Qualcomm Technologies has announced the tape-out of its 2nm semiconductor design. The technology is said to improve performance, power efficiency and integration, enabling more advanced AI processing on devices.
Srini Maddali, Senior Vice President, Engineering and HW Lead, Qualcomm India, outlined the significance of the 2-nanometre technology, discussing its capabilities, potential use cases and the differentiation it brings.
Can you give an overview of the 2-nanometre technology unveiled today?
2nm is the latest and cutting-edge technology that we unveiled today. The Global Qualcomm Organization has been working on this technology for some time. We then identified an opportunity where this technology will enable product differentiation, which is why we created a design, implemented it, and taped it out. We look forward to leveraging this technology in many products coming. One thing about generational technology is they give an advantage in performance, cost and power scale.
The performance will be significantly higher than earlier technologies, while power consumption will be lower by an order of magnitude. Costs could also see a meaningful benefit, as much more design can be packed into the same area, making it more cost-effective. As we move through this technology migration, it allows us to integrate far more complex hardware, enabling capabilities that were difficult to achieve in previous generations.
What are the key use cases for this technology?
One aspect is AI at the edge, where the absence of capability to run large models in real time forces devices to rely on cloud connectivity. This introduces challenges like latency and privacy. With newer, more complex technologies and deeper hardware integration, these models can now run directly on the device, without needing to connect to a network.
And so, you can do more operations on the device. The advantage also is privacy and security because your data is not going out. The uniqueness of this technology and our IP is that we have an NPU, a GPU and a CPU. All these are different capabilities, but our hardware software integration is such that, based on the workload or user requirements, our software decides where to run this particular activity in the die. In some cases, the same workload can be split into multiple parts, allowing it to run optimally on the CPU, the GPU, or a combination of both.
The key advantage is achieving the best performance at optimal power. Without this capability, running workloads on a single processing unit can be suboptimal—either draining the battery or reducing performance. By leveraging three different processing units, workloads can be distributed to deliver the highest performance at the lowest power consumption. This is the primary benefit, and it enables far more AI to be executed at the edge as well as in hybrid models.
In a hybrid approach, most workloads run on the device, while only specific tasks that require higher capability are sent to the cloud. This avoids transferring entire datasets, preserving privacy and security, while keeping the majority of sensitive processing on the device.
Which sectors or device categories are likely to rely on or benefit from this innovation?
Edge AI spans a wide range of devices, including smartphones, routers, and CCTV systems. Qualcomm’s approach is to build scalable IP that can operate across form factors. While the IP is not identical across devices, it is designed to scale effectively across this spectrum.
This scalability also means the same product is not limited to a single application. A chipset designed for a smartphone, for instance, can also support an in-vehicle infotainment system in a car, or be used in a laptop. The technology can address a range of use cases across devices.
The 2 nm technology, in particular, can support multiple applications, some of which may emerge only when customers begin exploring new possibilities. Often, customers identify new use cases. We are watching to see how customers adopt the technology and how new products and use cases emerge in the market.
Can you outline the contribution from the India teams for this particular development?
Qualcomm’s India team, over the last two decades, has developed IPs and products and contributed to them. Any product or technology that Qualcomm develops, some or major portions of it, depending on the case, our teams in India have contributed to it.
Qualcomm looks at the talent pool available worldwide as an R&D organisation that can be scaled up. They will use the available talent capabilities as per their requirement.
Have you worked with local OSAT players in any capacity?
Qualcomm works with manufacturing partners worldwide, as per customer requirements. In India, OSAT facilities are now being developed, and we are interested in these emerging capabilities. Given our diverse technology stack and chip development programmes, we need to assess what Indian manufacturers can deliver and how those capabilities fit with our products before leveraging them.
Qualcomm CEO Cristiano Amon had earlier met with the Prime Minister, during which it was conveyed that, as and when manufacturing capabilities become available in India, the company would be keen to utilise them.
With ISM 2.0 taking shape, what level of involvement does Qualcomm see for itself?
We are a fabless design company, and ISM 2.0 is not only focused on manufacturing and packaging but also on strengthening chip design in India. We have been working with various government organisations and industry bodies on the “design in India” concept. We have been designing here for some time now and will continue to.
Beyond developing IPs and products, we engage with the broader ecosystem, including universities, which we support through funding and collaborative research. University teams carry out research projects with us, and we also bring students in as interns before hiring some of them full-time. We believe in building an ecosystem that can scale and deepen capabilities over time. As part of ISM 2.0, we will work closely with the government and industry bodies to accelerate this effort, however possible.
