The inventor...looks upon the world and is not contented with things as they are. He wants to improve whatever he sees, he wants to benefit the world; he is haunted by an idea. The spirit of invention possesses him, seeking materialization. - Alexander Graham Bell
The figurine pictured above is a portrayal of the ancient Hindu deity Hanuman. The tale of Hanuman is a very ancient one, captured at least as far back as 500 BC in the Rig Veda and quite possibly predated by stories originating at least 1,000 years earlier.
Hanuman is an extraordinarily heroic figure - a demon slayer and leader of armies, possessed of great strength and immune to injury from fire or weapons, he is also a very scholarly figure and first among intellectuals. In short, Hanuman encompasses all of the highest virtues to which humanity aspires.
During his many adventures, Hanuman manifests a plethora of talents and abilities. His legendary swiftness of both mind and body draw comparison to the god of the winds, along with allusions to Hanuman being an agent of change, moving mountains and stopping the course of the sun across the heavens in order to smite evildoers and bring justice to the world.
In many of his travels, trials and quests, Hanuman demonstrates the ability to alter his form - sometimes expanding his physical self and at other times transforming into a different shape altogether. Due to this ability, Hanuman is considered to be one of the Vanara, mythical beings from Hindu scripture known for their shapeshifting abilities.
Though there is not yet a Hanuman figure among them, the leading companies in the Programmable Logic sector serve some of the same purposes as the Vanara. FPGAs and CPLDs can be reshaped at the will of the user - in some applications, even dynamically. Using SRAM to in effect reproduce the truth table of functions up to six inputs and employing configurable multiplexers to extract the output of such functions from a given set of inputs, programmable logic can emulate an incredible range of combinatorial, sequential and datapath circuits.
By emulating digital logic functions, FPGAs and CPLDs are much less efficient than their fixed function counterparts in the chip sector and are consequently much pricier, consume much more power and suffer from much lower performance. Nevertheless, the flexibility of such devices is so appealing that designers are willing to put aside their requirements for performance, cost and power, sacrificing such concerns in order to enjoy maximum freedom and reconfigurability.
As a result of this, programmable logic is most frequently found in markets that are are highly dynamic and experiencing healthy growth. In such markets, cost takes a back seat to feature growth & expansion, with the at-will configurability of PLDs helping system designers to pile on new features while simultaneously reducing TTM (Time To Market.)
The appeal of programmable logic has proven irresistible on occasion even to system designers in the supremely cost-sensitive Consumer segment. In the first decade of the new millenium, some HDTVs were released to market with Xilinx FPGAs in them. Lattice even had one of their extremely low cost and low power iCE devices show up in a mobile phone.
The 'shapeshifting' abilities of programmable logic also lend it to serving as an indicator of future winds of change. With a typical 6-9 month leadtime between an OEM sourcing their FPGA or CPLD needs and the device showing up on a PCB in a newly released system, the design-in activity of programmable logic companies and their products constitute harbingers of things to come.
To represent the Vanara of High Tech, I have stripped Xilinx out of the Stone Masons group and merged it with Lattice and Altera in this new category. There are product and technology strength differences between the three companies but, broadly speaking, they by and large cover the same markets - communications (both wireless and wireline infrastructure), consumer (to a limited extent), ISM, some experimental High Performance Computing, Mil/Aero and Automotive.
Now let's take a look at the numbers, from Q1 2008 to Q3 2014 (revenues in $B):
1. Altera and Xilinx are nearly in lockstep. Both took a beating in 2009 and experienced difficulties towards the end of 2011 and beginning of 2012.
2. Altera has clearly been trying to overtake Xilinx to repeat their performance of the latter half of the 1990's. The company came very close between Q3 2010 and Q32011 and seems to be on a trajectory to try again sometime in the second half of 2015.
3. All three companies have been effectively stagnant for the past four years.
Historically, Altera has been somewhat more adventurous than Xilinx in searching for new value-adding capabilities. The company put substantial effort into developing an ASIC-like migration path for its FPGAs to support high volume opportunities, but it never really caught on to any significant extent with customers. Altera also experimented with combining CPUs and FPGA fabrics with standard cell logic in smaller devices but could not finding a winning solution.
The serial weakness of Xilinx quarterly revenue reports this year might be attributable to their latest toolset. Field reports of the 2014 release of the Vivado suite suggest a lack of basic features and support for things such as file naming, OS compatibility and a variety of other capabilities. The reaction from field applications people has been scathing - a combination of incredulity and disgust.
