India's most-watched semiconductor project will begin making chips at the 90-nanometer process node — not the 28nm that Tata Group's chairman wrote into the company's own annual report, and that Taiwan's PSMC publicly committed to just two years ago. Commercial production, previously promised for the end of 2026, is now expected to begin in mid-2028. The two-node gap and 18-month slip were confirmed by people familiar with the matter and by Tata Electronics' own spokesperson, who told reporters on July 17 that the plan "was always to start with 55nm and 90nm, followed by 28nm."

That claim is difficult to square with Tata Group Chairman Natarajan Chandrasekaran's own words. "We have chosen to start our chip journey at the 28nm node — it is a stable node," Chandrasekaran wrote in Tata Sons' annual report for the year ended March 2025. PSMC Chairman Frank Huang told India's Economic Times that Dholera would debut with 28nm chips that could "later move to 22nm," a commitment consistent with PSMC's official technology transfer agreement describing 28nm as the most advanced licensed node. IT Minister Ashwini Vaishnaw said at the March 2024 groundbreaking that Dholera's first chip would arrive in December 2026 at the 28nm, 50nm, and 55nm nodes.

The disclosure lands two days after New Delhi approved a fresh ₹1.27 lakh crore ($13.3 billion) in semiconductor subsidies under India Semiconductor Mission 2.0, a program designed in part to keep the Dholera project competitive with global peers.

What 90nm and 28nm Actually Mean

Process nodes in the semiconductor industry are not simply size measurements — they are engineering generations, each requiring a different suite of manufacturing equipment, chemical processes, and process-control expertise. The 28nm node, which appeared in production around 2010, requires deep-UV immersion lithography and roughly twice the number of design rules needed at 80nm or wider nodes. It is also, according to research cited by chip designer Sophie Wilson, the node with the lowest cost per logic gate in the industry — the economic sweet spot where transistors are densest relative to process complexity. That combination of commercial maturity and cost efficiency makes 28nm the node of choice for smartphones, automotive control units, wireless chipsets, and communications equipment.

The 90nm node, by contrast, serves a narrower and less commercially competitive tier: analog and mixed-signal circuits, industrial sensors, embedded microcontrollers with non-volatile memory, and long-lifecycle automotive applications where reliability, not performance density, is the primary requirement. This is a real market — the 2021 global automotive chip crisis was a shortage of exactly these chips — but it is a smaller, lower-margin market than 28nm.

Crucially, moving from 90nm to 28nm is not a linear progression. It requires a fundamental change in photolithography equipment (from dry deep-UV to immersion systems), and roughly doubles the process-control complexity of every manufacturing step. The Tata-PSMC technology transfer agreement, signed in September 2024, licenses a portfolio spanning 28nm through 110nm — meaning the eventual 28nm goal remains within scope of the deal, but achieving it requires building a level of fab expertise that Tata does not yet have.

Learning Curve Logic: A Defense With Real Merit

Both Tata and PSMC frame the 90nm start as deliberate sequencing, not a broken promise. In chipmaking, this logic has genuine operational foundation. Semiconductor yields — the fraction of working chips per finished wafer — start near zero when any new fab opens and climb as engineers gain mastery over process chemistry, equipment calibration, and contamination control. Foundry industry research has consistently found that delaying a yield ramp by as little as six months can eliminate a significant share of a process node's expected profit, because most of a foundry's earnings on any node come during the brief window after yield surges but before capacity loosens.

Starting at a wider geometry — 90nm, with its more forgiving tolerances and shorter process step count — gives Tata's engineering teams time to calibrate equipment, build yield management expertise, qualify materials suppliers, and train technicians before attempting the tighter controls that 28nm demands. Tata has been recruiting executives with foundry operations experience, including from GlobalFoundries, a pure-play foundry that built its business serving exactly this tier of the market.

PSMC itself has operational 28nm capability: its Tongluo Science Park fab in Taiwan, which opened in May 2024, runs 28nm and wafer-stacking processes at commercial scale. The technology being licensed to Dholera is therefore proven in PSMC's own facilities, not merely aspirational. PSMC Spokesperson Eric Tang said the company's technology transfer to Tata "covers several technology nodes with the most advanced process being 28nm," and that beginning with mature nodes before advancing is standard foundry practice.

