The numbers Brief Episode map Timeline Takeaways Charts Quotes Clips Glossary Go next Transcript

Episode 159 · Deep tech · 41 min

Supply security is the product

India imports the chemistry inside its fertiliser and $42 billion of critical semiconductors a year, and sits on a US list that caps advanced GPU shipments at 4,000 annually while a single Indian data centre needs 12,000. The founders of BioPrime Agri and Morphing Machines argue the answer is not to copy the American playbook but to design inputs for Indian soil and Indian silicon, then find the first Indian customer willing to buy them.

Deep tech & hardware Supply chain & agri Regulation & policy

Dr Renuka Diwan and Deepak Shapeti

Founders, BioPrime Agri and Morphing Machines · with Vishal Krishna

41:27

Said in this episode

▶ 12:35

4,000/yr

Advanced GPUs India may import

Deepak says India's category-3 placement caps advanced GPU shipments at about 4,000 a year, one category above Russia and China, while a single Indian data centre typically needs 12,000.

▶ 18:44

$42B

Critical semiconductors imported a year

Excludes defence, which he calls a different ballgame; India captures only about 10–20% of the margin on that flow.

▶ 9:04

20%

Fertiliser use efficiency on Indian farms

The other 80% locks into the soil or washes into rivers, on fertiliser raw material India imports and subsidises from taxpayer money.

▶ 6:16

1.7 lakh

Farmers BioPrime has reached

About 80% of them small and medium holders; she immediately calls the number not even a drop against India's farming population.

▶ 37:08

~40%

Produce lost between farm and fork

Her blended estimate, higher for fresh fruit and vegetables; the host recalled 35% from his reporting days and an audience member put the global figure above 50%.

▶ 15:17

15 years

India's fabrication gap to Taiwan

He puts the US four or five years behind Taiwan and India about fifteen, dating the drift to a foundry fire in Chandigarh around 1992 and two dormant decades after it.

The brief

The argument in sixty seconds

The claim running through this Ankur Capital panel is that India's deep-science problem is an input problem. Renuka's case for BioPrime Agri is that every agricultural input the country uses (the fertiliser chemistry, the micronutrients, the pesticides) was designed somewhere else and merely manufactured here, for farms nothing like ours; fertiliser use efficiency sits around 20%, so four-fifths of what a farmer buys locks into the soil or washes into a river, subsidised out of the taxpayer's pocket. Deepak's case for Morphing Machines is the silicon mirror image: a $42 billion annual chip import bill on which India captures only 10–20% of the margin, a category-3 listing that caps advanced GPU shipments at 4,000 a year against the 12,000 one data centre needs, and the $20 billion team-and-licence deal he says Nvidia used to absorb Groq without an antitrust review. Both argue the fix is not a better lab but a tighter chain of design, a Tata foundry, domestic packaging and EMS integrators willing to put an Indian part in the bill of materials, plus the discipline neither ecosystem has yet: productisation, and the first sale. The hardest sell, they agree, is domestic belief. Renuka had to talk her own team out of wishing the technology were Australian before she could face a farmer with it. With roughly $25 billion of data-centre capex landing and 90,000 patents filed in 2024, the open question is whether India captures that value or assembles somebody else's.

Worth your time if you are

Chip founders weighing an India-first beachhead

Agri-input and biotech founders scaling out of a lab

Policy people designing India's deep-tech schemes

Investors sizing India's semiconductor value capture

Students trying to fund a first tape-out

Episode map

Where the conversation travels

Every block is a chapter, coloured by what it's about. Click any of it to jump straight to that minute on YouTube.

Deep tech & hardwareSupply chain & agriRegulation & policyIndia macroManufacturingFounder journey

