Case Replay: Inside India's ₹1.5–₹2 billion (~$15.8–$21 million) OSAT Entry — Bhiwadi, Sanand and the Anchor-Customer Test

Most coverage of the back-end build-out stops at the ribbon-cutting. This replay starts after it, because that is where the operational decision-making lives. India's outsourced semiconductor assembly and test (OSAT) expansion has now produced at least three distinguishable playbooks — a small-and-medium-enterprise (SME) packaging line, a state-subsidised flagship fab-adjacent plant, and a joint venture locked to a foreign device maker — and each one hands plant managers, procurement leads and capital-committee members a different answer to the same question: can a modestly funded line survive before a committed buyer appears?

The arithmetic framing the debate comes from Nitin Dodia, founder-director of Smart Fab Systems, in an interview reported by EE Times. For mature, non-advanced work, he put the starting band at ₹1.5–₹2 billion (~$15.8–$21 million) — more precisely ₹1.5 billion (~$15.8 million) to ₹2 billion (~$21 million), with the final figure set by product mix, tool list and local operating costs. “Simple wire bonding or leadframe-based packaging is less capital-intensive,” Dodia said. A May 2026 government notice supplied the working proof point: Sahasra Semiconductors, whose facility in Bhiwadi, Rajasthan has drawn more than ₹1.5 billion (~$15.8 million).

Case file 1 — Bhiwadi: how an SME template was assembled

Sequence first. Sahasra's packaging site was set up under the Ministry of Electronics and Information Technology's Scheme for Promotion of Manufacturing of Electronic Components and Semiconductors (SPECS), the component-window incentive track that sits beside the flagship fab and assembly subsidies run by the India Semiconductor Mission (ISM). Press material issued around the launch describes the Rajasthan unit as the country's first SME-led semiconductor facility, inaugurated in mid-May 2026 at Salarpur-Khushkhera, inside Bhiwadi's Electronics Manufacturing Cluster — an established electronics assembly belt rather than a greenfield park, which shortened the hunt for board handlers, metrology technicians and shift supervision.

The spec sheet of that entry case is deliberately unglamorous: the site covers 5,295 square meters, or 57,000 square feet, and lists an annual packaging capacity of 60 million units. Its product list is memory chips for microSD cards and flash storage, LED driver ICs, eSIMs and RFID products.

Volume, not ambition, is what sets the equipment bill. A company such as Micron, which handles millions of chips each month, must buy tools sized for that flow, Dodia argues; an OSAT aimed at lower-volume mature products can open with a far smaller outlay. For comparison, the global assembly-and-test services market that these entrants are fishing in was put at roughly $47 billion in 2025 and forecast to about $71 billion by 2030 by Mordor Intelligence — a demand pool big enough to absorb small lines, but one whose growth is concentrated in high-end formats.

Case file 2 — Mahape: the regional supplier that started before the incentives matured

RRP Electronics, meanwhile, runs an OSAT facility in Mahape, Navi Mumbai, Maharashtra, serving automotive, industrial and telecom customers. Trade reporting from Business Standard records it as Maharashtra's first operational OSAT/ATMP line when it launched in 2024, and the company has since extended into wafer-level packaging and secured land for a larger footprint in the state — a case of a board-level manufacturer moving upstream into back-end work before the national subsidy map was complete.

Scale contrast: what the flagship programmes are buying

The announced figures below show the two orders of magnitude that separate flagship lines from an SME entry case, and hint at why their risk profiles are not comparable:

Facility	Site	Announced investment	Capacity	Entry logic
Tata Electronics	Assam	₹270 billion (~$3.25 billion)	48 million units a day	National-scale line, subsidised fab programme adjacency
CG Power	Sanand, Gujarat	₹76 billion (~$910 million)	15 million units a day	Joint venture with a device maker that also buys the output
Kaynes Technology	Sanand	₹33.07 billion (~$395 million)	more than 6.33 million units a day	Electronics manufacturer extending into power packaging
Sahasra	Bhiwadi, Rajasthan	more than ₹1.5 billion (~$15.8 million)	60 million units a year, about 164,000 units a day	SME line on mature packages, SPECS support

Read as an operating lesson, the table says something uncomfortable: capital per shift-hour is not the scarce variable. The scarce variable is the order book that fills it.

