower-value mobile baseband business
CEVA's primary challenge is to diversify into new sockets such as audio and video on the phone, and into other areas like automobiles, M2M, foundries, etc. As mentioned last week, these addressable markets pale in size to the mobile baseband business. These new licenses may be 10M units per year, versus 50M or more units from baseband licensees. CEVA accelerated its non-baseband penetration last year, and now has about 25 customers. CEVA's chance of success increases as the number of non-baseband licensees continues to rise.
Some of these licensees' products may never reach the market, but some of them should. CEVA takes some risk by working with unproven startups in very promising areas. I saw some incredible demos using the CEVA-MM3101 DSP platform from startups like eyeSight and nViso. CEVA also added more mature licensees like Yamaha (OTCQX:YAMCY). In a worst-case scenario, none of the current 25 licensees or future licensees will be commercially successful. In a rosy scenario, some of them hit 50M units, many more hit 10M - 15M, and a few fail. This would produce $50M in annual royalty revenues in a few years assuming 500M units x $0.10 royalty rate (royalty rates are usually higher in these areas), roughly equal to 2013 total revenues. The base case may be $20M using today's current 25 non-baseband licensees.
Strong and diversified licensee growth is the biggest challenge for CEVA. While CEVA is dominant in baseband, CEVA has more competition on the audio socket from Tensilica. Similarly, in video/imaging mobile sockets and faster-growing non-mobile markets like M2M, CEVA does not have the incumbent advantage. Thus, CEVA needs to continue to grow its non-baseband licensee base and successfully penetrate the LTE baseband business. Otherwise, sources of future royalty growth flatten out and may eventually decline when existing royalty end-markets mature and inevitably face some pricing pressure. Even in this scenario, CEVA would be a highly profitable cash machine, but the growth would be less, leading to a lower valuation.
LTE (long-term evolution) 4G mobile baseband opportunity
According to management, CEVA has signed 20 licensees for LTE. Two names missing from that list are Intel and Broadcom. I spoke with an Intel LTE representative at the show. He said that this was due to CEVA not meeting certain size and high-volume data feed requirements within a certain time frame. So, Intel chose a less-optimal, hard-wired ASIC solution to meet its immediate LTE (4G) DSP chip needs. This raises more questions. How can Intel not use internal DSP capabilities for 2G and 3G where its uses CEVA, but hope to succeed with internal DSP capabilities for a more complex 4G standard? Is this a permanent strategy or a temporary one? Broadcom's move is more explainable with last year's acquisition of Renesas LTE assets, which had some DSP capabilities, although not at the level of CEVA's knowhow, and again is a more limited hard-wired ASIC solution. Thus, while it is still early days, a risk exists that CEVA will have lower baseband market share without Intel, Broadcom as well as Qualcomm.
On the other side of the coin, Samsung has already shipped LTE smartphones with CEVA DSP cores (see here for details). Even for Samsung's premium smartphone, which currently relies on Qualcomm, Samsung is considering using its own baseband chip in early 2014. If this happens, this could add about $5M in additional annual revenues just in 2014 (50M x $0.10 royalty), and multiples more in future years.
Qualcomm's current stranglehold on the LTE multi-mode mobile chip market is loosening. Several competitors are entering the LTE chip market; many of them are CEVA licensees. China is especially motivated to avoid paying high Qualcomm royalties after 3G. 100M LTE mobile phones are expected to be shipped in China in 2014, before a broader launch in 2015. Most of these will be low-to-mid-range smartphones outside of Qualcomm's realm, using the Chinese TD-LTE standard. If Spreadtrum, Leadcore, now MediaTek (see here), and other well-positioned Chinese CEVA licensees control just 30% of that market, CEVA would receive incremental royalty revenues of $3M in 2014 with much more to come in 2015. These upside figures are not factored into consensus estimates.
