4 min readJul 9, 2026 11:53 PM IST
First published on: Jul 9, 2026 at 05:51 PM IST
On June 23, at the International Supercomputing Conference, a machine almost no one had heard of vaulted to the top of computing’s most-watched scoreboard. LineShine, housed at the National Supercomputing Centre in Shenzhen, clocked 2.198 exaflops on the High Performance Linpack benchmark — more than 20 per cent faster than the United States’ El Capitan at 1.809. It is the first Chinese system to top the TOP500 since Sunway TaihuLight in 2017, and the first machine ever to cross two exaflops using only general-purpose CPUs, no GPUs.
That last detail is the story. China did not win by buying the best chips; it was barred from doing so. LineShine runs on a fully domestic stack — custom 304-core LX2 processors built on the Arm architecture, a proprietary LingQi interconnect, the homegrown LingKun platform and Kylin OS, across nearly 13.8 million cores. Beijing stopped submitting to the TOP500 as Washington’s export controls bit; re-entering the summit is a deliberate signal that the chokehold did not work as intended.
The export controls since October 2022, meant to freeze China’s ambitions, instead triggered a state-backed self-sufficiency drive. Chip self-sufficiency was just 33 per cent in 2024, and is targeted at 80 per cent by 2030. TrendForce expects domestic chips to supply half of China’s AI-chip market in 2026. Three toolmakers, Naura, AMEC and ACM Research, entered the global top 20 for the first time in 2025; Huawei’s AI-chip revenue is projected to jump 60 per cent to $12 billion.
This is the value-chain ascent in miniature: China climbing diligently from assembly into design, tooling, materials and architecture — the high-margin, hard-to-replicate links where strategic leverage sits.
For India, the moment is instructive. Launched in 2021 with a Rs 76,000-crore outlay, the India Semiconductor Mission has approved 13 projects across seven states. The Tata–Powerchip fab in Dholera, a Rs 91,000-crore, 28nm bet on 50,000 wafers a month, anchors the effort, with first silicon due around 2028.
Three lessons can be drawn from this result. First, an ecosystem is not a fab: China’s edge came from depth across equipment, materials, interconnects, software and design IP built over a decade. ISM 2.0’s refocus on equipment, materials and Indian IP suggests the lesson is absorbed; execution is the test. Second, design is leverage, up to half of value addition, and India already hosts the R&D centres of Nvidia, Intel, Qualcomm and Texas Instruments, a fabless strength that may pay faster than chasing leading-edge fabs. Third, patient capital and talent win: With a 700,000-strong skills shortage looming by 2030, the country that trains engineers fastest compounds its advantage.
The deeper takeaway is geopolitical. Washington’s denial strategy assumed that withholding technology would preserve its lead indefinitely. China’s answer is that denial accelerates substitution, fragmenting the global stack into parallel, incompatible ecosystems. For India, that fracturing is an opening: As multinationals seek alternatives to China-Taiwan concentration, a non-aligned, English-speaking India grows attractive, provided it does not mistake assembly for autonomy.
LineShine’s lesson is not that China has won the computing race. It is that a determined state, denied the best tools, can build adequate ones and weaponise the symbolism. Washington must weigh whether controls that catalyse a rival’s industrial policy are worth their cost. New Delhi should read the result plainly: Sovereignty in silicon is bought with sustained investment, indigenous IP and engineering talent, and not subsidies alone.
The writer is a defence and tech policy adviser and author of The Digital Decades: On 30 Years of the Internet in India
This is the value-chain ascent in miniature: China climbing diligently from assembly into design, tooling, materials and architecture — the high-margin, hard-to-replicate links where strategic leverage sits.(Representational Image) 
Share