01Cold open: from strangers to an ecosystem 0:00 Vishal opens with a memory of 1990s Indian scientists who believed knowledge and business could never coexist and filed their IP abroad, against 90,000 patents filed in India in 2024. The case is for scaling deep science the Indian way, not the American way. 02Papers, patents, and nothing beyond 1:37 Renuka describes founding BioPrime out of frustration that a career of hardcore lab research ends in papers or a patent count, with nothing reaching a farm or a food system. 03A grant, a field visit, and a gap 3:36 A Department of Biotechnology proof-of-concept grant let lab scientists test an idea, and the field visits that followed found farmers and even input retailers who had never heard of solutions that looked obvious from the science side, because Indian agri inputs have always been designed outside India. 041.7 lakh farmers and a 20% efficiency 6:16 BioPrime has reached about 1.7 lakh farmers, roughly 80% of them small and medium holders, going after climate-hit yields, soil as the farmer's only inheritable asset, and a fertiliser use efficiency of about 20% that leaves the rest locked in soil or in the rivers. 05One processor, and a 4,000-GPU ceiling 10:18 Deepak explains a ground-up architecture that runs CPU-like and GPU-like work on a single processor, then the geopolitics around it: Groq's $20 billion absorption by Nvidia, and India's category-3 placement capping advanced GPU imports at 4,000 a year. 06The foundry that burned, and 15 lost years 12:50 A single Indian data centre needs about 12,000 high-performance GPUs, and the sovereignty argument runs back to a foundry fire in Chandigarh around 1992 that he says began two dormant decades, leaving India roughly 15 years behind Taiwan in fabrication. 07Consolidation, and the first-customer problem 15:29 Post-COVID the India Semiconductor Mission mobilised the ecosystem, but designing a chip does not sell it. The primary problem is the first customer, and roughly $25 billion of domestic data-centre investment plus TCS's own $8 billion is the captive demand he wants pointed at Indian silicon. 08Two kinds of chips: $2 and $10,000 18:10 One class is high-volume, low-value parts for meters, fan motors and LED controllers that China dominates; the other is the $5,000–$10,000 low-volume, high-value class Morphing Machines targets for cybersecurity, 6G and physical AI, set against a $42 billion import bill on which India keeps only 10–20% of the margin. 09Agri 3.0 is exports and aggregation 20:56 Renuka argues no single system works across Indian agro-climatic zones, so the organising force is exports: exotics grown around Bengaluru, farmer groups formed at a scale that makes logistics work. 10Great businesses that aren't venture businesses 24:03 A Maharashtra collective built a fig export business to the Middle East and then juice and three or four more products, profitable and replicable and probably not venture-fundable, while the real capital gap sits past early stage, where at-scale manufacturing has to be built. 11Owning IP instead of selling services 25:55 A biotechnology manufacturing policy from the Department of Biotechnology and new agri processing funds mark a government thesis she calls commendable, that India's next chapter should be IP-owned biotech rather than another services story, though she wishes the speed were higher. 12Rice, soy, and the food weapon 27:44 Food security becomes explicitly geopolitical: the rice narratives during the Ukraine war, dumping accusations, soy kept out of India, and a New Zealand trade deal negotiated without opening the dairy sector. 13Productisation and the colonial mindset 29:00 Both founders name the same gap, that the idea is the easy part and Indian deep tech has not learned to productise, and Renuka adds the harder one: convincing her own team, and then the farmer, that made-in-India technology is worth as much as an Australian label. 14A student asks how to tape out a chip 32:55 Asked whether to tape out first or productise first, Deepak sends the student to the government's per-student tape-out programme and the Chips-to-Startups scheme, worth up to ₹3–4 crore with a free tool chain and validated RISC-V cores, then to find a differentiating slot in the ecosystem. 15The 40% that never reaches a plate 36:20 An audience question on farm-to-fork spoilage draws a blended loss figure of about 40%, higher for fresh produce, and a survey of sachets, coatings and ambient-temperature chambers before Renuka says only strong FPOs, processing near the farm and better shelf life at the input stage will move it. 16The report card: 100, and a seven 39:53 On government accountability and transparency, Renuka rates its importance at 100 and calls her sector's jarring shift from unregulated to regulated a process that could have been smoother; Deepak scores the last four or five years seven to eight out of ten, with excellent intent held back by execution.

Timeline

The arc, briefly

Two input problems, silicon and fertiliser, dated across four decades of Indian policy.

1987-90India is close on chips. Deepak puts the country a couple of years behind the American side. Not leading, but within reach of it.

1992The first foundry burns in Chandigarh. Some say foreign interests did it. Either way the sector went dormant from there until roughly 2010 to 2015, and India is now about 15 years behind Taiwan in fabrication. He puts the US four or five years behind Taiwan.

1992-2002A decade among scientists who filed somewhere else. Vishal spent it with Indian researchers who believed knowledge and business could never coexist. They went abroad and filed their IP there, and he was told he was not a patriot for asking why.

2021India doubles down after the chip shortage. Post-COVID, having seen how critical semiconductors were to every part of the economy, the government put structured support behind the India Semiconductor Mission and pulled the ecosystem together.

202490,000 patents filed in India. Vishal's marker that the strangers have become an ecosystem, and that people like Renuka and Deepak stayed to build here rather than leaving.

TodayTwo ceilings, one argument. India sits on the category-3 list, so advanced GPU shipments are capped at about 4,000 a year while a single Indian data centre needs 12,000, and the country imports $42 billion of critical semiconductors while keeping only 10 to 20 per cent of the margin. On the farm side, fertiliser use efficiency runs around 20 per cent, so four-fifths of what a farmer buys locks into the soil or washes into a river. Nvidia took the Groq team and licensed the technology for $20 billion, leaving a shell and no antitrust review.

