Friday, July 4, 2014

High Tech “White Elephants” – The Tragic Mistake of India’s Silicon Foundries

(Note: this article preempts the "Value Proposition and the Custom Chip Market" editorial series for this week. The next installment of the series will be focused on the SoC segment of the custom logic market and will be posted the following week, on July 11.)

The best laid schemes o' mice and men gang aft a-gley; and leave us naught but grief and pain for promised joy - Robert Burns

At the end of February 2014, the government of Manmohan Singh announced a program to attack the electronics portion of India's mushrooming foreign trade deficit through a domestic industrial development project. In partnership with two private cooperatives, New Delhi will sponsor the construction of two semiconductor wafer foundries. These foundries (commonly known in the chip industry as 'fabs') are intended to supply the chips for local component sourcing required by Indian trade regulations pertaining to electronic systems imports. The fabs would also serve as the centerpieces for further growth and development of India's expanding technology sector.

There are a variety of arguments for and against this initiative. This editorial will examine the proposal in detail, including factors such as financial and regulatory elements, economic impact, the role and effect of government policy on industrial development, technology issues, social implications, environmental effects and the significance to international trade. The treatise will conclusively prove that the entire effort is utterly puerile and propose an alternative course of action that promises much higher returns on investment to the nation.

Part 1 - Particulars

There are two private sector consortia participating in the program. One consists of Jaiprakash Associates, IBM and TowerJazz, a small independent silicon foundry based in Israel with additional fabs in the USA and Japan. This group intends to build an INR 334B ($5.9B USD) plant near New Delhi supporting process node geometries down to 28nm and volumes of 40,000 300mm wafers per month. 

The second is composed of HSMC (Hindustan Semiconductor Manufacturing Company) Technologies India, Silterra (an outdated independent Malaysian fab) and ST Micro, an $8B USD european semiconductor manufacturer with an assortment of fabs in France and Italy, serving the networking, wireless, automotive and consumer electronics markets with microcontrollers, discrete power devices and a variety of standard products. Though it will support the same manufacturing volumes as the first consortium's facility, this plant will implement a more advanced 22nm process. Based in Prantij, Western Gujarat, the facility is budgeted at INR 290B, or $4.93B USD.

New Delhi's involvement in the deal is massive. Each group will receive an interest free INR 51.42B loan and will have 25% of their capital requirements covered by the central government. There are additional tax incentives and a customs duty exemption for certain capital items (this is significant, as will be explained later in the editorial.) Not counting these tax subsidies, the direct investment support from New Delhi sums to INR259B, or $4.4B USD - approximately 40% of the total projected cost.

Part 2 - The Case For

Benefits expected to accrue from this program include the following:
1. Jobs - the project forecasts 22,000 new jobs created directly by the two fabs, with an additional 100,000 spawned from the rise of businesses and services supporting the new manufacturing plants.
2. Accelerating India's growth in the Technology industry to advance the nation further up the economic value curve. 
3. Fostering technology development that is India-centric.
4. The creation of centers of high technology enterprise that will spawn the development of a supporting ecosystem of companies, products and services.
5. Reducing India's dependence on electronic imports - already $31B USD in 2013 and expected to grow to $400B in 2020 - thru regulatory requirements for local content.
6. Create local sourcing for the electronics needs of India's rapidly modernizing armed forces. This is being viewed quite logically as a strategic necessity. India will need to fill a security role as a growing world power at least regionally and, quite possibly, farther afield in order to to safeguard the nation's interests and protect it against any potentially hegemonistic intentions of other BRICS nations.

On the surface, the program seems to have no faults and appears replete with beneficial results to the nation over both the short and long term. Yet as the old proverb says, "all that glitters is not gold." The advertised gains of the initiative simply do not hold up to serious scrutiny.

Part 3 - The Case Against

Forecasting by bureaucrats tends to be used for anxiety relief rather than for adequate policy making. - Nassim Nicholas Taleb

Despite making great strides over the last two decades, the people of India as yet have a long road in front of them to reach first world living standards. The current per capita income of $4300 USD is still an order of magnitude short of such a goal.

Using this number as a comparative measure, the government expenditure allocated for this program is unjustifiable as an employment booster. Generating 22,000 direct jobs from $4.4B USD of investment equates to $200,000 USD (INR 11.76M) per worker. Adding in the indirect job projection results in a cost of $36,000 USD (INR 2.12M) per worker. A government that was genuinely concerned about expanding the national labor force would not waste the Indian taxpayer's money so flagrantly.

However, the employment numbers themselves are dubious. As a comparative data point, GlobalFoundries Fab 8 in Malta, New York State, USA employs 5500 people and produces 60,000 300mm wafers per month in a 28nm process. Modern microelectronics fabs are extraordinary complex manufacturing facilities with a very high degree of automation. As such, they are not labor-intensive operations. The claim that the two planned Indian fabs producing 80,000 wafers per month will directly employ 22,000 people is specious.