Lesson 1 — Order book before equipment list

“The bigger point, however, is that success is not just about money,” Dodia said. “Whether you are setting up a smaller or a larger operation, you need to find an anchor customer.” In his account, the decisive variables are capacity utilization, experienced staff and product qualification; a new OSAT should line up at least one buyer willing to take output over three to five years, and that requirement bites harder when the opening investment is modest.

Early entrants such as Sahasra and RRP Electronics can point to customers already buying their parts. “That matters more than simply investing the money,” Dodia said. A buyer could spend ₹1 billion (~$11 million) to ₹2 billion (~$21 million) on tools, yet idle machinery earns nothing while no orders are queued.

The Sanand joint venture shows the logic executed. CG Semi — the CG Power venture with Renesas and Stars Microelectronics — moved its G1 line from groundbreaking in 2024, to test operations in August 2025, to commercial production in July 2026, when its first qualified chips went out to Renesas Electronics India, according to the companies' launch announcements. The plant did not have to discover a market after the fact; the technology partner sat on both sides of the table. Kaynes Semicon follows a similar pattern with its own technology partners and anchor buyers.

“Without that arrangement, there is little point in proceeding,” Dodia said. A global buyer such as Apple or Samsung might instead end up coordinating 10 smaller OSATs, each tied to a ₹1 billion (~$11 million) to ₹2 billion (~$21 million) build, to satisfy its needs. “This creates additional management work for the customer as well,” he noted.

Utilization outweighs opening capex, in Dodia's telling, because equipment economics are unforgiving at either end: costs are high whether a company spends ₹1 billion (~$11 million) or ₹10 billion (~$105.4 million). Idle tools lock up cash and still attract interest. “Therefore, what matters most is whether your products and equipment are being utilized to their fullest potential,” he said — and investors need patience, because returns are typically zero in the first year, with no meaningful results expected in the first two or three years.

Lesson 2 — The staffing line is the real critical path

“In my view, the workforce, particularly experienced personnel, is the most underestimated requirement,” Dodia said. Vendors will quote tools and materials, but people who have actually run semiconductor-grade equipment are scarce everywhere, and Indian OSATs bid against the U.S. and other countries expanding chip production to cut external dependence.

The macro picture supports his worry: Deloitte projects that the industry will need more than one million additional skilled workers by 2030, equivalent to over 100,000 a year, while a Deloitte-SEMI survey found 82 percent of executives naming a shortage of qualified candidates. Dodia expects the pool of semiconductor-grade operators and process engineers to thin quickly.

His reason is that much of the know-how is unwritten. Two identical tools from one vendor can diverge as residue builds inside chambers; an experienced engineer reads statistical process control charts and locates the source. “Most of the time, the knowledge exists in someone’s head,” he said.

Wasi Uddin, senior manager of R&D and training at Kaynes Semicon and a former OSAT and ATMP consultant at MeitY, framed it differently in an earlier interview: the “immediate gap is not materials or equipment; it is the expertise required to operate semiconductor-grade manufacturing facilities.” Much of that expertise arrives from the Philippines and Malaysia, or through Indian engineers who worked abroad and returned, and manufacturers must also outbid IT-sector salaries for the same people, Uddin said.

Lesson 3 — Qualification is capex you cannot see on the balance sheet

Dodia describes qualification as the second barrier. Cycles often run 12 to 24 months, so a new OSAT cannot ship the moment equipment arrives. A customer may request six or 12 consistent batches before accepting a supplier, and even then the OSAT may only be installed as a secondary source. Material suppliers can spend about 12 months qualifying with an OSAT.