Some may argue that a $0.10 royalty per LTE chip is not sustainable. On average, ARM (ARMH) earned $0.10 per primary function per chip, and $0.18 per chip with combined functions in 2012 (see here - slide 25). The value-add (video streaming, etc.) and chip and phone price points are much higher for LTE phones versus 2G and 3G phones. Even with cajoling management, I could not get them below 8 cents in a few years, and not lower than 6 cents at market maturity.
Tensilica is a formidable #2 competitor. Tensilica was acquired by Cadence (CDNS), a leading EDA (electronic design automation) company for $380M in 2013. Cadence paid an EV/revenue multiple of 8x ($355M/$44M). Earlier in 2013, Cadence management was fairly transparent about Tensilica's total revenue, projecting 2013 revenues of $57M, including expected royalty income of $13M. But more recently, Cadence stopped quantifying Tensilica results, even after being directly asked by analysts on quarterly conference calls, only stating that Tensilica performed ahead of plan. Furthermore, some of Tensilica's revenues come from configurable RISC processor cores for non-DSP applications. The DSP revenues are principally from audio, and to a lesser degree, video sockets, as opposed to baseband sockets. Thus, getting DSP core market share data is increasingly difficult. The latest Linley Group studies that I could find show CEVA with 90% DSP core market share according to this article, and Tensilica with 20% market share according to this release using the same Linley Group source. Thus, we can estimate that CEVA has a dominant position today, with CEVA stating that it is 3x larger than Tensilica. But with Cadence backing Tensilica, it has become a more formidable #2 competitor.
Another source of competition is in-house solutions from potential semiconductor customers. As discussed, the main semiconductor company today that does not use CEVA at all is Qualcomm. One risk is that in the future, some larger semiconductor companies will revert to in-house DSP design. Another risk is that Qualcomm may try to license its DSP design to third parties, although the company has not made any indications to do so and is unlikely according to this article. The countervailing factor is that CEVA has invested thousands of man years for over a decade into developing its ecosystem, and now has a massive developer community supporting its DSPs. This is why 5B CEVA-powered devices have been shipped to date and that CEVA's market position remains so high.
ARM is not a competitor, and apparently has given up on the DSP core market. ARM had backed a startup, Cognovo, which was involved in DSP cores. But Cognovo was sold to Swiss-based u-blox (UBXN.SW) in 2012. U-blox happens to be an interesting company in its own right, with solid technology, serving the fast-growing M2M market, and ramping revenues over the past few years.
In reviewing these risks in light of a deeper dive, I still think that CEVA continues to have an attractive risk-reward tradeoff. The biggest risk of not diversifying and growing licensee revenues is being thwarted by recent results. In the words of Marty Zweig - "the trend is our friend." The acceleration in licensees over 2013 culminated in record Q4 licensee revenue. Royalty revenue reversed its downward sequential spiral in 2013 Q4 as well. Furthermore, we have visibility for continued license growth in the current quarter. Today's license growth is tomorrow's royalty income.
Aside from ARM, and possibly Imagination (IMG.L) and Immersion (IMMR), CEVA is one of the rare semiconductor technology companies with a substantial royalty stream tied to growing end markets, ~90% gross margins, and strong cash generation. CEVA is a stock to hold for the long term.
Now is a good time to enter CEVA, as the market has not factored in the incremental royalty revenues that will come in Q4 2014 and 2015. Royalty revenues in the current quarter will still be below the year-ago figure, so we are not completely out of the woods yet. CEVA needs to continue to execute on licensing deals, garner market share in LTE baseband in China and elsewhere, and fend off competition. But we see no signs that management will break recent positive operating trends and not execute.
By the time that CEVA's rebounding revenues have been confirmed and royalties are on a full upswing, CEVA shares will be much higher. Over the next 18 months, CEVA shares should return to the mid-30s-highs that they achieved in 2011. Again, for more details on the CEVA investment thesis, please refer to last week's Seeking Alpha article.