India's State Incentive War: Gujarat's Execution Edge, Karnataka's FDI Inversion, and Why Nobody Can Tell You Who Actually Got Paid

The menus: what states are actually offering

Every major manufacturing state now runs its own layer of incentives on top of whatever the Centre offers, and the design choices differ meaningfully. Gujarat's 2026 industrial policy sorts investors into MSME/Large/Mega/Ultra-Mega tiers with capital-subsidy ceilings from 35% (MSME) up to 40–50% for Ultra-Mega and selected thrust sectors (toys, footwear, robotics, drones), plus a flat 7% interest subsidy and 100% electricity-duty exemption across every tier. Its dedicated Semiconductor Policy tops up the central ISM subsidy (50% of eligible capex) by an additional 40% of that central amount — 20 more percentage points — taking total capex support to roughly 70%. Uttar Pradesh matches that semiconductor top-up at 50 percentage points and has already made one large, named award under it — ₹7,037 crore in capital subsidy to Tarq Semiconductors, plus ₹321 crore in land-cost refund. Tamil Nadu takes a structurally different approach: investors choose one of four mutually exclusive options at the outset, the most distinctive being 100% SGST reimbursement on finished-product sales for a full 15 years, an option with no fixed rupee ceiling.

Terms above are the states' own stated policy language (Gujarat's Viksit Gujarat Industrial Policy 2026; UP's Semiconductor Policy 2024 and ECMS 2025; Tamil Nadu's Industrial Policy 2021; Rajasthan RIPS 2024; Punjab IBDP 2026, gazetted 8 March 2026; Odisha IPR 2022; Karnataka Industrial Policy 2025-30) — not verified against disbursal.

What actually landed: implementation, not intent

DPIIT's own IEM Part-B data — industrial investment that has moved from "intended" to "implemented" — tells a much more concentrated story than the incentive menus suggest. Across 4,859 implemented projects between 2021 and 2025, worth ₹19,51,828 crore nationally, just three states account for 63.3% of the total.

Two patterns jump out, and they hold up even after replacing the single-year FDI snapshot with a full five-year cumulative figure (DPIIT Table No. 5, calendar 2021–2025, computed by summing the annual state-wise columns across two report vintages). Odisha is the starkest case: 17.2% of India's implemented industrial investment on just $212 million of cumulative FDI — 0.08% of the national FDI total over the same five years. Its investment is almost entirely domestic capital (steel, metals, refining) landing and getting built, not foreign money. Gujarat tells a softer version of the same story: a real 33.8% implementation share against a comparatively modest 10.2% FDI share — still meaningfully below its execution dominance. Maharashtra and Karnataka run the opposite skew on paper: Maharashtra is both #3 by implementation and #1 by FDI (32.5% of the national cumulative total), while Karnataka is the clean inversion — 22.8% of India's five-year FDI inflow but only 2.2% of implemented investment. Money is arriving in Karnataka at real scale; comparatively little of it is showing up as landed factories in this dataset. Tamil Nadu, despite the most aggressive and flexible incentive menu of any state here, pulls a respectable 6.2% of national FDI yet sits mid-table on implementation (9th, 2.3%) — a reminder that its historically strong auto cluster mostly predates the current incentive window, so recent generosity hasn't yet visibly moved the ranking.

Maharashtra's FDI number has the same problem as Gujarat's

Calling Maharashtra a "genuine dual engine" needs a caveat, because its own DPIIT state synopsis shows the identical registered-office distortion already flagged for Gujarat's Jio Platforms case — just less widely reported. DPIIT's top-25 list of Maharashtra's largest individual FDI deals (October 2019 to December 2024) is dominated by transactions with no Maharashtra factory behind them at all. Three of the top five are a single company, Reliance Retail Ventures Limited (activity: "storage and warehousing"), receiving stakes from Saudi Arabia's Public Investment Fund (₹9,555 crore), Singapore's SLP Rainbow Holdings (₹8,813 crore) and Qatar Holding LLC (₹8,278 crore) — a nationwide retail and warehousing operation booked 100% to Maharashtra because RRVL is Mumbai-registered. Rank 3 is Ambuja Cements (₹8,341 crore, from Mauritius), whose actual plants sit mostly in Himachal Pradesh, Gujarat, Rajasthan and Chhattisgarh, not Maharashtra. Rank 5 is HDFC Credila Financial Services (₹7,642 crore, from the Netherlands) — a pure financial-services transaction with no physical project anywhere.

