Currently, India is a minor player, but not insignificant. The government has made it a priority to change this via the India Semiconductor Mission
The semiconductor design process is linear, and a single mistake can cost millions of dollars, if not more
Given the benefit of a vast, diverse market that throws multiple requirements and a considerable talent pool, India is a great place to apply agile methods to chip design
The size of the global semiconductor industry is more than $600 Bn and is estimated to cross a trillion dollars within this decade, as per a McKinsey report. This includes almost everything around us – consumer electronics, automotive electronics, computing devices, communication devices, industrial electronics, servers, and more.
The semiconductor industry started in the USA in the 1950s, expanding to Japan in the 1970s and then to Europe, South Korea, Taiwan, China, and other Southeast Asian countries. It is interesting to note that every country has a slightly different position in the global semiconductor space.
For example, Taiwan is home to TSMC’s mammoth fabs. Japan and South Korea started with DRAM manufacturing and have moved up the ladder with Sony (Japan) and Samsung (South Korea) leading the innovation. The Netherlands is home to ASML, which makes machines for semiconductor fabs. Arm is based out of the UK.
Currently, India is a minor player, but not insignificant. The government has made it a priority to change this via the India Semiconductor Mission – which is aimed at attracting semiconductor manufacturing to India. As of now, one of India’s strengths is its design talent. Every major semiconductor company has a design unit in India, including Texas Instruments, Intel and Samsung.
The Case For Agile Chip Development
We make the case that India is excellently positioned to be a leader in the semiconductor design space by bringing agility to chip design. From the outside, agile semiconductor design sounds paradoxical, the semiconductor design process is linear, and a single mistake can cost millions of dollars, if not more.
A new chip can take two to five years to bring from concept to production. Most practitioners would shudder at the thought of anything faster.
However, today’s fast-paced market demands mean that it is necessary to speed up chip design by introducing agility into the process. While challenging, this requirement gives a country like India an exciting opportunity to exploit.
The Democratisation Of Usage
Over a hundred years back, Henry Ford thundered, “They can have the Model T in any colour they want as long as it’s black.” The market has moved forward significantly since then – products are used differently by diverse customers.
Imagine a simple physical product like a fingerprint scanner or a CCTV camera. In a country like India, customers come in all sizes. Some customers are savvy; others are new to tech. The conditions at the installation locations differ – hot/cold, humid/dry, clean/dirty, sunny/shady, etc. Some can afford more than others.
Yet, there are only a few models of scanners and cameras out there – with almost no customisation options.
Suppose there were tens of models of each consumer electronic item, and then each industrial electronics item, and each computing item, and so on – that would create the kind of democratised usage that the market demands today. And it would mean thousands for different requirements for chips.
The Waterfall Model
Just as water finds it easy to flow downhill but can’t flow uphill, the waterfall model describes a development process where processes move forward and not backward (or iterative).
Semiconductor design involves several stages:
- High-level design
- The “Netlist” – a translation of the high-level design into manufacturable devices
- Layout of the Netlist on the silicon (“floor planning”)
- Placing and routing wires
- Final layout and signoff
Each step has to be done with extreme diligence and patience. Plus, it costs money and takes time – partly because of each step’s verification and validation processes.
The problem with the waterfall process is that a backtrace is impossible – in case of an error and a change in requirement. More than that, the process forces a long development cycle, exploding the time to market to anywhere between two to five years. This makes innovation slow.
If an engineer has a spark and comes up with a new way to do something, he/she has to wait for the next design cycle – which could take years. Further, when an updated product (or sub-product) finally hits the market, the customer’s needs might have changed!
Compare this with software, where our products are continuously updated via the internet, and never the “older version.” For tools like Gmail or Google Photos, we don’t even ask, “Which version is this?” Old-timers will remember – it wasn’t always so. In the 1990s, we purchased Windows 95, 98, and other software on CDs. But that’s a thing of the past.
The Agile Method
The agile method is a set of practices popularised in the last two decades, mostly in the software engineering universe. The Agile Manifesto, 2001, states four values as core to the agile method. They are:
- Individuals and interactions over processes and tools
- Working software over comprehensive documentation
- Customer collaboration over contract negotiation
- Responding to change by following a plan
One of the main outcomes of this method is that a software product is designed in modules that can be improved upon individually. These individual improvements are continuously delivered to the customer. This allows the TAT (turnaround time) from customer feedback to new feature development (or bug fix) to be very low.
In chip-verse, agility means running all design stages in parallel, feeding each off from the other. This is definitely possible in modular design where each module can be developed independently. Some components may even follow a hardware-software co-design methodology, allowing continuous upgrades, for example, in the camera unit of a mobile phone.
One of the most significant differences between agile and waterfall is that at every stage of the agile lifecycle, there is a strong emphasis on the user – what are their requirements, and what is their experience in using the product? Is the design conforming or adding to the final experience and meeting the final expectation at every stage? We need many real users to test our assumptions for this user-focused approach to succeed.
The Indian Advantage
Fortunately, we have a wide and diverse set of users. For semiconductor designers, the “users” are the engineers and programmers who buy chips and build devices for the “end-user”. And since there are infinite cohorts of end-users, thousands of engineering teams demand different kinds of chips in all kinds of sectors – defence, space, automobile, electronics, budget electronics, etc.
Not only does our country generate thousands of use cases, but it also generates fabulous semiconductor talent. Some of the brightest, most experienced hardware engineers worldwide are Indian, and they studied in Indian schools and universities.
Unfortunately, most of these engineers work in semiconductor services or as part of smaller teams in large MNCs. So, “product thinking” is low. But this is changing.
In Conclusion
Given the benefit of a vast, diverse market that throws multiple requirements and a considerable talent pool, India is a great place to apply agile methods to chip design. This would enable us to leapfrog ahead in the chip-verse.
Engineers would have to think differently to achieve this. Firstly, they have to design with the customer in mind. They need to study the customer’s needs in depth. Secondly, they must “own” the requirement/specification instead of bargaining with the customer or asking them to settle for a less optimal product.
Engineers also need to anticipate future requirements and the changing technology landscape. With all of this in mind, they need to design with a certain flexibility, i.e. agility.