For automotive work, the gate is even better defined. The Automotive Electronics Council's AEC-Q100 specification requires failure-mechanism stress testing across temperature grades from Grade 0 (−40 to +150 °C) to Grade 3 (−40 to +85 °C), with accelerated-life blocks commonly running 1,000 to 2,000 hours at zero failures for each lot condition — a test plan that alone consumes quarters, not weeks. That rigidity explains why buyers are slow to move and why a small line's first year is so often cash-negative.

Lesson 4 — Geography either compounds or quietly drains margin

In Dodia's model, smaller OSATs should ideally sit in clusters. Instead, states compete for semiconductor money and sites scatter nationwide. India has approved 12 projects under the India Semiconductor Mission; counting the SPECS unit in Rajasthan brings the operational project list to 13, per India Briefing's running tally. A chemical supplier serving these plants and companies such as RRP Electronics may need warehouses in eight or 10 locations, and some materials — including leadframes for certain relatively advanced packages — must come from within a few hundred kilometers of the OSAT because shipping can harm them.

Tata has one plant in Dholera, Gujarat, and another in Assam, so its suppliers must run logistics from west to east. Dodia accepts that states have reasons to share the industry, but says supply chains will mature more slowly: “Something that could have taken three years may now take five years,” he said.

Sanand illustrates the alternative. Gujarat has three OSATs in Sanand and three more in the surrounding areas, so a supplier facility nearby can serve three companies — and Micron's own assembly, test and packaging plant, inaugurated in Sanand in February 2026, deepens that same node further. Uddin adds that OSATs can import materials for the first three to five years while local suppliers develop. “Why is there no leadframe supplier till now in India? Because there was no OSAT, there was no market,” he said. “Micron, CG Semi, and Kaynes formed a consortium and held monthly supplier meetings so suppliers could see their combined demand,” Uddin said.

Where the openings actually are: retired lines and long-tail parts

Commodity packaging puts smaller OSATs against suppliers worldwide, Dodia says, while large OSATs steer growth toward advanced formats. Amkor's advanced packaging revenue reached $5.56 billion in 2025, whereas its mainstream wirebond business remained flat at $1.15 billion. Yole Group puts the whole advanced-packaging market at roughly $55 billion in 2025, heading above $109 billion by 2031 — the segment the giants are chasing, and the one a small line should approach only with a customer in hand.

Large OSATs also retire older lines. In 2022, ASE's Chung-Li plant in Taiwan decommissioned its punch quad flat no-lead line, forcing NXP to withdraw two parts. Dodia sees an opening for smaller OSATs to absorb discontinued products. “Many products have such well-established applications that demand continues for 10 to 20 years. Automotive is one example,” he said. NXP has committed to supplying several automotive microcontroller families for more than 15 years, and Rohm Semiconductor specifies supply periods of 10 to 20 years for some products, including automotive MOSFETs.

He warns against crowding, too: if five or 10 smaller OSATs all aim at automotive, they will race each other. “Ideally, they should coordinate their efforts, with one company targeting one segment and another focusing on a different segment,” he said. Some could chase advanced packaging when suitable customers appear, since startups and R&D teams often prefer newer technologies.

Segmentation and the buyers who actually show up

Uddin says the major players have already divided work by segment: Micron packages memory; CG Semi focuses on legacy packages such as QFN and thin shrink small outline packages (TSSOP); Kaynes concentrates on power packaging, with Alpha and Omega Semiconductor having booked 28% of its capacity. “It will be a competition, that is for sure, but it will not be a throat-cutting kind of a competition,” Uddin said.

Some customers pay for dedicated capacity. Defense buyers may not want products exposed elsewhere and can pay more for a separate line. R&D teams and startups need small lots quickly: if a large manufacturer receives a million-dollar order, a $10,000 order may wait months, and in those situations time matters more than unit cost. Fabless startups are probably the best customers for new Indian OSATs, Dodia says — serving them needs a flexible model able to support tens of processes, because their volumes are low and requirements move fast. Integrated device manufacturers and electronics manufacturing services firms can also become buyers.

