The Next Inflection In Supercomputing Is Intersection Of High-Performance Computing, AI And Quantum: AMD's Thomas Zacharia - ETV Bharat

ETV Bharat / technologyThe Next Inflection In Supercomputing Is Intersection Of High-Performance Computing, AI And Quantum: AMD's Thomas ZachariaAt the first Supercomputing India meet in Manipal, the industry leader said AMD is eager to partner India because of its strategic advantage and talent.C-DAC Director-General E Magesh welcomes Thomas Zacharia at Supercomputing India 2025 in Bengaluru (ETV Bharat)By ETV Bharat English Team Published : December 10, 2025 at 3:32 PM IST 6 Min Read By Anubha JainBengaluru: Delivering the keynote address at the inauguration of the first edition of 'Supercomputing India 2025 — HPC, AI and Quantum', which kickstarted at the Manipal Institute of Technology in Bengaluru today, Thomas Zacharia, Senior Vice-President (Strategic Technology Partnerships and Public Policy) of AMD, said, "Having worked in high-performance computing for over three decades, I recall how, in 2007, we first began thinking seriously about deploying an exascale system (supercomputers capable of performing over a quintillion calculations per second) within a 20 Megawatt power envelope. We ultimately achieved this — deploying such a system in the middle of the pandemic at under 20 Megawatts. After that, I left Oak Ridge National Laboratory [in Tennessee, USA] and joined AMD (Advanced Micro Devices)." He went on to talk about exascale computing, which was announced by the White House GENESIS Mission in November, and broader efforts to accelerate supercomputing in India, which is currently charting its own exascale journey and beyond. Zacharia said when supercomputers are imagined and deployed, it is not merely about procuring a system — but about shaping the technology as it evolves. He added that it takes about 3-5 years to design, specify, co-develop, build, and operate a system that is truly cutting-edge, and innovates on multiple fronts. Talking about the architecture of Summit, a supercomputer at Oak Ridge that was once ranked as the world's fastest, he said it was accelerated by GPUs featuring NVIDIA Tensor Cores, and was the first major supercomputer to explicitly support mixed-precision computation for large-scale scientific and AI workloads, while delivering significant performance gains where lower-precision arithmetic was suitable. Summit was decommissioned last year, while still ranking among the world’s top 10 fastest systems — highlighting its lasting impact on high-end computing, Zacharia added.He also quoted NASA’s Mars Mission, using the supercomputer Frontier as an example. He mentioned that to support the GENESIS mission, the US government announced two major supercomputers. After Frontier comes El Capitan, which will be based on MI430X GPUs and EPYC CPUs, designed for the convergence of HPC, AI, and quantum computing. He said that for AI-driven scientific workloads within the American Science Cloud, the MI355-based LUX system was announced. Finally, he said, "For future systems, we are beginning to think in terms of Megawatts per rack. Photonics will be key to driving out-of-scale systems, especially in environments interconnected with networked quantum systems."Zacharia later sat down for an exclusive interview with ETV Bharat. Excerpts: Anubha Jain: What will be the next major inflection point for high-performance computing?Thomas Zacharia: The next major inflection is really the intersection of HPC (High-Performance Computing), AI (Artificial Intelligence), and quantum [computing]. Traditional HPC has been focused on calculations. AI allows it to perform inference and model scaling in a very fast way. Quantum, of course, is a specialised case. Together, they offer tremendous opportunities for innovation.AJ: How should India strengthen its government-academia-industry ecosystem to accelerate innovation in AI, HPC, and quantum?TZ: I think the government clearly has an important role in aligning policy and having an all-of-government approach, particularly for use cases. We heard today about several use cases like digital public infrastructure, digital payment systems, etc. That’s an opportunity for the government, but also for [its] alignment with the private sector and academia. At AMD, which has a major presence in this country, we look forward to partnering and supporting that aspiration.AJ: How critical has compute capacity become for scientific and economic security, amid rising global competition in AI infrastructure?TZ: My 30-40 years of experience tells me that as we provide more compute capabilities, smart people figure out ways to deliver new and important services. A nation should aspire to have a compute infrastructure that matches its population, its ambitions, and its data needs. India is investing, and I’m really excited to see how India is leaning into this. I think it’s going to make a great difference.AJ: How would you rank India’s position in the global HPC and semiconductor value chain? What strategic partnerships does AMD consider crucial for advancing India’s compute infrastructure?TZ: India has an important role to play in the global HPC and AI ecosystems, and AMD is eager to partner. We believe that silicon diversity, open ecosystems and software are all stated priorities for the Indian government.AJ: What kind of public collaboration models are essential for countries like India to leapfrog to exascale and post-exascale computing? Are there specific lessons from global programmes that India can adopt?TZ: Partnerships are key. No one country or institution has all the capabilities. India has a strategic advantage because it has tremendous talent. So partnering within India — such as, through the MoUs signed today, but also partnering globally — is essential to being a leader in this field. India is in a great position. It needs to execute and aspire to be in that top tier. Full-stack, sovereign AI capabilities are achievable by building a diverse, open, and local AI ecosystem, deploying end-to-end AI computing to reduce costs, and empowering everyone with AI literacy.AJ: On exascale computing...TZ: Early systems like Jaguar would have required 7-9 Gigawatts — unfeasible at the time. Today’s AI datacenters operate at Gigawatt scales. Pushing to the next level will require breakthroughs in algorithms, processor design and architecture, to achieve targets like 100 Megawatt exascale systems.We need deep, long-term, trusted partnerships that challenge and push each other to achieve what would otherwise be impossible. Public–private partnerships, investment in cutting-edge R&D, and acquisition of advanced systems, are key to building sustainable, long-term engagements. Every time we deploy a new system, we add new capabilities, while leaving some areas where hardware–software co-design becomes crucial to delivering desired outcomes and achieving the next set of advancements in the field.AJ: What about convergences in future advances in computing technology?TZ: In the future, leadership science is going to require a convergence of HPC, AI, Quantum, and connected labs. Advances in HPC are dependent on AI. Zettascale isn’t a larger version of today’s systems. Achieving zettascale (10³³ FLOPS) requires new architectures, software, and system designs, not just scaling-up existing exascale systems. AJ: HPC and AI in medical science...TZ: From surgical robots to humanoid and industrial-scale robots, AI and HPC are poised to drive a massive revolution, especially in countries like the US, where manufacturing is a national strategic priority, and there are significant investments in robotics. Within the next decade, humanoid robots are expected to enter our homes. The world is evolving in fascinating and unprecedented ways.