Takeaways

Ideas to carry out of this hour

Supply security, not novelty, is now the deep-tech thesis

Deepak's argument is that compute has become a chokepoint the way oil once was. India sits on a category-3 export list that restricts advanced GPU shipments to about 4,000 a year, one rung above Russia and China, while a single Indian data centre typically needs 12,000 of them. He reads Nvidia's $20 billion absorption of Groq the same way: rather than risk an antitrust fight, the team was hired and the technology licensed, leaving a shell behind. The lesson he draws is not that India should out-engineer anyone, but that a country which cannot source its own compute can be stonewalled at will.

▶ 12:19 ▶ 12:35 ▶ 13:34

The first customer is the bottleneck, not the design

He is blunt that designing a semiconductor does not mean it sells. The primary problem is the first customer, and Morphing Machines wants that customer to be Indian because the volumes are already here: roughly $25 billion of data-centre compute capacity being built by domestic operators, with TCS alone putting in about $8 billion. What has to change is the bill of materials: design houses, a Tata foundry, domestic packaging and the EMS integrators all choosing each other, which is exactly the tightly knit arrangement he says gave Taiwan an advantage nobody has replicated.

▶ 16:00 ▶ 16:14 ▶ 17:01

India imports $42 billion of chips and keeps a fifth of the value

Excluding defence, India imports around $42 billion of critical semiconductors a year, and on that flow captures only about 10–20% of the margin; the rest stays with the original supplier of the chip or the module. His split of the opportunity is deliberate: high-volume, low-value parts at 50 cents to $2 apiece (electricity meters, BLDC fan motors, LED controllers) where import duties now make Indian competitors viable, and low-volume, high-value products with MRPs of $5,000–$10,000, where a reprogrammable architecture can compete globally. Capturing more of that chain, not winning a node race, is what he thinks produces durable advantage.

▶ 18:16 ▶ 18:44 ▶ 18:57

India's farm inputs were designed for someone else's farms

Renuka's founding provocation is that nobody in India had built an agri-input innovation. The fertilisers, micronutrients, insecticides and pesticides are manufactured here but the chemistry comes from outside, and it has never been asked whether it fits Indian crops, soils and practices. The measurable cost is fertiliser use efficiency of about 20%: four-fifths of what the farmer applies either locks into the soil or washes into rivers, on raw material India imports and subsidises out of taxpayer money. Her fix is to unlock what is already in the soil and get more of it into the grain, so the same input produces both a better margin and more nutritious food.

▶ 5:28 ▶ 5:59 ▶ 9:04

Science is 10% of the company; productisation is the rest

Deepak says science felt like 20% of the journey when he started and now feels closer to 10; Renuka independently puts idea and science at about 10% of what makes a company work. Their shared diagnosis is that Indian deep tech is still in its infancy at productisation, which for him takes in marketing, business development and solutioning, and which cannot be shortcut. His method is unglamorous: pitch customers the way you pitch investors, read the room, take the hits and stand back up until the weakness is fixed.

▶ 11:34 ▶ 29:00 ▶ 30:16

The hardest sale is telling Indians it was made in India

Renuka's biggest obstacle was not the science or the capital but belief. Everything BioPrime uses is sourced in India except one plant material from Australia, and her own team came back arguing the product would sell better if they simply called it Australian technology, because people would lap it up. She had to run mindset conditioning inside the company before she could stand in front of a farmer who also did not believe it. Abroad the question flips to disbelief that so little money could produce the result, with an implication that corners were cut; her only answer is relentless work and a narrative of trust and performance built over time.

▶ 30:32 ▶ 30:46 ▶ 31:36

Post-harvest loss is a systems problem, not a product problem

Asked where spoilage sits on the food-security ledger, Renuka puts blended farm-to-fork losses at about 40%, and higher for short-shelf-life fruit and vegetables. An audience member with international experience pushes the global figure past 50%. Startups are attacking it from every angle: sachets that suppress the ripening enzyme, coatings, ambient-temperature storage chambers, and input-side work like hers that lifts shelf life so produce is pre-booked and lifted at the farm gate for two or three rupees over market price. Her verdict is that none of it closes the gap alone; it needs strong farmer producer organisations, processing sited near the growers, and shelf-life gains at every level at once.

▶ 37:08 ▶ 38:18 ▶ 39:22

The numbers, drawn

What the episode measures

Every figure below was said on air, with timestamps included and caveats kept.