DPIIT's own published sector breakdown for Maharashtra confirms this isn't cherry-picked: across October 2019–December 2024, Services (23.26%), Computer Software & Hardware (13.77%) and Trading (5.47%) together account for 42.5% of all FDI reported at Maharashtra — higher than the equivalent 39% share those same three categories hold nationally. Construction (Infrastructure) adds another 16.13%, itself a mixed bag of real and financial activity. Automobile Industry, the only unambiguous plant-and-machinery manufacturing sector in Maharashtra's own published top five, is just 4.33%. None of this means Maharashtra has no real manufacturing investment — its #3 implementation ranking is separately measured, real IEM data — but its headline FDI dominance is substantially a Mumbai head-office and financial-transaction artefact, the same distortion this article already documents for Gujarat, just not previously named for the state that tops the FDI table.

A sector-adjusted comparison: how much of each state's FDI is actually manufacturing?

DPIIT publishes an official "top five sectors" breakdown, but only for its own top five FDI-ranked states nationally — Table 6.3(i) through 6.3(v), covering exactly Maharashtra, Karnataka, Gujarat, Delhi and Tamil Nadu, and no other state. That is the complete set; there is no equivalent document for Odisha, Andhra Pradesh, Uttar Pradesh or any other state, confirmed by checking directly. Four of these five states already appear in this article's implementation table; Delhi did not, since it isn't one of the top implementation states, but is included here for completeness since it's a genuine top-5 FDI state by DPIIT's own ranking. All five synopses cover October 2019–December 2024, except Delhi's, whose most recent published synopsis runs only to December 2023 — a one-year-older vintage, noted rather than smoothed over.

On this sector-adjusted view, Tamil Nadu is the clear outlier in the right direction: Automobile Industry and Prime Mover (industrial engines and turbines) together account for roughly a quarter of its top-five FDI, by far the highest manufacturing-coded share of any of DPIIT's own top-5 FDI states — consistent with its real, longstanding auto-component cluster, even though it ranks only 5th by FDI volume among them. Gujarat's 47% Computer Software & Hardware share is almost entirely the Jio Platforms remittances already documented above; its next two sectors, Non-conventional Energy and Power, are genuine physical infrastructure (this is where the Adani renewable-energy deals in Gujarat's top-25 list sit) but still not manufacturing in the plant-and-machinery sense. Karnataka, Maharashtra and Delhi all show single-digit-to-low-teens manufacturing-coded shares once the one unambiguous plant sector in each state's published top five (Automobile Industry for Karnataka and Maharashtra, Electrical Equipment for Delhi) is isolated — the clearest confirmation that Bengaluru's, Mumbai's and Delhi's FDI totals are dominated by their roles as corporate-registration and financial-services hubs rather than by industrial capacity landing in those cities. No sixth state has an equivalent published breakdown: DPIIT's sector-detail reporting is, by design, limited to its own top five FDI states nationally, which is itself a data-availability finding worth flagging rather than a gap in this research.

The national PLI report card

Zooming out from states to the Centre's flagship Production-Linked Incentive schemes, the disbursal pattern is just as uneven. As of 31 December 2025, ₹28,748 crore had been disbursed nationally against an outlay PIB itself states inconsistently as either ₹1.91 or ₹1.97 lakh crore — roughly 15% either way.

Solar PV's zero is a design feature — the money is deliberately deferred until a year after commissioning, and ~30GW is already built. ACC Battery's zero is the opposite: a genuine execution failure, with one flagship project stuck in pilot-scale production and another that exited the scheme without the word "withdrew" ever appearing in a government document.

Why the state-level numbers can't be trusted at face value

DPIIT's own metadata document for its FDI statistics states, in its methodology section: "Documentation on methodology: Not applicable." That's not an oversight worth glossing over — it explains a specific, checkable distortion. State-wise FDI is attributed by the investee company's registered office address, not the location of the actual factory or project (this was true even more coarsely before October 2019, when attribution ran by RBI regional office, several of which span multiple states). NCAER flagged this exact problem in a 2016 report and recommended switching to project-location attribution; the recommendation was never adopted.

The clearest live example is Gujarat's own FDI table. At least 9 of Gujarat's top 10 individual FDI remittances between October 2019 and December 2022 were a single company — Jio Platforms Limited, registered in Ahmedabad but operationally headquartered in Navi Mumbai, Maharashtra — receiving its well-publicised multi-investor capital raise from Google, Meta, Saudi PIF, Silver Lake, Mubadala, KKR and others. "Computer Software & Hardware" made up 66% of all Gujarat FDI in that window — not manufacturing. It's also why Gujarat topped the national state-FDI table in FY2020-21 with a 36.79% share, two full years before its Semiconductor Policy 2022-27 existed; crediting that FDI surge to a chip policy would be a basic sequencing error.

