Understanding the Future of AI Chips: A Look at Sagence AI and Analog Technology
Graphics processing units, or GPUs, are the powerful chips that run most artificial intelligence (AI) models today. However, these chips are also known for consuming a lot of electricity. According to estimates from Goldman Sachs, the increasing use of GPUs in data centers will lead to a staggering 160% rise in electricity demand by the year 2030. This trend raises concerns about sustainability and the future of energy consumption in technology.
Vishal Sarin, a designer with over a decade of experience in the chip industry, believes this trend is not sustainable. To address these issues, he founded a company called Sagence AI (formerly known as Analog Inference) that aims to create energy-efficient alternatives to traditional GPUs. Sarin explains that the current technology limits the potential applications of AI because the devices processing data cannot meet the necessary performance standards. His goal is to overcome these limitations while being environmentally conscious.
Sagence develops both the hardware (chips and systems) and the software needed to run AI models. While many companies focus on creating custom AI hardware, Sagence stands out because it designs analog chips instead of digital ones.
What’s the Difference Between Analog and Digital Chips?
Most chips, including GPUs, operate using digital technology, which represents data as binary code—essentially strings of ones and zeros. In contrast, analog chips can represent data using a range of values, which allows them to perform certain calculations with fewer components. This can make them more efficient and faster because they do not need to constantly move data back and forth between memory and processors, a process that often slows things down.
Read Also: Customer Deposits In Eight Banks Reached N85tn In Q3
Although analog chips were widely used from the 1930s to the 1980s for tasks like modeling the electrical grid, digital chips have dominated the market since then. However, the limitations of digital chips are making analog technology appealing again. Sarin points out that many traditional chip manufacturers are still using outdated designs, which he believes is hindering the progress of AI technology.
The Advantages and Challenges of Analog Chips
One of the main advantages of Sagence’s analog chips is that they can process data more quickly and efficiently. These chips, known as “in-memory” chips, do not need to transfer data between memory and processors, which can lead to faster task completion. Additionally, because they can store data in a range of values, they can hold more information in a smaller space compared to digital chips.
However, analog technology does have its challenges. For instance, achieving high precision can be more difficult with analog chips because they require more careful manufacturing. They can also be more complex to program. Despite these challenges, Sarin believes that Sagence’s chips will complement, rather than replace, digital chips, particularly for specialized applications in servers and mobile devices.
Sarin emphasizes that Sagence’s products are designed to reduce power consumption, costs, and delays associated with GPU hardware while still delivering high performance for AI applications. The company plans to launch its chips in 2025 and is currently working with several customers to ensure that its technology fits into existing systems.
Looking Ahead: Funding and Competition
Since its founding, Sagence has raised a total of $58 million from various investors, including prominent names like Vinod Khosla and Aramco Ventures. The company is looking to raise more capital to expand its team, which currently consists of 75 people. Sarin notes that their approach to chip manufacturing is cost-effective because they are not focused on chasing the latest performance benchmarks, which can be expensive.
The timing of Sagence’s efforts may be favorable, as funding for semiconductor startups is on the rise after a slow 2023. In the first half of 2024, venture capital-backed chip startups raised nearly $5.3 billion, surpassing the total raised in all of 2023.
However, entering the chip market is not without its challenges. Manufacturing chips is costly, and international trade restrictions can complicate matters. Additionally, companies like Nvidia have established strong market positions, making it difficult for newcomers to gain traction. For example, Graphcore, another AI chipmaker, faced financial difficulties and filed for insolvency after struggling to compete effectively.
To succeed, Sagence must demonstrate that its analog chips consume significantly less power and offer greater efficiency than existing options. Additionally, the company will need to secure enough funding to scale its production effectively. If they can achieve these goals, Sagence could play a pivotal role in shaping the future of AI technology.
