Semicon 2.0 and ECMS: the plant software India's 75 approved electronics projects still need

Summary. On 15 July 2026 the Union Cabinet approved Semicon 2.0 with an outlay of ₹1.275 trillion, roughly US$13.23 billion, extending support past chip fabrication into equipment, specialty materials, advanced packaging, design and engineering services. Under the separate Electronics Components Manufacturing Scheme, 75 projects worth ₹61,000 crore were approved as of 30 March 2026, expected to create around 65,000 direct jobs, from 249 applications proposing ₹1,15,351 crore. Budget 2026-27 raised the ECMS outlay to ₹40,000 crore. India has 12 approved semiconductor manufacturing projects carrying more than ₹1.64 trillion in cumulative investment, with commercial production already running at Micron, Kaynes and CG Semi. The capital is committed. The plant software is not, and that is where new component and assembly plants lose their first two years of yield.

The gap between an approved project and a running line

An ECMS or Semicon 2.0 approval funds capital equipment, buildings and, partly, employment. It does not fund the layer that decides whether the line reaches its rated output: the manufacturing execution system, unit-level traceability, statistical process control, and the analytics that turn machine data into a yield decision.

That layer is easy to defer during construction, because nothing visibly breaks without it. What breaks later is specific and expensive. A customer audit asks which reel of capacitors went into a batch of failed boards, and the answer takes three weeks of paper records instead of three minutes of query. A yield drop appears in final test with no way to attribute it to a machine, a shift or a component lot. An automotive or defence customer requires unit-level traceability as a supply condition, and retrofitting it onto a running line costs more than building it in.

Union Electronics and Information Technology Minister Ashwini Vaishnaw, announcing Semicon 2.0, said India "will be self-reliant in the production of indigenous chips by the end of this programme". Self-reliance at component and assembly level is a data problem as much as a capital one, because the customers these plants want are the ones who audit the data.

Vaishnaw also described India's progression on the ECMS approvals: finished products, then modules, then sub-modules, and now components, with semiconductor chips developing in parallel as the active part of the component ecosystem. Each step down that stack tightens traceability requirements, because component failures propagate upward into products the plant never sees.

What the schemes actually cover

Programme	Approved	Outlay	Scope	Status as of mid-2026
India Semiconductor Mission (ISM 1.0)	2022	Original ISM allocation	Chip fabrication capacity	12 projects approved, over ₹1.64 trillion committed
Semicon 2.0	15 July 2026	₹1.275 trillion (US$13.23 bn)	Design, fabs, ATMP/OSAT, equipment, specialty chemicals and gases, materials, engineering services, research, talent	First silicon fab expected operational 2028
ECMS	Launched April 2025 by MeitY	Raised to ₹40,000 crore in Budget 2026-27	Electronic components and sub-assemblies	75 projects, ₹61,000 crore, ~65,000 direct jobs
Design Linked Incentive	Under ISM	Part of ISM	Semiconductor IP, SoC and chip design	24 design projects approved; 105 startups and MSMEs given EDA tool access
PLI, SPECS, EMC 2.0	Earlier	Various	Large-scale electronics, components, clusters	Electronics production up seven-fold over 11 years per government data

Semicon 2.0 explicitly names engineering services and industrial automation among the segments it supports, which matters for plants budgeting a software programme: the digital layer is not outside the policy scope.

The four systems a new ESDM plant needs, in build order

Most plants try to buy all of this as one platform and stall in evaluation for a year. Sequencing it works better.

1. Unit-level traceability, first

Traceability is the foundation because everything else reads from it. IPC-1782, the Standard for Manufacturing and Supply Chain Traceability of Electronic Products, sets minimum traceability requirements scaled to perceived risk as agreed between user and supplier. That risk-based framing is useful: a consumer charger and an automotive power module do not need the same granularity, and pretending otherwise inflates cost.

Practically this means assigning a unique identifier at the earliest possible point, usually a laser-marked or printed 2D code on the panel or carrier, and recording, for each unit: the material lots consumed, the machine and nozzle or feeder that placed each component, the process parameters at that moment, the operator and shift, and every test result. Design the identifier scheme before the first machine is installed, because changing it afterwards means reworking every downstream integration.

