โโโโโLet me describe a founder I have met many times, in different forms. She has a sensor concept, a convincing simulation and a seed round. She spends eighteen months and most of that money securing a single overseas foundry run. The die arrive unpackaged.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ There is nowhere in the country that will package two hundred of them, no accredited laboratory to qualify them, and nobody to explain why a third of them fail. Without qualified parts she cannot win a customer, and without a customer she cannot raise the next round. The company closes. Her technology was never the reason.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Across the hundreds of hardware startups I have mentored over two decades, this is the most common way a good one dies. It deserves attention now, because the country has just made its largest commitment yet to semiconductors. SEMICON India 2026, themed โSilicon to Systems: Building the Ecosystemโ, opened in New Delhi this month, and the theme names exactly the journey our startups cannot yet complete at home.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Two ends, no middle
India has made real progress at both ends of the chain. On manufacturing, the first phase of the mission approved twelve projects carrying more than โน1.64 lakh crore of investment: a silicon fab, a silicon carbide fab, a gallium nitride micro-LED display fab and nine packaging units, with Micron, Kaynes and CG Semi already in commercial production.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ On design, over 105 startups attempted chip design in that first phase and about twenty raised venture funding of around โน800 crore; the new phase aims for at least 200. Semicon 2.0, approved in July with an outlay of โน1,27,500 crore, spreads its support across six pillars: design, machines and materials, fabs, assembly and packaging, R&D, and talent.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Between a design and a fab lies the stretch where a concept becomes a qualified part. It requires prototype fabrication on a committed schedule, packaging at volumes of tens or hundreds, reliability qualification to recognised standards, and failure analysis when things go wrong.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ None of this is glamorous, and all of it is mandatory. Today most of it is either unavailable in India or available only through academic facilities that cannot promise a founder a date.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ The gap between 105 attempts and twenty funded companies has many causes. In my experience the missingโ middle is one of the largest. The two hundred design companies the mission now targets will meet the same wall, because a taped-out chip still needs prototype packaging, characterisation and reliability data before a serious customer will design it in.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
What โtranslationalโ actually means
A translational facility is defined by its commitments more than its tools. A university cleanroom optimises for publications and cannot promise schedule or yield. A commercial fab optimises for volume and will not run two hundred pieces. The translational layer sits between them and sells deliverables: processed wafers, a characterisation dossier, packaged parts, a qualification report. Each comes with a named engineer and a published price.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Five capabilities matter most to a startup. The first is maskless direct-write lithography. A mask set for even a modest process runs into tens of lakhs of rupees, and every revision means buying part of it again; removing masks turns quarterly iteration into weekly iteration. The second is a regular multi-project wafer shuttle for MEMS and sensors, which India does not have, so that eight or ten startups can share the cost of one run.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ The third is a published process design kit, so a founder who is not a process engineer can design for the line. The fourth is accredited reliability qualification to JEDEC and AEC-Q standards, because an accredited report is a document a customer can place in a procurement file. The fifth is failure analysis inside the country, so that the answer to โwhy did it fail?โ arrives in weeks rather than months.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Start before the wafer
The more provocative argument concerns what happens upstream. India added more than 20 gigawatts of solar capacity this year, yet it produces no polysilicon at all, and its first commercial producer does not expect to start until 2028.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ The renewable energy ministry said in August that it is planning a production-linked incentive for solar-grade polysilicon, aimed at roughly 10 gigawatts of capacity. That is welcome, and it is solar grade. Semiconductor grade is a different order of difficulty.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ A translational facility will not make eleven-nines silicon. That takes industrial chlorosilane chemistry, with hydrogen and pyrophoric gases handled at a scale no incubator should attempt. But the rungs up to five or six nines are within reach at laboratory scale, and they are worth climbing.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ They serve real markets today: silicon for battery anodes, precursors for silicon carbide, high purity quartz for crucibles and optical fibre, and feedstock for solar research. They also build what a country needs before domestic electronic-grade material can be trusted: laboratories that measure boron and phosphorus at parts per billion, engineers who understand feedstock, and a qualification chain for Indian substrate.โโ
Boron shows why this knowledge has to be learned rather than bought. During Czochralski growth most metallic impurities stay behind in the melt, so crystal pulling removes them almost for free. Boronโs segregation coefficient is close to 0.8, so it enters the crystal nearly as readily as silicon itself.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ It is also a dopant. At a few tenths of a part per million it fixes the waferโs resistivity before any designer has chosen one. Whoever learns to control boron from the quarry onward holds a capability that cannot be imported in a crate.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Why the crystal decides everything
