DeepX CEO on Business Strategy, New Chip

Key Concepts

• On-Device AI: Processing AI tasks directly on the device (robots, smartphones, etc.) rather than relying on data centers.
• Low-Power Consumption: A key differentiator of the company’s chips, enabling applications in battery-powered devices.
• Generative AI on Device: Running large language models (up to 100 billion parameters) locally on devices.
• TOPS (Tera Operations Per Second): A measure of computing performance, used to demonstrate compliance with geopolitical regulations.
• Samsung Foundry: The manufacturing partner for the company’s chips, utilizing advanced technologies like Fiber Nano and 2nm processes.
• Embodied AI: AI integrated into physical systems like robots and drones, enabling interaction with the physical world.
• IPO (Initial Public Offering): The process of offering shares of a private company to the public for the first time.

Chip Differentiation and the Rise of On-Device AI

The company distinguishes itself in the semiconductor market by focusing on low-power AI chips, a significant departure from the high-power consumption of traditional GPU-based solutions like those offered by Nvidia. Historically, capabilities like perceiving and reacting to the physical world required GPUs, which consume over 100 watts of power, making them unsuitable for many device applications. The company’s chips, utilizing technologies like Samsung Foundry’s Fiber Nano and 2nm processes, consume as little as five watts, enabling AI functionality in robots and other battery-powered devices. This is described as a “cutting edge technology” enabling automated labor processes. The speaker believes “right now it is a time to use the CPU” – referring to their low-power technology – for these applications. The core benefit is enabling AI functions to run directly on devices, a trend referred to as “in device A.I.” and “embodied have been really key things” for the company this year.

The Dex Chip and Addressing AI Costs

A major challenge in the AI field is the high cost associated with the current data center-centric model. To address this, the company developed the Dex chip, designed for “generating AI on device.” This chip can process up to 100 billion parameter generative AI models locally, eliminating the need for constant data center reliance. An example given is processing touch ID verification without requiring data center connectivity. The Dex chip is manufactured in partnership with Samsung Foundry.

Navigating Geopolitical Challenges and Expansion Plans

The company is actively expanding globally, with a presence in China, the U.S., South Korea, Taiwan, Japan, and Europe. Despite geopolitical regulations surrounding semiconductor exports, the company maintains freedom to sell its products worldwide because its chip capacity is under 100 TOPS (Tera Operations Per Second). Specifically, their chips provide 25 TOPS and 80 TOPS, staying below the regulatory cap. The speaker acknowledges a push for onshore semiconductor capabilities in both the US and China, but believes the company’s focus on low-power, on-device AI positions it favorably. In China, there’s a strong movement towards on-device AI services, while the US is focused on advanced AI algorithms like those used by NASA. The company is seeing demand from manufacturers in the US, Japan, and South Korea, including Panasonic, LG, and Samsung.

Competitive Landscape and Future Outlook

The speaker doesn’t believe the company will directly compete with Nvidia, acknowledging Nvidia’s strength in high-performance computing. However, Nvidia’s chips, consuming around 100 watts, are unsuitable for battery-powered devices. The company is focused on the growing trend of on-device usability and anticipates a future where handheld devices will widely utilize this technology. The speaker predicts that “the major part of the AI will be the physical damage embody embody the, you know, generative A.I. of capability.”

IPO and Acquisition Considerations

The company has received acquisition offers, which were rejected. They are currently considering an IPO, having engaged with the NYSE, Nasdaq, and the Singapore Exchange. However, they are taking a “thoughtful approach,” prioritizing a strong business foundation before proceeding. The IPO timing is projected to be within the next two to three years. Valuation discussions with investors are ongoing, with a market valuation around $1 billion. The company prefers an IPO but is open to considering acquisition opportunities that align with their long-term vision. They are looking for partners who can contribute beyond chip design.

Client Portfolio and Growth Projections

The company boasts a diverse client portfolio of over 50 customers worldwide. Notable collaborations include a large-scale deployment of AI-enabled IT services with Baidu and a partnership with Hyundai Motor Robotics Group for mass production of commercial robots. The company’s chips are being deployed in various industrial automation applications. The speaker notes that they are “not hired, you know, sensitive to the investment or actually we have there, you know, some capital money” and are supported by foundry partners, allowing them to focus on partnerships and product variation. They anticipate significant growth driven by the increasing demand for on-device AI technology.

Logical Connections

The discussion flows logically from establishing the company’s core differentiator (low-power AI chips) to detailing their new Dex chip, expansion plans, competitive positioning, and future financial strategies. The geopolitical context is interwoven throughout, highlighting the challenges and opportunities in the global semiconductor market. The emphasis on on-device AI consistently ties together the various sections, demonstrating a clear strategic focus.

Conclusion

The company is strategically positioned to capitalize on the growing demand for on-device AI, offering a compelling alternative to traditional high-power solutions. Their focus on low-power consumption, coupled with partnerships with key players like Samsung and Baidu, provides a strong foundation for future growth. While navigating geopolitical complexities, the company remains committed to expanding its global presence and is carefully considering its financial options, including a potential IPO. The core takeaway is that the future of AI is increasingly moving towards localized processing on devices, and this company is at the forefront of that trend.