India's ₹1.275 trillion chip programme is, in practice, a power procurement decision

India's Union Cabinet approved Semicon 2.0 on 15 July 2026, a ₹1.275 trillion programme, roughly $13.23 billion, spanning chip design, fabrication, advanced packaging, materials, equipment, R&D and workforce development ^[1][2][3]. Fifteen days later, on 30 July, Indian power companies including Tata Power and Reliance announced nuclear investments following legislative changes that ended the state's monopoly over nuclear energy ^[10][20]. Cryptobriefing, crediting the Carnegie Endowment, frames the two as one decision: advanced fabs cannot run without enormous and reliable power ^[13][22].

Take the framing seriously and the arithmetic gets uncomfortable. India operates 24 reactors totalling roughly 8.18 to 8.78 GW, with eight more under construction adding 6.8 GW ^[6][7]. That pipeline lands the country near 15 to 15.6 GW, against a stated target of 22.5 GW by 2031-32, leaving about 7 GW that is neither built nor being built ^[8][16]. The government says it plans to commission 21 additional reactors by 2031, which means 13 units beyond the current construction fleet would have to be started and finished inside about five years ^[9][18]. The 100 GW target for 2047 requires roughly 12 times today's capacity, a compound rate near 12 to 13 percent a year sustained for 21 years ^[8][17].

Meanwhile the chips do not wait. Twelve to 13 semiconductor projects have been approved across six states with combined investment above ₹1.64 lakh crore, and the flagship Tata-PSMC fab at Dholera in Gujarat is expected to produce first silicon by December 2026 ^[4][5]. That is roughly five years before the nuclear step-up is scheduled to arrive ^[19]. Whatever powers Dholera and the projects behind it through the late 2020s will be the existing grid, with its existing reliability profile. The nuclear opening is insurance for the second decade of this build, not the first.

The subsidy intensity is the other number worth sitting with. ₹1.64 lakh crore is ₹1.64 trillion, so the Semicon 2.0 envelope is equivalent to about 78 percent of the capital already committed across approved projects ^[14]. No year-by-year disbursement split has been reported ^[22]. The implied conversion in the announcement is about 96 rupees to the dollar, which matters for anyone modelling equipment imports priced in dollars against incentives paid in rupees ^[15].

For operators, the useful signal is not the headline number but the instrument mix. The government is exploring Bharat Small Reactors and Small Modular Reactors specifically designed for industrial applications, including powering semiconductor plants ^[11]. Dedicated behind-the-fence generation is a different proposition from adding gigawatts to a national grid: it changes who carries outage risk, who negotiates tariffs, and whether a fab's power cost is a policy variable or a contract. Both programmes sit under Viksit Bharat, the framework targeting developed-nation status by 2047, the centennial of independence ^[12].

Three things to watch. First, whether Dholera actually produces silicon by December 2026, because that date is the only near-term test of execution in the whole package ^[5]. Second, whether the Tata Power and Reliance announcements convert into sited, licensed projects rather than memoranda, since 13 unstarted reactors by 2031 is the binding constraint on the 22.5 GW figure ^[10][18]. Third, whether any Bharat Small Reactor or SMR is contracted to a named fab; that would be the first hard evidence that the energy plan and the semiconductor plan are the same plan, rather than two announcements that read well together ^[11].