The GPU ceiling versus one data centre

advanced GPUs

Both figures as stated on air: India placed on the US category-3 list, one rung above Russia and China, against the roughly 12,000 high-performance GPUs a typical Indian data centre requires.▶ 12:35

Where the fertiliser actually goes

% of fertiliser applied

Taken up by the crop20%

Locked in soil or washed away80%

Renuka puts fertiliser use efficiency in India at about 20%; the 80% share is the balance she describes as locking up in the soil or washing into river bodies.▶ 9:04

How far behind Taiwan, in fabrication

years behind Taiwan

As stated on air: the US 'four or five years' behind Taiwan (upper bound shown), India about fifteen, after the Chandigarh foundry fire he dates to 1992 and the dormant years to roughly 2010–2015.▶ 15:17

Five links that have to pick each other

what turns a chip design into supply security

Deepak's answer to the first-customer problem. He wants the buyer to be Indian because the volumes already exist, and describes the tightly knit version of this chain as the advantage Taiwan has that nobody has copied.▶ 17:01

Worth keeping

Lines that stay

Our systems, our farmers, our crops, the way we farm is very different from anywhere else in the world. So why do we have to adapt? Why can't we have technologies developed specifically for our farmers, for our ecosystems?

— Renuka ▶ 5:59

India has been placed on the category 3 list for GPUs, so any advanced GPU shipments to India are restricted to 4,000 a year. We're just one category above Russia and China.

— Deepak ▶ 12:19

The idea is the easy part. Getting that first sale is incredibly critical — and we still don't understand, as founders, how to productise something very well.

— Deepak ▶ 29:00

Our team came back and said it would be so much better if we only said it was Australian technology — people would just lap it up. I had to start mindset conditioning from there, and then go to the farmer and explain that no, this is made-in-India technology.

— Renuka ▶ 30:46

There's no quick answer, no quick fix. The only way to do this is relentless work, constantly proving ourselves, and building that narrative of trust and performance.

— Renuka ▶ 32:08

Clips that travel

Short on time? Start here

Agri-input founders and food-system investors

Twenty per cent of the fertiliser actually works

The import-and-subsidy loop laid bare: 80% of applied fertiliser lost to soil or rivers, and why the fix is efficiency rather than volume.

7:52 → 10:18 · 2 min ▶ Watch clip

Chip founders weighing an India-first beachhead

One processor, and a 4,000-GPU ceiling

A CPU-and-GPU-in-one architecture, the Groq shadow acquisition, the category-3 cap, and the 1992 foundry fire that set India back fifteen years.

10:48 → 15:29 · 5 min ▶ Watch clip

Investors sizing India's semiconductor value capture

Who buys the first Indian chip

The first-customer problem, $25 billion of domestic data-centre capex, and why a $42 billion import bill leaves India only a fifth of the margin.

15:29 → 20:56 · 5 min ▶ Watch clip

Deep-tech founders stuck between lab and market

Made in India, and nobody believes you

Both founders on productisation as the real gap, then the team that wanted to call it Australian technology, and the disbelief that greets a cheap Indian result abroad.

29:00 → 32:55 · 4 min ▶ Watch clip

Anyone working on cold chain or post-harvest loss

The 40% that never reaches a plate

A candid tally of farm-to-fork spoilage, a tour of the startups attacking it, and the admission that none of it is close to solving the problem.

36:29 → 39:53 · 3 min ▶ Watch clip

Glossary

The jargon, unpacked

Fertiliser use efficiency

The share of applied fertiliser a crop actually absorbs, about 20% in India, with the balance locked into the soil or washed into rivers.

Tape-out

The moment a finished chip design is handed to a foundry for fabrication; the government programme cited on air funds one free tape-out per student team.

C2S (Chips to Startups)

An Indian government scheme that gives academic and early teams up to ₹3–4 crore, a free EDA tool chain and pre-validated RISC-V cores to reach a first tape-out.

AI inferencing

Running a trained model to produce answers, as distinct from training it. This is the market Groq led, and the one Morphing Machines can serve but deliberately does not target.

Category 3 GPU listing

The US export tier India is placed in, capping advanced GPU shipments at roughly 4,000 a year, one rung above Russia and China.

Radiation hardening

The qualification process that lets a chip survive the radiation of space; its cost, against very low volumes, is why space is not Morphing Machines' beachhead.

FPO (Farmer Producer Organisation)

A registered collective of farmers that aggregates produce so logistics, processing and exports become viable at scale, the model Renuka says post-harvest loss actually needs.

Bill of materials

The component list an electronics manufacturer builds a product from; getting an Indian chip onto it is the step Deepak says converts design capability into supply security.

Connections

If this resonated, go here next

EP 17 · Deep tech

Full transcript

The whole conversation, searchable

Auto-generated captions, lightly cleaned. Click a timestamp to open that moment on YouTube.