2. Machine connectivity via a standard, not point integrations

The expensive mistake is a bespoke driver per machine. IPC-2591, Connected Factory Exchange or CFX, standardises machine-to-machine communication so that placement machines, ovens, inspection systems and testers publish structured events on a common bus. That data then feeds closed-loop feedback, live production dashboards, IPC-1782 traceability, MES control, quality management and production control from one integration rather than many. IPC-HERMES-9852 handles board handover between machines in the line, and pairs with CFX.

For a plant procuring equipment right now, this is a purchasing lever: put CFX and Hermes support in the equipment specification before the purchase order, not after commissioning. Retrofitting connectivity onto machines chosen without it is the single most common cost overrun in these programmes.

3. MES, scoped to what the line actually needs

A full MES is a large system, and new plants rarely need all of it on day one. The modules that earn their place immediately are work order execution and routing enforcement, material verification at the feeder, in-line defect capture from automated optical inspection, and non-conformance handling. Scheduling optimisation, advanced planning and full ISA-95 level 3 to level 4 integration with the enterprise resource planning system can follow once the line is stable.

The integration question to settle early is which system owns the master data for parts, bills of material and routings. If the answer is ambiguous, you will spend the first year reconciling two versions of the truth.

4. Yield and OEE analytics on top

Only once traceability and machine data exist does analytics become useful, because a yield model without lot-level attribution can tell you that yield fell but not why. Overall equipment effectiveness reporting, statistical process control charts on the parameters that actually drift, and first-pass yield attribution by machine, lot and shift are the outputs that change decisions. Anything more sophisticated, including machine-learning defect prediction, depends on having a clean labelled history, which you only accumulate after the first three layers run.

Build, buy or hybrid

Decision vector	Commercial MES suite	Custom-built plant software	Hybrid: commercial MES, custom analytics and integration
Time to market	Slowest to configure for non-standard processes	Fastest for a narrow, well-understood scope	Moderate; core arrives configured, edges built
Upfront cost	Licence plus implementation, usually the largest line item	Engineering time, no licence	Licence for core, engineering for the rest
Maintenance overhead	Vendor-managed core, you own the configuration	You own everything, including upgrades	Split, which needs clear boundaries in the contract
Data control	Depends on hosting and vendor data terms	Full, including where data resides	Full for the custom layer, negotiated for the core
Fit to Indian scheme reporting	Rarely built in, usually custom-reported anyway	Built to fit	Built to fit in the custom layer
Best fit	Multi-plant groups standardising across sites	Single plant, distinctive process, strong internal IT	Most first-time ESDM plants

For most plants receiving a first ECMS or Semicon 2.0 approval the hybrid is the honest recommendation. The core execution functions are commodity and worth buying. Traceability schema, scheme and customer reporting, and yield analytics are where the plant's actual differentiation sits, and those are worth building against your own process.

Where Indian plants differ from the reference architecture

Three practical differences show up repeatedly in Indian ESDM projects.

Mixed equipment vintages. New Indian plants frequently combine current-generation placement and inspection equipment with older, cheaper machines for secondary operations. Those older machines often predate CFX. Plan for a small protocol-translation layer rather than assuming a uniform bus, and budget for it explicitly instead of discovering it at commissioning.

Connectivity assumptions. Plants in emerging industrial corridors cannot assume stable, low-latency links to a distant cloud region. Execution-critical functions belong on-premises with the cloud used for analytics and aggregation. A line that stops when the internet does is an architectural failure, not an infrastructure one.

Scheme and customer reporting. ECMS incentives are turnover-linked, capex-linked or hybrid, with parts tied to employment generation, over a six-year tenure with a one-year gestation for turnover-linked incentives and five years for capex. That reporting cadence is a data requirement, and it is far cheaper to design the plant data model so those numbers fall out of existing records than to assemble them by hand each cycle.

On personal data, most plant systems hold worker records alongside production data. Under the Digital Personal Data Protection Act 2023, with the DPDP Rules notified in November 2025 and full compliance required from May 2027, operator identity in traceability records is personal data and needs a retention and access decision made deliberately rather than by default. Our DPDP Act engineering playbook for Indian startups covers the general sequencing.

