Chinese tech firms OpenAI reshapes global AI race!
By Álvaro Alfaro
Beijing, Apr 6 (EFE).—
With low-cost, open-source AI models flooding the market, Chinese tech firms are mounting a serious challenge to the likes of OpenAI by reshaping the global AI race and forcing a strategic reset in the industry.
The rise of Chinese AI company DeepSeek and its affordable language model has unleashed a wave of competitive launches, prompting some to rethink the dominance of US-based leaders like OpenAI and accelerating a shift in how AI is built, priced, and shared.
Several Chinese services have matched the capabilities of OpenAI’s ChatGPT, according to industry analysts, at a fraction of the cost which has ignited fierce price wars and driving innovation toward greater openness.
Rapid-fire Releases In March, tech giant Baidu, often dubbed “China’s Google”, released its Ernie 4.5 and Ernie X1 chatbots at prices 50 percent lower than even DeepSeek’s already budget-friendly R1 model.
Not to be outdone, Tencent recently introduced its Hunyuan T1 reasoning model, offering similar capabilities and pricing, while touting “reliable results” and “low hallucination rates,” a reference to fabricated outputs that have long plagued generative AI systems.
Other major players including Bytedance, the parent company of TikTok, and Alibaba have also rolled out new AI chatbot models in recent weeks, fueling a technological arms race.
Open-Source Push Most of these Chinese models are open-source, a strategic move that has won praise from government officials.
In March, Chinese legislature spokesperson Lou Qinjian drew parallels with the mobile OS industry, recalling how closed systems like Symbian fell to open platforms like Android.
Following DeepSeek’s lead, Alibaba released parts of its Qwen model, and Bytedance did the same with Doubao.
The open-source approach, experts say, is being championed by a new generation of Chinese entrepreneurs eager to prove China can lead in innovation, not just replication.
“Entrepreneurs born in the ’80s and ’90s want to show that Chinese companies can be global innovators,” analyst Grace Shao told the South China Morning Post.
“Being cited or used by developers abroad is more exciting to them than just turning a profit.”
OpenAI itself has recently followed suit, launching a more open model, a notable pivot for the company, long known for its closed-code technologies.
Lean Development, Lower Costs DeepSeek says its R1 model was trained in just 55 days at a cost of $5.57 million using Nvidia’s H800 processors, a stripped-down version of its flagship chips.
That cost is reportedly less than one-tenth of what OpenAI spent training its GPT-4o model.
Tencent has also emphasized the memory efficiency of its Hunyuan model, addressing a key industry challenge tied to AI’s massive energy demands.
This newfound efficiency has rattled the market: Nvidia stock plunged as much as 12 percent after DeepSeek’s debut, raising concerns over the sustainability of expensive, hardware-intensive AI development models.
According to analysts at Janus Henderson, a more efficient approach to AI challenges the logic of investing billions in infrastructure and IP.
Soaring profit expectations have justified sky-high valuations, making investors vulnerable to any shake-up.
Challenges Remain Yet, not everyone in the industry is convinced this proliferation of models is sustainable.
Minimax CEO Yan Junjie told Chinese media he believed only a handful of companies worldwide will lead the next generation of large models, and cautioned that “accuracy must improve before AI can take on broader, traditional applications.”
Global expansion may also be hindered by state-imposed censorship. China’s AI regulations, introduced in 2023, require models to adhere to “core socialist values” and prohibit content that threatens national security or social stability.
As a result, users abroad have encountered refusals from Chinese chatbots to discuss politically sensitive topics like the 1989 Tiananmen Square crackdown or Taiwan’s status. EFE
aa-sk
