India's Public Compute Layer Reaches 68 Petaflops - Capacity Is Up 70% Since December on Three More Systems
- The National Supercomputing Mission reports 40 deployed systems with 68 petaflops combined as of September 2026.
- That compares with 37 systems and 40 petaflops reported in December 2025, so capacity is up about 70% on three more systems.
- Thirteen systems exceed one petaflop, and the plan is 50 systems with over 123 petaflops.
- The mission names indigenous Rudra servers, 100 and 200 Gbps interconnects and local cooling, but does not disclose the processors or accelerators inside.
- Aggregate petaflops are not usable capacity; utilisation, queue times and access will decide practical value.
- The National Supercomputing Mission reports 40 deployed systems with 68 petaflops combined as of September 2026.
- That compares with 37 systems and 40 petaflops reported in December 2025, so capacity is up about 70% on three more systems.
- Thirteen systems exceed one petaflop, and the plan is 50 systems with over 123 petaflops.
- The mission names indigenous Rudra servers, 100 and 200 Gbps interconnects and local cooling, but does not disclose the processors or accelerators inside.
- Aggregate petaflops are not usable capacity; utilisation, queue times and access will decide practical value.
- All India Radio / News On AIR (4 Oct 2026): https://newsonair.gov.in/national-supercomputing-mission-strengthens-indias-indigenous-supercomputing-ecosystem/
- NewKerala / IANS (4 Oct 2026): https://www.newkerala.com/news/a/indias-supercomputing-capabilities-expanding-support-ai-weather-forecasting-692.htm
- OpenGov Asia (4 Oct 2026): https://opengovasia.com/india-expands-national-supercomputing-capacity-under-nsm/
- IBEF on the December 2025 figures (37 systems, 40 petaflops): https://www.ibef.org/news/37-supercomputers-with-a-total-computing-power-of-40-petaflops-deployed-under-the-national-supercomputing-mission-nsm
The development
A government update reported on 4 October 2026 says the National Supercomputing Mission (NSM) had deployed 40 supercomputers with a combined capacity of 68 petaflops as of September 2026. The mission is steered jointly by the Department of Science and Technology and the Ministry of Electronics and IT, and implemented by C-DAC Pune and the Indian Institute of Science, Bengaluru. It was launched in April 2015 with an outlay of about Rs 4,500 crore. Reporting also says 13 of the systems exceed one petaflop, and that the plan is for 50 systems with over 123 petaflops of cumulative capacity.
What changed since December
In December 2025 the mission was reported at 37 systems and 40 petaflops. The September 2026 figures are three more systems and 28 more petaflops, a rise of about 70% in capacity on an 8% rise in system count. That pattern points to larger machines being added rather than more small clusters, which is the direction the 123-petaflop target also implies. The December figures come from earlier government reporting; the 2026 figures come from a government update carried by IANS and other outlets, so both are official statements rather than independent measurements.
Usage and the indigenous stack
The mission reports support for more than 16,000 researchers across over 400 institutions, over 15 million compute jobs run and about 1,990 research publications. On the hardware side, the mission names the PARAM series and the indigenous Rudra server line, high-speed interconnects tested at 100 and 200 gigabits per second, and locally developed cooling. The coverage does not say which processors or accelerators sit inside these machines or how much of the silicon is domestic, so sovereignty here means system design, integration and operation, not necessarily the chips.
Why aggregate petaflops need care
Petaflops measure peak computation, and a total across 40 sites is not one machine. What matters for researchers and companies is how much of that capacity is scheduled for them, queue times, software support and data movement. It also should not be compared with commercial data-centre capacity quoted in megawatts. S-103 covered the commercial data-centre layer; this signal covers the public research layer, and the two serve different users.
What to watch
- Utilisation and queue times, and the share of capacity open to start-ups and industry.
- Progress from 40 to 50 systems and from 68 to over 123 petaflops.
- Disclosure of processors and accelerators used, and any domestic chip or server procurement.
- Whether common software, scheduling and data-governance arrangements stop the systems from becoming 40 separate islands.
Track the systems we watch
Signals, reports and briefings on India’s industrial transformation.