The Invisible Killer of Carbon Emissions: The Double Dilemma of Code and Chips

Hello, everyone. I’m Ian. Today, let’s talk about semiconductor chips and low-power code. This topic might sound a bit high-tech, but it’s actually closely tied to every aspect of our daily lives. First of all, if we use our chips every day, they’d have to run 24 hours a day for testing. In our country, chips tend to degrade within two years of continuous use because the heat generated from running nonstop is too intense. So, if we could use low-carbon solutions to reduce power consumption, we might be able to extend their lifespan. This isn’t just about physically prolonging the chip’s life—it also involves practical applications. If we follow this line of thinking, we’ll be able to conduct more experiments in the future, right? Of course, the issues we face now aren’t limited to chips—we also need to consider code. Code runs on computers, and every line of code consumes electricity. That electricity ultimately comes from power plants, which burn coal and oil, emitting carbon that ultimately contributes to climate change. So, should we reduce data traffic to minimize code execution? Actually, no—code logic itself can be low-carbon. For example, optimizing code and reducing redundancy can lower carbon emissions. Our company is researching exactly this kind of code to see how much we can optimize it. Take ChatGPT, for instance: developing its model consumes as much electricity as a small town. The volume of data and energy consumption behind it are truly staggering. Research by the Mozilla Foundation found that carbon emissions from communication technologies are actually on par with those of the aviation industry, ranking fourth globally. This reveals a fact: the internet’s impact on the environment is far greater than we imagine. In the past, people thought data was invisible and intangible, but in reality, every step of its production, transmission, and processing is accompanied by carbon emissions. For example, receiving a single email results in 5 grams of carbon dioxide emissions. Sending a single advertising email may result in 50 grams of carbon dioxide emissions—and that doesn’t even include images and videos, which could add even more. According to statistics, the number of emails sent globally each year may exceed 1 trillion, many of which are advertising emails. For individuals, the proportion of advertising emails far exceeds that of emails they actually need to read. This is not just a personal issue; it is a challenge for the entire industry. Returning to chips: if our country’s chips are used daily, they’ll likely need to be replaced within about two years. If we adopt a low-carbon solution—such as reducing energy consumption—the chips could last longer. However, such solutions are costly and time-consuming, so we must first find a temporary solution—for example, using “low-carbon bags” to reduce the chips’ thermal effects through low-carbon methods and extend their lifespan. This is a solution at the physical level. Now let’s turn our attention to code. Logical coding practices also contribute to low-carbon solutions—for example, reducing redundant code and optimizing algorithms can both lower carbon emissions. It’s like a weight-loss pill: it can be used in certain areas, but not everywhere. Our company’s Greenewab platform is like such a “weight-loss pill.” It can reduce data volume by 97.25%, thereby cutting electricity consumption by 45%. The reason behind this is that much of the code in these systems was written thirty years ago without considering the future. Every time a new person joins or a new outsourcing firm takes over, they pile more code on top, resulting in a massive chunk of “legacy code” that’s difficult to clean up. We start by transforming large systems, optimizing them bit by bit. If the risk is low and the impact on the company’s core business is minimal, then it’s worth a try. Most companies won’t let us directly modify core code, but we can teach their programmers how to optimize it. For example, we helped an oceanography company optimize its intranet, which had been in use for decades. After our optimization, we found that we could remove a lot of unnecessary data, improve logical performance, and reduce energy consumption. This process is time-consuming, but the results are significant. Therefore, future development is not just a technical issue—it’s also a matter of sustainable development. Both code and chips are calling for new solutions, and we’re working hard to provide answers to this challenge.