For decades, the semiconductor industry has followed a simple rule: shrink the transistor, make the chip faster. But Huawei's top chip scientist says that era is giving way to a new one, where the key to performance lies not in making transistors smaller, but in moving data across the chip more quickly.
In a four-hour interview on the Chinese video platform Bilibili, Liao Heng, a key figure in Huawei's chip efforts, laid out what he calls the "Tau Scaling Law." The approach, he argued, will drive the next big leap in chip performance. It's already being used in a technique called LogicFolding and is slated to appear in an upcoming Huawei smartphone, according to a Bloomberg report on the interview.
Why the old playbook is running out of steam
For most of the industry's history, chipmakers relied on what's known as "geometric scaling." That's the practice of packing more, smaller transistors onto a piece of silicon. Each new generation of manufacturing technology—often measured in nanometers—brought predictable gains in speed and energy efficiency. Smaller transistors switch faster and use less power, so the industry could count on steady improvements.
But that well is running dry. Each new manufacturing step is harder and more expensive to pull off. The cost of building a leading-edge chip factory has ballooned into the tens of billions of dollars, and the physics of shrinking transistors further is getting increasingly difficult. As a result, the pace of improvement from geometric scaling has slowed.
That's where Liao's Tau Scaling Law comes in. Instead of focusing on making transistors smaller, it focuses on time—specifically, how quickly data can move around a chip. In modern processors, data often has to travel long distances across the chip, and that travel time can become a bottleneck. By optimizing how data flows, chips can get faster without needing to shrink the underlying transistors.
One technique that embodies this idea is LogicFolding, which essentially reorganizes a chip's logic circuits to reduce the distance data has to travel. The result is a chip that can do more work in the same amount of time, or do the same work faster, without the huge cost of moving to a smaller manufacturing node.
What this means for the chip industry
If Liao's approach proves out, it could have big implications for the entire semiconductor industry. For one, it suggests that the race to ever-smaller transistors—a race that has become enormously expensive—may not be the only path forward. Companies that can master data-movement optimization could gain a competitive edge without needing to build the most advanced factories.
This is especially relevant for Huawei, which has faced restrictions on accessing the most advanced chipmaking technology from U.S. suppliers. By focusing on architectural innovations like LogicFolding, Huawei may be able to squeeze more performance out of chips made with older, more readily available manufacturing processes. That could help the company stay competitive in smartphones and other devices, even as it navigates those constraints.
The broader chip market has been volatile lately, with chip stocks swinging five times faster than the broader market, a 30-year high in volatility. That volatility reflects the uncertainty around AI demand, supply chains, and the limits of traditional scaling. Liao's comments add another layer to that picture: the industry may be entering a phase where design innovation matters as much as manufacturing prowess.
Investors have also been watching how government policies shape the chip landscape. The U.S. CHIPS Act, for example, has targeted $874 million at smaller AI chip suppliers, an acknowledgment that the industry's future isn't just about the biggest players. If alternative scaling approaches gain traction, smaller firms with specialized design expertise could become more valuable.
What it means for investors
For everyday investors, Liao's argument is a reminder that the semiconductor industry is not a monolith. The companies that win in the coming years may not be the ones with the most advanced factories, but those that can innovate at the architectural level. That could shift the balance of power among chip designers, manufacturers, and equipment makers.
It also underscores the importance of following technological trends, not just stock prices. When a major player like Huawei signals a shift in how chips are designed, it can ripple through the entire supply chain—from design software to testing equipment to the materials used in manufacturing.
That said, it's worth keeping perspective. Liao's Tau Scaling Law is one approach, and it's not yet clear how widely it will be adopted or how much of a performance boost it will deliver in practice. The industry has seen many promising ideas that didn't pan out. But the fact that Huawei is already using it in a commercial product suggests it's more than just a theoretical concept.
For now, investors should watch how Huawei's upcoming smartphone performs, and whether other chipmakers start to adopt similar time-scaling techniques. If they do, it could signal a broader shift in how the industry thinks about performance—and where the next wave of growth will come from.