The program is intended to further stimulate the growth of India's technology sector. If this is so, then the locations picked for establishing the fabs are rather surprising. The overwhelming majority of semiconductor design work is being done by chip firms in Bangalore and Hyderabad to the south, with major growth in other southern cities such as Chennai and Pune. There is a critical link between semiconductor design and manufacturing which provides significant competitive advantage, as proven by technology heavyweights such as Intel, IBM and Samsung. Chip fabs do indeed require significant water supplies, but the centers of design in the south all have adequate water sources. This makes the choice of location for the fabs not only counterproductive, but suspect as well.

While we're on the topic of location, there are other considerations regarding the establishment of a semiconductor foundry. Fabs don't just need water - they require highly purified and de-ionized water. An advanced water treatment facility will have to be built for each fab. 

That water will have to be recycled as well, because modern chip fabs are notorious polluters. The pollutants they produce are not your typical soot, sewage and garbage, either. Semiconductor processing employs a plethora of highly reactive and even toxic substances - gasses, acids and heavy metals - that can cause cancer, birth defects and a variety of illnesses and injuries which are difficult to treat. Without water recycling, careful pollution controls and safeguards, the fabs run the risk of contaminating the local water table and destroying farmland.

One of the major goals of the program is to reduce the electronics portion of India's current account deficit in international trade by requiring greater local content. The very endeavor of building these fabs runs somewhat counter to that goal. 75%-80% of a fab's construction costs are for equipment. This equipment is essentially space age in its complexity & sophistication and includes items ranging from $500,000 USD to $50M USD in cost. All of that equipment is made overseas - mostly in America and Japan.

The fab equipment cost can be understandably dismissed as something of a one-time expense. However, the program's current account problems do not end there. In order to make an appreciable dent in the electronics current account deficit, fully 100% of the content of those systems will have to be locally sourced. This means the two fabs will have to produce all the electronic components for these systems - DRAM, SRAM, flash memory, analog, mixed signal, SoC, graphics, audio, video, voice, discretes, RF, MCU, optical, DSP, CPU, TTL and so forth. It would take a miracle for the fabs to successfully produce all of these different components.

Let us assume, however, that miracles do indeed occur on occasion and that the fabs can meet the very broad range of process requirements to successfully produce all the above mentioned devices. At 80,000 wafer starts per month combined, the new India fabs would represent approximately 1.5% of the world's total semiconductor foundry capacity. If we pretend that the fabs were already in full operation today, they would be able to produce 1.5% of the world's semiconductors - a $315B USD market in 2013. This would amount to about $4.7B USD of chips. Thus, last year's $31B USD of electronic imports would drop to about $26B USD. Nonetheless, the current account deficit reduction would be more or less immaterial in 2020, as it would change from $400B to roughly $395B USD. In order to keep pace with growing imports, the fab capacity would have to increase by a factor of 13 over the next 5-6 years, and the investment cost would correspondingly multiply.

This brings us to the central issue of the entire discussion. Exactly how useful would these fabs be to India's technology sector?

Independent chip foundries are currently evolving into substantially different entities than the straightforward manufacturing plants with basic business models that they were just ten years ago. Fabs are increasingly moving into embedded IP development and engineering services in order to enhance their value to semiconductor design companies. Simply providing a production line is becoming less useful by the day in this industry. 

As previously mentioned, the foundries will support 22nm and 28nm lithographies. The project will break ground for both facilities in early 2015 and is planned to reach full production capability sometime in 2017 or 2018. However, the worldwide semiconductor industry has reached a crisis point over this past year. The traditional Moore's Law improvements to the three P's (price, performance and power) that were normally expected with each move into the next smaller process node began to show diminishing returns as the industry entered deep submicron feature sizes at the beginning of the millenium. Once the 28nm node was reached, further shrinkages to 21nm and 14nm actually resulted in negative returns, as one parameter seemed to irredeemably worsen when the other two were optimized in a given chip design. 

This has triggered a ferocious debate in the chip industry over whether Moore's Law is dead or not. A majority of voices believe silicon has reached the end of its effective life somewhere around 14nm-28nm, while a few voices proclaim that with a final monumental process engineering effort they may be able to extend the law down to 10nm and, perhaps, even 5nm by 2020. Furthermore, the chip industry appears poised to move to 450mm wafer sizes for increased yield and manufacturing efficiency by no later than 2016. 

Whatever the case may be regarding whether silicon has reached the dreaded Last Node, the issue itself, combined with the upcoming transition to larger wafers, together cast a long shadow over the entire India fab project. No matter how you slice it, the reality of the matter is that the fabs will likely be completed at a time in the future where, in comparison to the rest of the industry, they will be supporting mature semiconductor technologies on a relatively inefficient manufacturing line. Spending $4.4B USD of taxpayer money on a program that, upon its completion, is very likely to be perceived by the rest of the technology sector as a White Elephant forces any reasonable observer to question the fundamental point of the initiative.

The danger that this may be a waste of public funds is magnified by the very presence of the central government in the project. There isn't a nation on the face of the earth that hasn't seen the ugly consequences of politicians disbursing significant sums of the government purse and attracting a flood of highly self-serving chrony capitalists like pigs to a full trough. It is a normal experience in any country to witness anything a central government touches become corrupt and grossly inefficient thru graft, chrony capitalism and nepotism. Initiatives allegedly begun for the public good become transformed into a circus of fraud where private interests fatten up on the taxpayer's dime with no benefit accruing to the public and the ill-gotten gains flow to political figures and their corporate sponsors.

