ISSCC 2026 closed on 19 February 2026 in San Francisco. Two independent bodies of reporting now exist — the IEEE's own conference report and the Chinese-language coverage that followed the ISSCC 2026 China briefing held in Nanjing on 24 November 2025 — and they reconcile well enough to be used together.
This article puts the numbers side by side, records where sources differ rather than picking one, and draws out what the distribution actually says about where circuit research sits in 2026.
Submissions and Acceptance
Both sources agree on submissions: 1,025 papers, a record and roughly 12 per cent above ISSCC 2025's 914. The acceptance count differs only by definition.
The China briefing reported 257 selected papers from 108 institutions across 15 countries, explicitly excluding four keynote talks and eight invited papers, giving an acceptance rate of 25.1 per cent against 27.2 per cent the previous year. The IEEE report counted 265 accepted technical papers including eight invited, at a 26 per cent rate. The two reconcile: 257 plus 8 invited equals 265.
The affiliation split from the China briefing is the most useful figure in the whole dataset: 193 papers from universities (78.10 per cent), 51 from industry (19.84 per cent) and 13 from research institutes (5.06 per cent).
Regions Were Redrawn This Year
ISSCC 2026 reorganised its regional committees. The Far East region became APAC, covering East Asia, Southeast Asia, South Asia and the South Pacific. The European region became EWAA, covering Europe, West Asia and Africa. The North America region became AM, covering the Americas.
On the new boundaries, APAC accounted for 169 papers, the Americas 51, and EWAA 37. Within APAC, 107 papers came from China including the mainland, Hong Kong, Macao and Taiwan.
Counting by first-author institution's country gives a slightly different and lower figure for China — 96 papers, reported as the fourth consecutive year at the top and close to 40 per cent of accepted papers. Both the 107 and 96 figures appear in reputable coverage and reflect different counting bases, not an error in either.
Institutional Leaders
Tsinghua University led globally with 18 papers by first-author institution. Among companies, Samsung led with 13. Among research institutes, the Institute of Microelectronics of the Chinese Academy of Sciences led with 5.
Chinese industry participation was reported for a seventh consecutive year, with Huawei HiSilicon, UNISOC and SmartSens named among companies with accepted papers.
Seventy highlight papers were selected across the 34 technical sessions, drawn from 45 institutions in nine countries. China accounted for 24, of which 16 were from the mainland: Tsinghua with five, Southern University of Science and Technology with two, and one each from Peking University, Southeast University, Fudan University, the State Key Laboratory of Solid-State Microwave Devices and Circuits, East China Normal University, Nanjing University, Shenzhen University, the University of Science and Technology of China and Sun Yat-sen University.
The Future-Technology Track
The most strategically interesting result is the Technical Directions track, which covers quantum computing, neuromorphic chips and flexible electronics. China placed four papers there for the first time, while Korea recorded none — a contrast that drew comment in Korean media.
Treat this as a research-direction signal rather than a capability verdict. Four papers indicate a research presence in frontier areas, not production readiness.
Two Papers Worth Reading
A concrete example of the academic-to-application gap: the University of Science and Technology of China reported a 20 GHz 2TX×4RX fully on-chip radar in 22nm CMOS at 128 mW total chip power, using a forward ΔΣ-DPLL-locked multi-injection rotary travelling-wave oscillator to achieve ADC-free phase-domain digitisation and phase-shifter-free beamforming. It was presented in Session 34 on integrated radar and UWB transceivers and demonstrated in-cabin vital-sign monitoring.
From Tsinghua, a 4D Gaussian Splatting processor in 28nm measuring 3.65 mm² achieved 0.24 mJ per frame with reported efficiency above 16 TOPS/W using quadratic frame interpolation and tree-based parallel rendering.
Both are real measured silicon. Neither is a product. That distinction is the whole point of reading ISSCC correctly.
Where Sources Disagree
Reported figures for the United States differ between sources: one account gives 56 papers by first-author country, another gives 50. China's count appears as both 107 and 96 depending on counting basis. These are recorded here as reported rather than resolved, and should be checked against the conference's own final record before use in any formal document.