A realistic first-year plan

Phase	Focus	Output
Pre-commissioning	Equipment specification with CFX and Hermes support; identifier scheme design	Procurement conditions and a data model
Months 1-3	Machine connectivity and raw event capture	A working event bus and stored history
Months 3-6	Traceability records to IPC-1782 at the agreed risk level	Genealogy query answering a customer audit
Months 5-9	MES core: work orders, routing enforcement, material verification, defect capture	Controlled execution, fewer misbuilds
Months 8-12	OEE, SPC and first-pass yield attribution	Yield decisions with named causes
Month 12 onward	Predictive quality on accumulated labelled data	Models trained on your own line

The overlap between phases is deliberate. Waiting for one to finish before starting the next is how a plant software programme takes three years instead of one.

This sits alongside the broader shift covered in our work on AI-enabled predictive maintenance for Indian manufacturing and smart manufacturing with IoT and AI in Indian factories. Predictive maintenance in particular becomes viable only once machine event data is already flowing, which is another argument for doing connectivity first.

The three failures that show up in month nine

Reading the plan above, most teams agree with it and then hit one of three specific problems. They are worth naming in advance.

The identifier scheme is wrong and it is too late

A plant chooses a serial format that encodes the production date and line number, then adds a second line, then a second shift pattern, then a variant that carries two sub-assemblies with their own identifiers. The scheme runs out of room and the genealogy has to be rebuilt. Design for the plant you will have in year three, not the one you commission. Reserve fields you do not yet need, and never encode meaning that can change into the identifier itself.

The MES and the ERP disagree about a part number

Two systems hold parts, bills of material and routings. Both are edited. Nobody wrote down which one is authoritative. Six months later the line builds to a superseded revision because the MES copy was never updated. This is not a technology problem and no product fixes it; it is a decision that must be made, documented and enforced with a one-way sync before either system goes live.

Analytics arrives before clean data

A dashboard project starts because leadership wants visibility, and it runs on whatever data exists. The numbers look plausible and are wrong, because defect records are not linked to lots and downtime is entered by hand at end of shift. The dashboard then loses credibility permanently, which makes funding the real thing harder. Instrument first, visualise second.

Groups running several sites hit a fourth version of this, where each plant solves the same problems differently and the corporate view never reconciles. That is the case for a standard core across sites, which is also the argument in our Industry 4.0 and manufacturing AI service work in India.

What to ask a software partner before you sign

1. Which IPC standards do you implement, and can you show a genealogy query running against a real line rather than a demo dataset?
2. How do you handle a machine with no CFX support, and what does that translation layer cost to maintain?
3. What runs on-premises and what runs in the cloud, and what happens to the line when the link drops?
4. Who owns the master data for parts, bills of material and routings, and how does that reconcile with our ERP?
5. What does the scheme reporting extract look like, and is it a report or a manual assembly job?
6. What happens to operator personal data in traceability records, and how is retention configured?

A partner who answers question 2 and question 6 without hesitation has built one of these before.

India-specific considerations

SEMICON India 2026 runs 17 to 19 September 2026 in New Delhi, which is a practical deadline for plants that want equipment vendors to commit to connectivity support in writing. Vendor conversations are easier when a purchase decision is live.

Second, the DLI scheme has approved 24 semiconductor design projects and given 105 startups and MSMEs access to industry-standard EDA tools, which means a growing population of Indian fabless firms will need packaging and assembly partners locally. Plants that can demonstrate traceability and yield data to a design customer's satisfaction will win that work ahead of plants that cannot, regardless of equipment quality.

Third, Semicon 2.0 supports specialty chemicals, industrial gases and semiconductor-grade materials. Suppliers in those categories face their own traceability requirements from the fabs they serve, so the same data architecture question applies one tier up the chain.

How eCorpIT can help

eCorpIT is a Gurugram-based technology organisation, founded in 2021, with CMMI Level 5, MSME and ISO 27001:2022 certification and partnerships including Microsoft, Google and AWS. Our senior-led, multi-disciplinary engineering teams build the plant software layer described above: machine connectivity over CFX and OPC UA, unit-level traceability designed to IPC-1782 risk levels, MES integration with existing ERP systems, and yield and OEE analytics on top. We design systems aligned with DPDP Act 2023 requirements for the operator data these systems hold. If you have an ECMS or Semicon 2.0 approval and no plant IT plan yet, contact us and we will start with your equipment list and identifier scheme, which is the decision that constrains everything after it.