"It is common that the introduction of the technology platform will be gradual, starting with more mature nodes," Tang told reporters.

IT Minister Vaishnaw confirmed the mid-2028 commercial target on July 15, placing the full elapsed time from Cabinet approval (February 2024) to first commercial output at approximately four years — broadly consistent with global norms for greenfield fabs, though notably longer than what the government's earlier communications implied.

Why the Gap Between Promise and Delivery Still Matters

The learning-curve argument is legitimate. The gap between what Tata and PSMC said in 2024 and what they are doing in 2026 is also real. Chandrasekaran's "stable node" line in the 2025 annual report was not a statement that 28nm would be the fab's ultimate destination — it was a statement about what the fab would produce from day one. Vaishnaw said "first chip in December 2026 at 28nm." That is the record. Tata's spokesperson's claim that 90nm "was always the plan" contradicts those public statements without explaining the discrepancy.

The practical consequences extend beyond corporate accountability. Supply chain managers at Indian automotive companies, electronics manufacturers, and government procurement agencies who made decisions based on Dholera's stated 28nm timeline will need to recalibrate their planning horizons. The difference between 90nm output (power management ICs for industrial applications) and 28nm output (chips for smartphones, 5G infrastructure, modern automotive control units) is not a minor technical footnote — it is a different customer base, a different competitive landscape, and a different level of strategic value to India's electronics ecosystem.

The 18-month timeline slip compounds this. Industry sources, prior to the July 17 report, had already noted that first silicon was likely to arrive in the first half of 2027 rather than December 2026. The July 15 confirmation of mid-2028 for commercial production — as opposed to trial output — suggests a further gap between when Dholera produces its first working wafer and when it is manufacturing chips in sufficient volume and quality to supply real customers.

A $735 million loan package secured from HSBC, MUFG, First Abu Dhabi Bank, DBS, and ANZ in February 2026 came with conditions that analysts described as "unconventional": lenders effectively relied on Tata Group's brand reputation rather than conventional project economics, with the project's 163.5-acre Dholera land pledged as collateral at a valuation of zero, reflecting its undeveloped status.

What 90nm Competes Against: Chinese Foundries Already Running at Scale

The strategic context for Dholera's 90nm debut is significantly less favorable than it was in 2023, when the project was conceived. Chinese foundries — Semiconductor Manufacturing International Corporation (SMIC), Hua Hong Semiconductor, and Nexchip — have been aggressively expanding mature-node capacity with state subsidies, targeting precisely the 40nm–90nm range that Dholera will initially serve.

Hua Hong's FAB9 facility in Wuxi is expected to approach 83,000 wafers per month at 90nm–350nm specialty processes by 2027. SMIC's Beijing expansion is targeting 100,000 wafers per month when complete. Nexchip is ramping to 100,000 wafers per month at 55nm and 40nm. All operate under explicit government subsidies and at cost structures that Dholera, as a new entrant, cannot immediately match.

SMIC's co-CEO Zhao Haijun said in the company's first-quarter 2026 earnings call that AI demand had "directly pushed power-management and other mature capacity into shortage," prompting customers to seek capacity — predominantly in mainland China. Average utilization at older-generation 8-inch wafer fabs hit approximately 90% in 2026, up from 80% in 2025, according to research firm TrendForce.

The demand is real, in other words. But so is the Chinese supply. Tata will enter the 90nm market in 2028 with capacity that Chinese foundries already established by 2025. India's surviving competitive argument is not price — it is geopolitics: the security and resilience value of sourcing chips from a democratic, non-adversarial partner. That argument is genuinely valuable to Western customers and to Indian defense and infrastructure operators. But it is a fundamentally different value proposition from the one the government's public communications have emphasized.

