Mythic AI, a US semiconductor company developing energy-efficient chips for artificial intelligence, has launched a Centre of Excellence in Bengaluru as part of a broader expansion in India.
The company plans to build an AI chip development team in the city and to quadruple its India workforce by the first quarter of 2027, according to YourStory's September 24 roundup. Over time, it expects to expand the Bengaluru team into product development, testing, systems engineering and intelligent sensing chips.
The move deepens Mythic's engagement with India's semiconductor and AI ecosystem at a time when global chipmakers are increasingly turning to the country for engineering talent.
A different kind of AI chip
Mythic's technology differs sharply from the graphics processing units (GPUs) that dominate AI today.
The company develops analog compute-in-memory technology. In a conventional digital chip, data is stored in memory and moved back and forth to a separate processor for computation. That constant movement consumes a large share of the energy used in AI workloads. Compute-in-memory designs perform calculations inside the memory itself, reducing data movement. Mythic goes further by doing some of that computation in the analog domain, which can be highly efficient for the matrix operations at the heart of neural networks.
Mythic claims that its approach can deliver up to 100 times greater system-level efficiency than conventional digital or GPU-based inference, according to Communications Today. Such claims are difficult to compare directly across products and workloads, but they illustrate the scale of the efficiency gains the company is targeting.
The company's Vanguard platform combines analog and digital computing on a single chiplet — a small, modular piece of silicon that can be combined with others in a larger package. It is aimed at applications including advanced driver-assistance systems (ADAS) in cars, defence and edge AI, where devices must run AI models locally with tight power limits.
What the Bengaluru team will do
According to YourStory, Mythic's Bengaluru team will contribute across architecture, front-end design, physical design and design-for-test, while also advancing the integration of analog and digital chiplets.
These are core chip-design disciplines. Architecture determines how a chip is organised; front-end design translates that architecture into logic; physical design lays out the circuits on silicon; and design-for-test ensures chips can be checked for defects during manufacturing.
The India team already contributes about a quarter of the work on the Vanguard platform, and that share is expected to rise to around half as operations expand, according to Communications Today.
Three target markets
Mythic is focusing its India expansion on three markets: data centres, automotive and robotics.
In data centres, the company's pitch is about energy. AI inference — running trained models to answer queries, generate content or analyse data — is becoming one of the fastest-growing sources of electricity demand in computing. Mythic argues that the benefit of more efficient chips is not just lower electricity bills but the ability to run more AI within existing power constraints. As data-centre operators around the world struggle to secure enough electricity, that argument is gaining weight.
In automotive, Mythic has a joint development agreement with Honda for an AI system-on-chip aimed at software-defined vehicles, with deployment targeted for the late 2020s or early 2030s.
In robotics and defence, the company is interested in autonomous systems, drones and intelligent sensing — applications where devices must process data in real time while running on batteries.
Mythic chief executive Taner Ozcelik has expressed a personal goal to eventually see Mythic chips powering autonomous systems on Bengaluru's roads, according to Communications Today.
Why India
Mythic's move follows a well-established pattern. Many of the world's largest semiconductor companies, including Intel, Qualcomm, Texas Instruments, AMD and Nvidia, have large engineering centres in India, particularly in Bengaluru and Hyderabad. India is widely estimated to account for a significant share of the world's chip-design workforce.
That talent base has become more important as chip design grows more complex and companies compete for engineers globally. For a company of Mythic's size, a Bengaluru centre offers access to experienced designers at a scale that would be harder and more expensive to build elsewhere.
The Indian government has also been working to build a broader semiconductor ecosystem. Through the India Semiconductor Mission, launched in 2021, it has provided incentives for chip fabrication, assembly, testing and packaging, and for design start-ups. Several assembly and testing facilities and a fabrication plant are under development.
Mythic's current production takes place in Germany and Arizona, with packaging carried out elsewhere in Asia. Over the long term, the company has indicated that it could consider chip manufacturing and advanced packaging in India as the ecosystem matures, according to Communications Today. It is also developing die-to-die packaging technology that could compete with TSMC's CoWoS, the advanced packaging method widely used for high-end AI chips.
The competitive landscape
Mythic operates in one of the most competitive segments of the technology industry. Nvidia dominates the market for AI training and much of the market for inference in data centres. A wide range of start-ups and established companies are developing alternative architectures, including in-memory computing, photonics and custom accelerators, each promising better efficiency for particular workloads.
Analog computing has long been attractive in theory but challenging in practice. Analog circuits are sensitive to noise, temperature and manufacturing variations, which can affect accuracy. Companies in this space must show that their chips deliver reliable results at scale, integrate easily with standard software tools and offer a compelling total cost of ownership.
Mythic's focus on edge devices, cars and power-constrained data centres suggests it is targeting markets where energy efficiency is valued most highly, rather than competing head-on with GPUs across every workload.
What it means for India
For India, Mythic's expansion is another sign that the country is moving up the value chain in semiconductors. Engineering centres that contribute to core architecture and design — rather than only supporting tasks — help build deep expertise that can later feed into domestic start-ups and manufacturing.
For Indian engineers and the global Indian diaspora in the semiconductor industry, the growth of such centres creates opportunities to work on frontier technologies without leaving the country.
And for the wider AI industry, Mythic's bet on efficiency speaks to a growing reality: as AI demand collides with limits on electricity supply, the chips that do the most work per watt may become some of the most valuable in the world. Part of that work will now happen in Bengaluru.