Both Xilinx and Altera have standardized on ARM for an embedded CPU core in their high end FPGAs and are positioning themselves as "Programmable SoC" houses, with high performance SERDES I/O, embedded multipliers, feature-rich programmable cells, hierarchical interconnect schemes and a wealth of home grown and 3rd party embeddable IP. Also, both companies have come to depend on a high end complex FPGA line for the bulk of their sales, supplemented by one or two derivative product lines that incrementally reduce features and performance for moderate improvements in cost and power. For the better part of two decades, the offerings of the two companies have been so much alike that major accounts often switch off between the two regularly in order to keep each of them honest.
The two competitors are, at the end of the day, nearly identical to each other in every respect. Since between them they dominate roughly 80% of the programmable logic market, the companies cast a shadow of technology stagnation across the sector, compounding their own problems in competing against each other or finding avenues for genuine value-add and organic growth.
This brings us to the runt of the litter - Lattice. The company's history over the last 15 years has been checkered, to say the least.
Lattice bought Vantis (the AMD spinoff of its programmable logic division) in 1999 and Agere's FPGA division in 2002 but could not turn those acquisitions into permanent revenue gains. After a string of questionable actions and statements by CEO Cyrus Tsui culminating with a lavish birthday party for his wife thrown at company expense (an event that apparently featured an appearance by 1960's crooner Paul Anka and which cost the company $1M, including $300k for Anka's appearance), Tsui was fired in 2005 and may have actually fled the country with federal investigators hot on his tail concerning potentially illegal activities regarding stock option grants and technology exports to China. Several CEOs came and went over the next few years.
Finally at the end of 2011, Lattice bought Silicon Blue, a startup that had developed an extremely lean FPGA architecture that emphasized very low power and low cost. The acquisition proved to be an excellent one, as Lattice essentially ran away uncontested with the low end of the programmable logic market.
The single greatest problem with all three companies is that they are still hung up on the programmable paradigm after 30 years. The idea that the flexibility of configurable logic is a supremely valuable capability to system designers continues to captivate the marketers and executive management of programmable logic firms in the same way that scripture dominates the thoughts of true believers in a religious cult, to the exclusion of any outside ideas that do not conform to doctrine.
What It All Means
With the singular exception of Lattice's iCE line, the offerings of all three companies fail by and large on all three P's - price, performance and power. Thus, their only real claim to value-add in system design is flexibility. The fact that the last four years have witnessed desultory financial performance from the three firms, however, suggests a significant decline in the appeal of flexibility in system design in favor of lower cost and power.
Stated differently: programmable logic design activity is a bellweather for the relative health of the High Tech industry. The revenue numbers over the last half decade indicate that there is a lot less new, original design work being done. Most system engineering efforts are now devoted to optimizing current products in terms of cost and reliability while incrementally adding features (within the bounds of a static or even reduced cost profile) in order to eke out short-lived competitive leverage in the marketplace.
This is a bad sign both for programmable logic and the industry in general. With configurability having taken a back seat to power and cost considerations, the resultant revenue numbers for programmable logic provide yet another confirmation that the global High Tech industry has entered a new phase in its history - the Lean Times of Pharaoh discussed in previous editorials.
The soul, secured in her existence, smiles
At the drawn dagger, and defies its point. - Joseph Addison
When the winds of change alter course and design wins for programmable logic companies begin to grow substantially, we will see a recovery of the growth prospects for both programmable logic companies and the High Tech industry in general. In the meantime, though, the runt of the litter is actually the company to watch. Though Lattice's numbers are hardly impressive, they are positioned far better than the other programmable logic vendors to participate in the budding IoT market thru their iCE offerings. This is the only market segment that shows any potential for genuine growth in the near and medium term and is likely Lattice's last chance to break out and legitimately challenge the current segment players for leadership.
If any company is in the right spot at the moment to actually take advantage of the industry's current circumstances, it would be Lattice. It is incumbent on them to recognize the opportunity and seize it. Now is indeed the time for boldness and fierce resolution, as Lattice has nothing to lose. The executive team in Hillsboro would do well to embrace the ideals of Alexander Graham Bell in both thought and spirit, creating new products & offerings and energetically exploiting their leadership in power & cost in pursuit of opportunities in the IoT.
Next week's editorial will explore a new group of companies who hearken back to an age where men dreamed great things and the stuff of heroic sagas were born. :-)