After half a year, will China's DeepSeek change the AI industry?

DeepSeek - Ảnh 1. — DeepSeek has proven that developing an effective AI model does not require huge investments - Photo: REUTERS

It's been six months since DeepSeek shocked the world . Today, China's groundbreaking app is no longer a hot topic. But that doesn't mean it's going away.

Although there are no exact figures, many Americans are still using DeepSeek, according to the BBC. Some Silicon Valley startups have chosen to use DeepSeek instead of more expensive American AI models to save costs.

DeepSeek changes perceptions of AI

For years, US tech giants like OpenAI have led the AI race with the “bigger is better” view, meaning the bigger the model, the more data and the more computing resources, the smarter it is.

This has led to a costly race, spending billions of dollars to build giant data centers and buy the most expensive chips.

DeepSeek broke this stereotype with a development cost of only about $5.6 million, while its performance was on par, even surpassing, Western models in some criteria.

This is seen as a clear demonstration that “smarter engineering” can produce superior performance.

One of the technical factors that helps DeepSeek do this is the use of a "Mixture-of-Experts" (MoE) architecture, which allows the model to only activate the parts that are most necessary for a particular task, saving significant resources.

The DeepSeek incident sent shockwaves through the stock market, with Nvidia’s $600 billion loss in a single day highlighting the market’s sensitivity to any shift in the AI balance of power.

The event has forced investors to reconsider the value of chip and AI technology companies, which have been highly valued thanks to the belief in the race to "pour tons of money" to build large models.

China's position in the AI race

Geopolitically , DeepSeek has changed perceptions of China's position in the AI race.

China has traditionally been seen as a “catch-up,” but DeepSeek has shown the true potential of Chinese tech companies, prompting the U.S. government and tech companies to step up efforts to maintain their lead.

The Trump administration's subsequent statements about the importance of "America remaining the dominant power in AI" also showed that DeepSeek had become a symbol of the competition between the two superpowers.

However, DeepSeek's success has also come with serious concerns about data security. The fact that DeepSeek is based in China and its privacy policy suggests that user data may be stored and processed there has raised concerns among many companies and individual users.

Many people have sought to run this model on their own devices to avoid the risk of data being secretly shared.

The Future of AI: Revival or Return to the Old Way?

The AI industry has been transformed since DeepSeek was born. OpenAI unexpectedly released two models free and open source, a move that many saw as an admission that smaller, more efficient models have an important role to play.

At the same time, however, OpenAI continues to pursue GPT-5, an even larger model, showing that the trend of "pouring big money" is still very strong.

Nvidia’s stock rally and record highs reinforce the notion that the “big is good” race isn’t over yet. Companies like Meta are also investing billions in AI and competing for talent with pay packages as high as $100 million.

While DeepSeek has demonstrated a promising new direction, tech giants still believe in the old path of more data centers, more chips, and more power.

Finally, even DeepSeek is facing challenges. According to professor Marina Zhang of the University of Technology, Sydney, DeepSeek's next model, DeepSeek-R2, has been delayed due to a lack of high-end chips.

This shows that even a company with an efficient approach cannot completely escape hardware supply challenges, a major hurdle in today's complex geopolitical landscape.

According to the BBC, DeepSeek delivered a much-needed jolt, forcing the AI industry to rethink its basic assumptions. It opened up a debate about efficiency and cost, and demonstrated that innovation can come from anywhere.

However, it seems that the traditional path of Western AI companies is still dominant.

Back to topic

EAST SEA

Source: https://tuoitre.vn/sau-nua-nam-deepseek-cua-trung-quoc-co-lam-thay-doi-nganh-ai-20250811171149177.htm