The restrictive labor laws and regulatory climate of India, along with the bureaucracy's legendary inefficiency, further erode confidence in this project's ultimate success and viability. Such programs easily fall prey to the stealthy corruption of nepotism, so that incompetent and unqualified lackeys and relatives get appointed to prestigious and lucrative supervisory & managerial positions. In this way, large public projects become bloated, corrupt, incompetent and ineffective. Stated differently: if having large local fabs in India was such a good idea, each of the two private consortia could either attract more partners or combine forces with each other and do this on their own. There should be no need whatsoever for financial support from New Delhi.

Part 4 - The Way Forward

Bureaucracy is ever desirous of spreading its influence and its power. You cannot extend the mastery of the government over the daily working life of a people without at the same time making it the master of the people's souls and thoughts. - Herbert Hoover

This meddlesome intrusion of the central government into an already strong and vibrant portion of the private sector is, in fact, highly contraindicated. It is presumptuous of New Delhi to assume that exerting an overt influence on India's high technology sector is needed and beneficial. On the contrary - it's presence is distinctly detrimental and unwelcome.

Indian universities have produced a tsunami of technical talent that has spread itself across the globe and is busy employing its talents in building a better tomorrow for mankind. It is key to understand and accept the specific fact that a great deal of India's talent has felt it necessary to leave home and hearth to exercise their talents and energies. They devote their skills and ambitions to success in foreign lands because their pioneering spirits are stifled by an inept, corrupt and fossilized bureaucracy emanating out of New Delhi. The organs of the state impede Indian entrepreneurs from inventing the future within India's borders and effectively force most of these adventurers to seek their fortune elsewhere.

One does not plan and then try to make circumstances fit those plans. One tries to make plans fit the circumstances. - General George S. Patton

New Delhi's policy initiative on this matter lacks foresight, technical understanding, market awareness and, ultimately, ambition. The nation would be served far better by pursuing an alternate course of action.

First and foremost, it is essential for New Delhi to get out of the way. India naturally produces ingenuity in astonishing abundance. The very nature of the country itself produces inventive, flexible, curious, entrepreneurial and open-minded people. It is a country with not one, but two common languages - Hindi and English. The states are as different from each other in history, viewpoints, language and culture as are the nations of the European Union, but with a much stronger spirit of cooperation and coexistence. In that respect, India resembles the United States a bit more than it does Europe.

Thus, the proper role of the central government is not one of providing direction, but of staying out of the way. Stated differently - New Delhi should not lead, but serve. India's technology entrepreneurs know what needs to be done and can form ad hoc or more permanent electronics-oriented industrial associations to organize and drive their common causes.

Where government needs to show initiative is in making every effort to keep out of the way of India's technology sector. This includes regulatory and labor law reform to minimize the negative presence of government bureaucracy and its singular mastery of generating oceans of red tape. By letting high tech firms manage their labor and business affairs with maximum flexibility, they will be free to demonstrate the same sort of dynamism which allowed Silicon Valley to blossom and reshape the world we live in. A more hospitable business environment will also greatly encourage the repatriation of Indian technology workers, whose seasoned managerial and technical skills will prove invaluable.

If government is to play a role in advancing High Technology in India, it must do so in a way that absolutely minimizes the ability of politicians to tamper with the effort. There are examples of extremely successful technology developments done in partnership with government that followed this principle - NASA's Moon program, the Manhattan Project and various DARPA programs (ARPAnet, the SR-71 and the "Have Blue" and "Senior Trend" projects) stand out as exemplary success stories in this regard.

One cannot help but notice that all three of the above mentioned examples were heavily influenced by or completely maintained under the suzerainty of the military. They involved close collaboration between technical and scientific experts in the military and academia to define a technology goal and a program to achieve it which was then contracted out to the private sector. The development effort was then supported and managed largely by the military, who brought to the program a sense of duty & mission along with a resolute commitment to the long term strategic interests of the nation, rejecting any petty, short term political concerns.

The question then becomes one of selecting what technology such an Indian "Manhattan Project" should concern itself. The answer becomes obvious when one observes what the leading strategic technology interests are of Apple, Google, Samsung, IBM and China's Ministry of Science and Technology: graphene, silicon carbide, carbon nanotubes and quantum computing. Rather than trying to chase after the accomplishments of others in silicon (a technology which may be on its last legs), India should be gathering its resources and talents to pursue new, untapped technology that will launch it into a position of leadership. The above mentioned technologies are also highly complementary and hold the kind of potential for the future that will make the silicon-driven changes of the last four decades look trivial by comparison.

In summary: India should not concern itself with following in the footsteps of others, but leaping ahead of them. The talent and initiative are there at home and in the technology 'diaspora' of citizens working abroad. It boils down to a matter of vision & will, while also taking steps to ensure that politicians don't muck it all up.