Why Baidu Defends AI Spending Despite DeepSeekās Success
The debate over computing power requirements in AI development has intensified globally, as established technology companies face pressure from emerging competitors.
Until now, the AI sector has traditionally relied on substantial computing infrastructure, with companies investing billions in data centres and specialised processors – and this approach has been championed by industry leaders including OpenAI and Google.
Yet recently, a particular AI startup, DeepSeek, has challenged this assumption.
The company has risen in the AI game, claiming new innovation and strategies around language model (LM) development and achieving cost savings.
Now, as companies across the world navigate US technology restrictions limiting access to advanced semiconductors, new methodologies focusing on efficiency are heating up.
Computing power remains critical to Baidu’s AI strategy
Baidu is China's largest search engine and AI company and according to Reuters, its substantial investment in computing infrastructure remains essential despite emerging technologies that promise greater efficiency, according to CEO Robin Li.
Reportedly speaking at the World Government Summit in Dubai, Robin emphasised the fundamental role of processing power in developing AI systems – saying: “The investment in cloud infrastructure is still very much required.
“In order to come up with models that are smarter than everyone else, you have to use more compute.”
Now, the company's position reflects ongoing competition in China's AI sector, where established technology companies face pressure from emerging startups developing more efficient systems.
DeepSeek innovation raises questions about infrastructure spending
Robin's comments follow recent developments at DeepSeek, which claims to have developed LMs matching the capabilities of established systems while requiring substantially less computing power.
Yet DeepSeek's breakthrough raises questions about the necessity of extensive infrastructure investments by companies such as Baidu.
- DeepSeek has developed LMs matching OpenAI's GPT performance while using less computing power
- Baidu's CEO maintains high infrastructure investment is still necessary despite DeepSeek's efficiency breakthrough
For instance, Baidu launched its Ernie large language model (LLM) following the release of OpenAI's ChatGPT in 2022 – and while the company reports that Ernie matches the capabilities of GPT-4, OpenAI's most advanced system, public adoption has not matched initial expectations.
Furthermore, the emergence of DeepSeek contradicts Robin's previous assessment of China's AI sector.
Robin had stated that the emergence of another company matching OpenAI's impact was unlikely, demonstrating the rapid evolution of technology across the world at the moment.
Sanctions reshape China’s AI development
US technology restrictions on Chinese companies have created additional pressure to develop efficient AI systems, as these sanctions limit access to advanced semiconductors and computing components, forcing companies to adapt their development approaches.
Robin notes that these constraints have catalysed innovation within China's technology sector, saying: “You just don't know when and where innovations come from.”
The shift towards open source development
Although Robin still advocates for closed-source development as the only viable path, he also acknowledges the potential benefits of open-source approaches – and this shift reflects broader industry trends, where open-source development enables wider participation in AI advancement and accelerates technology adoption across sectors.
“If you open things up, a lot of people will be curious enough to try it. This will help spread the technology much faster,” he tells Reuters.
Therefore, the strategic change to open-source approaches suggests not just recognition of changing market dynamics, but emerging companies like DeepSeek that demonstrate the potential for innovation through alternative development approaches.
Explore the latest edition of Technology Magazine and be part of the conversation at our global conference series, Tech & AI LIVE.
Discover all our upcoming events and secure your tickets today.
Technology Magazine is a BizClik brand