FAQ

What did Semicon 2.0 actually approve?

The Union Cabinet approved Semicon 2.0 on 15 July 2026 with an outlay of ₹1.275 trillion, about US$13.23 billion. It extends support beyond chip fabrication to design and intellectual property, ATMP and OSAT packaging, semiconductor manufacturing equipment, specialty chemicals and industrial gases, semiconductor-grade materials, engineering services, research and workforce development.

How many ECMS projects have been approved?

As of 30 March 2026, 75 projects worth ₹61,000 crore had been approved under the Electronics Components Manufacturing Scheme, expected to generate roughly 65,000 direct jobs. The scheme received 249 applications proposing ₹1,15,351 crore of investment. Budget 2026-27 raised the ECMS outlay to ₹40,000 crore.

Why start with traceability rather than MES?

Every other system reads from traceability records. Yield analytics cannot attribute a defect without knowing which lot, machine and shift produced the unit, and customer audits query genealogy directly. Building the identifier scheme and genealogy first means the MES and analytics layers integrate against a stable data model rather than forcing a rework.

What is IPC-CFX and why does it matter for procurement?

IPC-2591, Connected Factory Exchange, standardises machine-to-machine communication on the shop floor. One CFX integration feeds traceability, MES control, dashboards, quality management and closed-loop feedback. Specifying CFX support in equipment purchase orders costs nothing upfront and avoids building a bespoke driver for every machine after commissioning.

Should a new plant build or buy its MES?

For most first-time ESDM plants a hybrid works best: buy the commodity execution core, then build the traceability schema, scheme reporting and yield analytics against your own process. Full custom suits a single plant with a distinctive process and strong internal IT. A full commercial suite suits multi-plant groups standardising across sites.

How long does a plant software programme take?

A realistic first year runs machine connectivity in months one to three, IPC-1782 traceability in months three to six, the MES core in months five to nine, and OEE, statistical process control and yield attribution in months eight to twelve. Phases overlap deliberately. Running them strictly in sequence is what stretches these programmes past three years.

Does DPDP apply to factory systems?

Yes, where those systems hold personal data. Traceability records commonly include operator identity alongside machine and lot data, which makes them personal data under the Digital Personal Data Protection Act 2023. Retention periods and access controls for that field should be decided deliberately, given the compliance deadline in May 2027.

What about older machines without CFX support?

Plan a protocol-translation layer and budget for it explicitly. Indian plants frequently mix current-generation placement and inspection equipment with older machines for secondary operations, and those predate CFX. Assuming a uniform bus and discovering the gap at commissioning is a common and avoidable cost overrun.

References

1. India Semicon 2.0 Approved with US$13.2 Billion Outlay: Key Details — India Briefing, 15 July 2026.
2. Cabinet approves Rs 1.27 lakh crore for Semiconductor Mission 2.0 — The Federal, July 2026.
3. Govt approves Semicon 2.0 to strengthen semiconductor ecosystem — Akashvani News, July 2026.
4. 75 projects worth Rs 61,000 cr approved under ECMS programme — Akashvani News, 30 March 2026.
5. Budget 2026: Govt hikes ECMS outlay to Rs 40,000 crore — Business Today, 1 February 2026.
6. ECMS attracts Rs 1.15 lakh crore in investment proposals — Deccan Herald.
7. Electronics Components Manufacturing Scheme — Press Information Bureau, February 2026.
8. IPC-1782: Standard for Manufacturing and Supply Chain Traceability of Electronic Products — IPC.
9. IPC-2591: Connected Factory Exchange (CFX) — IPC.
10. Digital Manufacturing and the factory of the future — IPC.
11. IPC-1782 Standard overview — GlobalSpec.
12. SEMICON India 2026 — India Semiconductor Mission.
13. DPDP Act enforcement dates and compliance timeline — ConsentOS.

Last updated: 4 August 2026.