Global companies such as Renesas may qualify a smaller Indian OSAT for two reasons, he adds: incentives may let it offer better prices, and a dedicated partner can provide planned capacity over three to five years without competing against the hundreds of customers served by ASE or Amkor. The conclusion he draws is a tiered ecosystem — large OSATs for volume, smaller ones for faster turnaround, security or specialised work, a structure some mature markets abroad already run. If the ecosystem serves only high-volume customers, startups, defense buyers and R&D teams will not obtain what they need. “We definitely need both types of OSATs,” he said, with older products and mature technologies still able to meet many Indian requirements.

Five misconceptions these case files expose

• “Cheaper capex means lower risk.” A ₹1 billion (~$11 million) line with no anchor buyer fails for the same reason a ₹10 billion (~$105.4 million) line does: idle assets still carry interest and depreciation.
• “The tools define the plant.” Divergence between two identical chambers is a maintenance-knowledge problem, not a procurement problem — and that knowledge is largely undocumented.
• “Capacity equals orders.” Nameplate numbers such as 60 million units a year mean little before six or 12 consistent qualified batches have been accepted by a customer.
• “Automotive is an open field.” Qualification rigour plus several already-committed entrants means late-comers compete on yield history they do not yet have.
• “Location is a subsidy decision.” State-by-state competition spreads suppliers across eight or 10 locations and lengthens supply-chain maturation by years.

Action checklist for operations, procurement and capital teams

1. Write the demand letter before the tool PO. Seek a named buyer committed to output over three to five years, or a technology partner that is also the anchor customer — the pattern that took CG Semi to commercial production in about two years.
2. Model utilization, not investment. Board papers should lead with line-fill rate by quarter and treat capex as the second slide.
3. Build a two-shift qualification calendar. Assume 12 to 24 months before first acceptance; budget for the six or 12 consistent batches and for material suppliers spending about 12 months in parallel.
4. Recruit the operator layer early. With Deloitte forecasting over one million additional skilled workers needed globally by 2030, hire for chamber-level experience before ramp, and codify tacit settings into written control plans.
5. Co-locate or coordinate. Prefer a cluster such as Sanand, where nearby suppliers serve three companies; if a dispersed site is unavoidable, price the warehousing and logistics penalty for materials that cannot travel far, such as certain leadframes.
6. Pick a lane in the long tail. Retired package types, automotive parts on 10 to 20 year lifecycles, defence-dedicated lines and small-lot startup work are differentiated positions; commodity wirebond is not.

FAQ

How much does a mature-packaging line actually take to start?

Nitin Dodia's range is ₹1.5–₹2 billion (~$15.8–$21 million) for non-advanced work, with wire bonding and leadframe-based packages at the lower end; Sahasra's Bhiwadi site, at more than ₹1.5 billion (~$15.8 million), is the published reference case.

What most often stalls an OSAT ramp?

Not equipment delivery. The binding constraints are anchor demand, semiconductor-grade staffing and qualification cycles of 12 to 24 months.

Can a small plant survive against the giants?

Only in defined niches: legacy packages the majors retire, small lots and fast turnaround for startups and R&D, security-sensitive defence work, and power or automotive long-tail parts. Amkor's 2025 split between $5.56 billion of advanced packaging and flat mainstream revenue at $1.15 billion shows where the majors' attention is.

Is local material supply a precondition?

No — Uddin allows imports for the first three to five years while local suppliers develop, which is precisely what the Micron, CG Semi and Kaynes supplier consortium is trying to shorten.

Should new entrants chase advanced packaging?

Dodia's advice is conditional: follow chiplet, system-in-package and fan-out work only when suitable customers appear; otherwise compete for specialised or long-tail segments instead of head-on volume.

For industrial operators, the transferable point is narrower than a national policy debate: back-end capacity is a service business wearing a manufacturing coat. Its first fixed cost is trust — in a tool's yield history, in an engineer's judgement, in a customer's forecast — and trust takes longer to install than a die bond takes to arrive.

Case analysis for WorldIndustrialWeekly's General desk by Michael Torres, Industrial Weekly. Corrections and confidential industry tips: news@industrialweekly.com