The distortion cuts the other way too. Tata Electronics' ₹27,000 crore Assam ATMP semiconductor project is domestic balance-sheet capital and therefore cannot appear in DPIIT's FDI-equity series at all — Assam's actual recorded FDI equity inflow in FY2024-25 was ₹24.26 crore ($2.88 million) against that ₹27,000 crore headline project. The same applies to Tata's Agratas battery gigafactory, Reliance's Jamnagar gigafactory, and several other marquee domestic-capital projects: real, large, and invisible in the one dataset most commonly cited as evidence of "investment attractiveness."

That opacity is a choice, not a technical limit: the government's own Animal Husbandry Infrastructure Development Fund publishes a public, no-login, nightly-refreshed dashboard with full state/district/bank-wise funnel data, including rejections (11,026 applications → 1,592 eligible → 910 sanctioned → 357 disbursed). Nothing prevents the same standard being applied to PLI or FDI data — it simply hasn't been.

The arrears divide: who admits what they owe

Beyond the incentive terms and the FDI attribution problem sits a third, quieter divide — which states actually disclose what they've paid, and which have a documented backlog of what they've promised but not delivered.

Two economic concepts that explain why this data looks the way it does

Charles Goodhart's observation, made about UK monetary targets in 1975 but generalised far beyond it, is usually summarised as: "when a measure becomes a target, it ceases to be a good measure." State-wise FDI was never designed as a competitive scorecard, but it has become one — states cite it in budget speeches, investors cite it in site-selection decks, and journalists cite it as a proxy for "investment climate." Once a number is being watched that closely, the entities being measured (in this case, companies choosing where to register a holding entity, not where to build a factory) have every reason to optimise for the number rather than for the underlying activity it was meant to proxy. Gujarat's Jio Platforms remittances and Maharashtra's Reliance Retail/Ambuja/HDFC Credila registrations are not fraud — they are the entirely predictable result of a registered-office attribution rule sitting underneath a widely-watched target, exactly as Goodhart's Law predicts.

The states' incentive menus themselves are a live example of Charles Tiebout's 1956 model of competition among local jurisdictions: in Tiebout's framework, mobile residents (or, here, mobile capital) "vote with their feet," and jurisdictions compete by offering different bundles of taxes and benefits, which in theory should sort investment toward its most efficient location. The menu comparison in this piece — Gujarat's tiered capital subsidy, Tamil Nadu's SGST-reimbursement option, Odisha's uncapped capex subsidy, West Bengal's wholesale revocation — is a real, live version of that competition. But Tiebout's model assumes the "voters" (investors) have full information about what they're choosing between, and assumes jurisdictions face a real cost for over-promising. Neither assumption holds well here: the arrears data (Telangana's ₹3,736 crore, Andhra Pradesh's ₹3,000–5,000 crore) shows states can offer generous terms with limited consequence for under-delivering on them, and the FDI-attribution problem means investors and journalists comparing states are often comparing a Mumbai-registered-office count, not a real menu of executed projects.

Sources: DPIIT IEM Part-B implementation data and DPIIT AR 2025-26 Appendix V (via the author's own trade-policy research pipeline); DPIIT "Metadata for Foreign Direct Investment Equity Inflow in India" (June 2025); DPIIT FDI State Synopsis tables; RBI Form FC-GPR filing instructions; NCAER, "Enhancing the Scope and Quality of Indian FDI Statistics" (2016); PIB PRIDs 2230621, 2246089, 2246085 (Feb–Mar 2026) on PLI disbursement; Gujarat Viksit Gujarat Industrial Policy 2026; Uttar Pradesh Semiconductor Policy 2024 and ECMS 2025; Tamil Nadu Industrial Policy 2021 (official text, storage.investingintamilnadu.com); Rajasthan RIPS 2024; Punjab IBDP 2026 gazette; Odisha IPR 2022; Karnataka Industrial Policy 2025-30; West Bengal Bill No. 5 of 2025. Several figures (Gujarat/UP battery-specific subsidy rates, MP's incentive formula, some biotech disbursal claims) are flagged by the underlying research as secondary-sourced or unverified and are excluded here. This post is AI-assisted analysis of public and primary-research data, not investment advice.