On DeepSeek's reverse innovation

On DeepSeek's reverse innovation

Jan 28th 2025

DeepSeek, the small Chinese AI startup, has been all over the news lately—and for good reason. It’s fascinating not just because they managed to knock the benchmarks off, but also because they took a completely different approach from what we typically see. Instead of following the usual pattern of Chinese companies merely replicating successful Western models, DeepSeek built an advanced AI model for around just $6 million.

What’s really cool here is that while many have been doubling down on aggressive, GPU-based development for large language models (LLMs), DeepSeek came up with a new approach that maximized limited resources. In a way, they’ve set an example of “reverse innovation” where a solution tailored for an emerging market ends up having a ripple effect on the global tech landscape.

Traditionally, the global tech narrative has been one-directional, with breakthrough ideas spreading out from the major hub–SF. But DeepSeek’s strategy challenged that notion entirely. They managed to create new value and even disrupt the US markets, erasing billions of dollars in one day.

One of the most intriguing aspects of DeepSeek is how they could reframe our understanding of globalization. Often, we think of innovation with globalization as a horizontal process where working innovations flow from a central source to the rest of the world. They simply showed how innovation is far more multidimensional. When companies start building smarter, cost-effective models that are tailored to their own market conditions, they don't just copy/participate in a global race—they can also redefine it.

I would always argue the potential of emerging markets when they have the necessary resources. But this is an aggressive example challenging the assumption that only massive investments can yield breakthrough technology—the SF brute force of high-budget projects. They started a trial for smaller players around the world, showing that with the right approach, they too can also have big goals and compete with anyone.

Further, they weren’t only an example of enabling small players, but another one of how globalization can be a positive-sum game. When smaller companies start new approaches, they force the larger, more established players to rethink their own. In this way, DeepSeek made a more dynamic path towards the gains of AI development. When more minds are at work solving the same problems, the end result is bound to be exciting and, often, unexpected abundance for all.