Bhat's Notes

Coverage of “Is Electronics the next trillion dollar in Indian exports by 2030” key note by Sridhar Vembu at FICCI DT7 on 7th July 2023

Dhoti-clad CEO of Zoho, an Indian business magnate unassumingly walked into the session and made mental notes of the speaker’s lines. Even his assistant Krishnan was Dhoti-clad. He doesn’t run his billion dollar in annual revenues from a metro. Tenkasi, a district town in southern Tamil Nadu is where he runs his business from.

Vembu speaks with intent, clarity and from the bottom of his own experience. He is impromptu and didn’t mince his words. There are two role model profitable unicorns in India, that stay cleared from the questionable practices of some of the other unicorns. Zerodha is one and Vembu’s Zoho is the other.

Sridhar Vembu: Taiwan, China, South Korea and Japan from east Asia are our competition. These countries executed with absolute perfection the semiconductor strategy.

USA is trying to catch up with these countries at the University level. It is going to fail and even our country is doing the same and outcome would be no different. Valuing degrees over practical industry efforts won’t help. In our industry, it matters a lot. An 18 year programmer from USA is writing cutting edge compilers for us in Tenkasi.

These East Asian countries didn’t go with PhDs first but much later. Microsoft Research got many PhD AI scientists but Microsoft ended up funding them OpenAI and licensing their technology. This is a slap for them. This is repeatedly true in Silicon Valley and in our industry in India. East Asians have gone through execution experience, not academic excellence. Universities can’t build a commercial product through fine-tuning. 90% of the work this industry needs is not invention but the fine tuning. Getting it right in fab is $1 billion difference per fab per year. Most of the patterns are of fine tuning. Fine tuning the laser to reduce the inefficiencies is where the secret sauce is. It takes time to master all of it.

A relative of mine I keep analysing his musical mastery but copying the way he plays or just or just watching it won’t make me expert. I have to go play and make mistakes and then I learn. Same thing with anything complex we do. We overvalue the formalised knowledge, we de-value the experiential or contextual knowledge. Asians from these countries have mastered through experiential contextual knowledge, not academic knowledge. Doing and learning from it is where the requirement.m lies. Our cloud at Zoho is based on this.

There may be 150 steps involved in making Silicon a chip. Same with layers in the cloud. We need intricate awareness. Profit comes from the mastery of those layers by marking them repeatedly more efficient. Where are the costs going? Intricate knowledge is what defines efficiency. FAANG companies make $500 B in profits per year. These profits are making their concentration stronger. 80% of the profits of smart phones are going to Apple. So is the case of Operating Systems. Contextual experiential knowledge matters a lot.

Cost of living in the valley is so high. How can that region compete? Our iPhone includes $50 tax for the landlords of the Silicon Valley and their living, figuratively. Such high cost of operation let’s you compete as there is no close competition in the world. Google and Apple accounted for the entire mobile ecosystem. Entire operating system industry is between Microsoft and Google. Two companies are just controlling it.

Setting up more universities means making more talent in the valley. Each brain we export is very costly about many jobs that we have not created. It is important we keep the talent here. Those East Asian countries have done it. Their GDP per capita reflects it. Forgot the fab, look at the nail clipper we use in India. It is made in China or Korea. Small country with small population is shipping such every day products to India. Why are we not doing those value steels that require industrial knowledge and not the metallurgical departments in the universities. It is Rs70 product. A nail cutter used in a remote Indian village is made in South Korea product!

AI LLM also has the similar deep layered models. The AI infrastructure is expensive and the battle gets the expensive. If you are not in the high margin industry, you can’t afford those high capital fab or LLM industries. A LLM model is easily at $100 million . NVidia chips are now extraordinarily expensive with 85% gross margin, unheard of in the semiconductor industry. We need those chips and there is a 3 month waiting list and you wait as customer. We are use to waiting for our customers paying after three months here in India. At NVIDIA, You don’t negotiate with them. That’s where India’s challenge is.

We have to ask what are we doing here in our Indian colleges. We had a glorious past, do we have even future? Can India has future if it keeps exporting talent all the time. We must take pride in talent retention and talent promotion. This is a recursive problem. Great challenges attract great talent and great talent attracts great challenges. Attract the US talent. They come to India if there is cutting edge R&D part in India. Industrial R&D is must and is boring in a presentation but it is crucial. 1000 steps. We filed 8 patents last 6 months and I can’t explain them easily. That is what we need. This industrial R&D could be done in 1-2 crores. Not that expensive.

In liberalised era, start thinking about import substitution. Jeff Bezos said your margin is my opportunity. If NVIDIA gets 80% margin, where is my opportunity in it? Look at the imports that too high-margin ones and have 5 member R&D industrial teams to do mastery. Attract and retail talent through cutting edge challenge. Gross margin goes up means more salaries paid to the talent and the more you attract. This is repeat cycle. This path will take us to the trillion dollar.

Unlike Taiwan, we have demand here. We have to master. We are doing that in medical devices. India’s first MRI got medical grade regulatory approval. This is 4-5 year of effort by 40 people through industrial R&D. Meta idea is the same.

If the market is there, restart the project and try again. In fab, Money goes to buildings and equipment and not in R&D of the fab.

Thee is lot of talent and look beyond the universities. Looking at universities via a formalist trap and avoid it.