If I had to name the single step that matters most, it would be the moment silicon becomes a single crystal. Everything a device engineer later relies on is fixed there, and nothing downstream can repair it.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Orientation is decided at nucleation. In Czochralski growth a seed crystal cut to the chosen orientation is dipped into molten silicon and slowly withdrawn, and the whole ingot inherits its lattice. A neck a few millimetres thick, the technique W. C. Dash introduced in the late 1950s, lets dislocations from thermal shock grow out to the surface. After that, the job is to ensure nothing else nucleates.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Two threats lead straight back to the sand. The fused-quartz crucible devitrifies at siliconโs melting point, and cristobalite particles that break away can end dislocation-free growth. Impurities rejected at the freezing front build up in the melt until the crystal loses its structure. The crucible also dissolves, adding 1017 to 1018 oxygen atoms per cubic centimetre: useful for strength and gettering, harmful as thermal donors if uncontrolled. Cleaner quartz and silicon mean longer ingots that stay single-crystal.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ The crucible is itself a strategic dependency. An estimated 70 to 90 per cent of the worldโs high purity quartz comes from one district, Spruce Pine in North Carolina, where both producers stopped work when Hurricane Helene struck in 2024.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Orientation then governs how devices behave. On (100) silicon, potassium hydroxide etching stops on {111} planes at 54.7 degrees, forming the tapered cavities under most pressure-sensor diaphragms; on (110) silicon it cuts vertical walls. Piezoresistance is strongly directional, so sensing resistors follow chosen crystal directions, and MOS devices use (100) because it has the fewest interface traps. A MEMS designer who cannot specify orientation, off-cut and resistivity cannot really design.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
That is why a translational facility should own crystal growth before it can make electronic-grade feedstock. The judgement it builds does not come on a datasheet.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Put it where the minerals and customers are
Where such a facility sits matters as much as what it contains, and Gujarat illustrates the logic. The state produces most of Indiaโs salt and therefore anchors its chlor-alkali industry, the source of the chlorine and hydrogen chloride that chlorosilane chemistry consumes.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ It holds the countryโs principal fluorspar deposit, the starting point for hydrofluoric acid, the most important single reagent in silicon purification and wafer cleaning. It has extensive silica sand, industrial power and deep-water ports. And within a couple of hoursโ drive sit the Dholera fab project and the Sanand packaging plants, exactly the customers who need failure analysis and qualification done locally rather than abroad.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Other regions will have different combinations. The principle travels: put translational capacity where the feedstock, the chemistry and the first customers already are.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Six rules for building one
Having worked through the planning of such facilities in detail, I would offer six rules to anyone attempting one.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Revenue before cleanroom. Commission characterisation, test and failure analysis first. They earn money within months, build customer relationships and teach the organisation how to operate. A facility that builds its cleanroom first spends years with neither revenue nor users who know how to use it.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Demand before tools. Release no purchase order for a front-end tool until anchor customers have prepaid for services. If industry will not commit money in advance, the demand thesis is wrong, and it is far cheaper to discover that before the tools arrive.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Automate the administration. Self-service booking, tool interlocks tied to user certification, and automatic invoicing reduce staffing by a quarter or more. In shared facilities, headcount is usually the difference between breaking even and depending on grants indefinitely.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Publish the tariff and the IP policy. Users must own the intellectual property from their own runs, unconditionally, and no customer should be able to buy a place in the queue. Neither corporations nor founders will commit to a facility that is vague on either point.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Accredit early. ISO/IEC 17025 accreditation takes a year or more. It turns a laboratory that performs tests into a body whose reports are accepted in procurement files, and competitors cannot shortcut it.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Hold the scope. Mature processes between one micron and 180 nanometres cover MEMS, sensors, power devices, photonics and detectors, which is where Indian hardware startups actually work. Advanced CMOS logic will absorb any budget offered to it and return a press release.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
The test Semicon 2.0 has created the instruments. Its machines and materials, R&D and talent pillars can fund precisely this kind of infrastructure. What remains is to apply them to the middle of the chain, so that the design companies the mission is targeting, and the hardware founders who are not yet counted anywhere, can reach a qualified part without leaving the country.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
The test I would apply is simple. When a founder in Ahmedabad, Pune or Kochi with a sensor concept and forty lakh rupees can reach a qualified, shippable device at home within eighteen months, the missing middle will have been built. The conference got the theme right. The work now is building the rooms where that journey actually happens.โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
