Recently, two examples left a deep impression on me. Although the industries they involve may seem distant from everyday life, they’ve given me a lot of inspiration in terms of innovation and business thinking. Both examples are about “thinking differently” and “breaking away from traditional paths.”

The first example comes from the chip industry.

As we all know, under the original technological path, we’ve faced limitations in advanced manufacturing processes. Although there might be breakthroughs in the future, there’s no clear sign of that happening yet.

However, recent reports have revealed a counter-strategy: we’ve lowered the price of 28nm chips.

Here’s the logic behind it: while advanced nodes like 7nm are indeed “cutting-edge,” the greater demand in the market lies in mature nodes. Our competitors, despite focusing on high-end processes, still rely on the profits from mature processes to fund their R&D. This can be seen from the number of fabs they’ve built. Since we can’t compete directly in the short term on high-end nodes, why not leverage our cost advantage and reduce the price of mature chips to squeeze their profit margins and weaken their ability to invest in advanced nodes? Isn’t this a kind of “tower-stealing” strategy?

The second example is from the AI field, and the protagonist is DeepSeek.

Recently, DeepSeek V3 performed remarkably well in testing—outperforming open-source models like LLaMA 3.1 405B and even rivaling closed-source models such as GPT-4o and Claude 3.5 Sonnet. What’s even more astonishing is that its training cost was only $5.576 million—significantly lower than the $1 billion budgets of models like GPT-4o. This shocked many in the international AI community.

DeepSeek's success lies in its use of software optimization and algorithmic innovation to drastically reduce the dependence on high-end hardware for training and inference. In other words, DeepSeek took a different path: through software, it achieved top-tier performance under limited hardware resources. Without diving into technical details, the core idea is simple: low cost ≠ low performance.

This strategy greatly reduces reliance on high-end GPUs—a space where companies like NVIDIA haven’t exactly been friendly to us.

While others are burning money stacking up hardware, DeepSeek focuses on the “soft” side—boosting efficiency through smarter algorithms. It made me realize that when we feel like we’ve hit a wall or the path ahead seems blocked, it might be because we’re still walking the path others laid out for us. Sometimes, stopping to look from another angle might reveal new opportunities.