The Future of Embedded AI: Inside STMicroelectronics’ Vision for the Semiconductor Industry

Pioneering Change: How STMicroelectronics is Using AI to Revolutionize the Semiconductor Industry

Few companies embody innovation like 意法半导体 in the fast-evolving world of semiconductors. Recently, I had the privilege of speaking with Marc Dupaquier , a leader in AI solutions at STMicroelectronics, about the groundbreaking work happening in the industry. Marc’s insights into how AI is driving a new era for embedded technology revealed a fascinating shift in both the practical applications of AI and the way semiconductor companies like STMicroelectronics envision the future.

STMicroelectronics: An Industry Legacy Adapting for a New Age

Founded from the merging of French and Italian technology firms, STMicroelectronics has become a global leader in microcontrollers, serving industries as diverse as automotive, consumer goods, and industrial manufacturing. However, the company’s mission has expanded far beyond the hardware of traditional semiconductors. Today, under leaders like Marc, STMicroelectronics is at the forefront of integrating AI into semiconductor design to bring enhanced functionality directly to edge devices.

Marc joined STMicroelectronics after acquiring Cartesian, an AI startup he founded that specialized in AutoML for embedded AI. “We founded Cartesian with a vision to bring AI closer to the device to make it affordable and energy-efficient,” he explained. This focus on cost-effective, embedded AI reflects the industry’s broader challenge of delivering high-tech capabilities without relying on energy-intensive cloud infrastructure.

Making Everyday Devices Smarter with Embedded AI

The potential for embedded AI has opened new possibilities across countless consumer products. Marc describes a future where, for example, a coffee machine can adjust its heat settings based on the specific coffee capsule used or alert the user when it’s time for cleaning. "For many years, the value of products was simply how well they performed their primary function," Marc noted. "But now, the inclusion of AI allows us to add new layers of value, such as predictive maintenance and customization."

Edge devices, including household appliances, cars, and industrial tools, are integrating AI to become smarter and more responsive. STMicroelectronics’ chips allow AI algorithms to run directly on devices, saving energy and reducing the need for constant cloud connectivity. This transformation means manufacturers can enhance product features without dramatically increasing costs—a critical consideration for market success.

Driving Growth in the Semiconductor Market with Edge AI

With industry analysts predicting that the semiconductor market could double by 2030, edge AI stands out as a key growth driver. According to Marc, the need for efficient, embedded AI has prompted the development of small, powerful chips that can run complex models. “For the industry, it’s a perfect storm,” he shared. “We now have product managers who understand the value AI can bring, and hardware that’s powerful enough to make it possible.”

STMicroelectronics has seen rising demand from product managers eager to integrate AI into devices that were once purely mechanical, like pumps or drills. These AI-enabled tools can now offer predictive insights—stopping a drill if it detects a potential hazard or alerting the user to maintenance needs. The convergence of affordable AI technology and client demand propels growth and reshapes expectations across consumer and industrial sectors.

The Shift from Hardware to a Hardware-Software Stack

In today’s semiconductor landscape, hardware is only part of the equation. Marc highlighted how STMicroelectronics has shifted focus from solely producing chips to developing a holistic software stack that supports embedded AI functionality. "Hardware manufacturers are becoming software people," Marc said, underscoring the company’s expansion into the software domain.

This shift enables STMicroelectronics to provide comprehensive support, from optimizing AI algorithms to delivering model compression and quantization. By merging software expertise with chip manufacturing, STMicroelectronics offers clients a complete solution that meets both their technical and financial constraints. As a result, even cost-sensitive industries can incorporate advanced AI without straining their budgets.

Sustainable AI: A Key Priority for the Future

The environmental impact of AI, particularly the high energy demands of data centers, is a growing concern for the semiconductor industry. For Marc, sustainability is essential, especially as the number of AI-driven devices skyrockets. He shared an anecdote about a large-scale AI deployment in a major U.S. city: “They wanted to use AI on 16,000 surveillance cameras to detect abandoned objects and enhance public safety. While feasible, the energy demands would have been immense.”

STMicroelectronics’ solution was to enable edge computing for these cameras, allowing them to perform AI analysis locally rather than sending constant data streams to a central data center. This shift dramatically reduces energy consumption and operational costs. Marc estimates that such an approach could save as much power as removing all yellow taxis from New York City’s streets. It’s a prime example of how edge AI can transform industries sustainably and economically viable.

Enabling AI for Everyone: Addressing Data and Development Challenges

Despite the progress in AI and hardware, many of STMicroelectronics' clients still face challenges in acquiring the data needed for effective AI training. “The number one problem for many of our clients is access to data,” Marc explained. In cases where AI requires specific environmental data—like training a lawn mower to recognize obstacles in a garden—collecting suitable data can be arduous.

To overcome this, STMicroelectronics has launched initiatives to generate synthetic datasets, accelerating the training process. Marc believes that by creating artificial data based on real-world scenarios, clients can develop effective AI models much faster. This approach has cut some development timelines from months to days, significantly expediting the path from concept to market-ready product.

Pioneering Chip Design Breakthroughs

Looking ahead, STMicroelectronics is exploring groundbreaking chip designs that could redefine the semiconductor industry. One area of focus is the development of Neural Processing Units (NPUs) designed to deliver AI capabilities on extremely compact chips. "We’re working on NPUs that are less than one square millimeter in size but offer powerful AI capabilities," Marc shared. These NPUs are engineered to perform AI computations at the device level, maximizing efficiency and reducing power consumption.

STMicroelectronics is also investing in digital in-memory computing (DIMC), which could yield even greater gains in processing efficiency. DIMC chips can handle advanced applications like vision transformers and language models by embedding AI processing directly into memory while consuming minimal energy. Marc hinted that some of these innovations will be on display at upcoming industry events, where STMicroelectronics will demonstrate the future potential of DIMC and NPU technologies.

Looking Ahead: Shaping the Next Generation of Embedded AI

As STMicroelectronics pushes the boundaries of semiconductor design, Marc remains focused on collaborating with clients to bring AI innovations to life. His team engages closely with client engineering teams, helping them navigate the complexities of AI integration. “For us, it’s about understanding the ‘art of the possible,’” he said. "When we work with clients, we learn as much as we teach. It’s a partnership that drives both of us forward."

STMicroelectronics is more than a chip manufacturer—it’s a leader in AI-driven change, committed to shaping the future of technology. With innovations in edge computing, sustainable AI, and embedded intelligence, the company is laying the groundwork for a smarter, more efficient world where AI is seamlessly woven into everyday products.