ISM 2.0 Arrives Two Days Too Late to Rewrite the Fab's Opening Chapter

The Union Cabinet's July 15 approval of ISM 2.0 — ₹1.27 lakh crore across chip design, manufacturing equipment, and supply-chain development — underscores how seriously New Delhi takes the semiconductor bet over the long term. The program extends ISM's lifecycle from five to twelve years, introduces a 30% incentive on manufacturing equipment, chemicals, gases, and materials, and is explicitly designed to attract incremental investments beyond what the original ₹76,000 crore ISM framework covered.

What the new subsidies cannot do is retroactively offset capital already committed to Dholera, or compress the timeline between 90nm production and 28nm production. The ₹1.27 lakh crore applies to future, incremental investments. The Dholera fab — now consuming $11 billion in a 2028 trajectory — will arrive at 28nm when the engineering team is ready, not when the policy calendar says so.

As of April 2026, construction was approximately 50% complete and cleanroom installation was under way. Three OSAT and assembly-and-test facilities in the broader ISM ecosystem are already operational: Micron Technology's $2.75 billion facility in Sanand, inaugurated on February 28, 2026, and now shipping DRAM and NAND modules; Kaynes Semicon's packaging and testing plant in Sanand, operational since March 31, 2026; and CG Power's joint venture with Japan's Renesas in Sanand, scaling assembly operations. These facilities, which package and test completed chips, will eventually process wafers from Dholera — but only when Dholera is producing them.

What to Watch

The 28nm roadmap is the single most important unanswered question. Neither Tata nor PSMC provided a specific timeline for when 28nm production will begin. Analysts and procurement managers should watch for customer tape-out announcements, process qualification certificates, or trial-run completion notices that would signal progress toward the more advanced node.

Yield data at 90nm will tell the fuller story. A new fab that achieves consistent, commercially viable yields at 90nm within its first 18 months of trial production is demonstrating the operational muscle that 28nm requires. Yields that lag — or that take two to three years to stabilize — would imply a longer-than-expected path to the node the project was sold on.

ISM 2.0's incentive uptake over the next 12 to 24 months will reveal whether other foundry partners or technology providers see Dholera as an anchor worth building around. A second or third major fab commitment to India — ideally at a node closer to 28nm from the outset — would significantly change the strategic picture.

Frequently Asked Questions

When will India's Dholera fab produce its first chips, and at what process node?

Trial production is targeted for late 2026, but commercial output is not expected until mid-2028, according to IT Minister Ashwini Vaishnaw's July 15 statement. The opening production node will be 90nm — not the 28nm that Tata and PSMC publicly committed to in 2024. Tata's spokesperson says 28nm remains a planned part of the fab's eventual offering, but no specific timeline for reaching it has been disclosed.

Why does the gap between 90nm and 28nm matter for India's chip strategy?

The difference is not purely technical — it is market-defining. The 28nm node serves smartphones, wireless chipsets, modern automotive control units, and consumer electronics: the chips India most needs to stop importing. The 90nm node serves industrial microcontrollers, power management ICs, and older automotive applications. Starting at 90nm means India's first domestically produced wafer-fab chips will address a narrower market segment than promised, and one where Chinese foundries are already significantly over-supplied.

Can India's 90nm chips compete with Chinese foundries?

On price, it will be very difficult. SMIC, Hua Hong, and Nexchip are all expanding 40–90nm capacity with state subsidies and established customer bases, and will have years of head start by the time Dholera is at commercial scale in 2028. India's competitive advantage at 90nm is not cost — it is supply-chain sovereignty: Western customers, defense organizations, and Indian government agencies that need to source chips from a geopolitically trusted partner. That is a genuinely valuable niche, but it is a different argument from the one the Indian government's public communications have made, which emphasized cost and commercial competitiveness.

What protections does ISM 2.0 offer buyers or supply-chain planners counting on Dholera?

ISM 2.0's ₹1.27 lakh crore ($13.3 billion) in new subsidies applies to future, incremental investments and does not retroactively cover costs or timelines already committed to the Dholera project. Supply-chain managers who were planning sourcing decisions around a 28nm supply from Dholera by late 2026 should treat mid-2028 as the earliest realistic date for commercial-scale 90nm output, and should plan separately for when 28nm capability might be available — likely no earlier than 2030, and not yet confirmed with any milestone schedule.