Alibaba and Baidu Adopt Homegrown Chips for Next-Generation AI ModelsDrawn by demands of an increasingly internet-driven economy, it’s become much more viable to grow your own.

Sept. 13, 2025 – Two of China’s most valuable technology companies, Alibaba Group and Baidu Inc., have leapfrogged the West to train some of the world’s most powerful computers while developing their own chips to do so—a significant step toward technological independence in an intensifying arms race for tech supremacy. Now those chips are being rushed into both companies’ services, The Information reported today, as the firms seek to bypass supply limitations and strengthen their competitiveness in a rapidly changing AI market.

A Strategic Pivot in Global Chip Tensions

For years, Chinese tech giants have leaned on high-performance chips from global suppliers—particularly U.S.-based companies such as Nvidia—for the vast computing power needed to train large-scale AI systems. But geopolitical tensions and export controls have limited access to that hardware, pushing companies to create it themselves.

By bringing their own chips into the training pipeline, Alibaba and Baidu are signaling a strategic change of course. Rather than waiting for imported GPUs or relying on third-party providers, they are designing semiconductors tailored to their specific workloads and AI frameworks.

“Making proprietary chips is no longer simply optional — it’s turning into a must-have for key AI players in China,” said a Shanghai-based semiconductor analyst. “Being able to manage and command chip design and production can significantly reduce dependence on foreign technology and assure a steady supply for critical A.I. projects.”

Alibaba’s T-Head Division Is the Pioneer

• T-Head, Alibaba’s chip-development unit, has been quietly but surely building up semiconductor prowess for years.
• The company already announced the Yitian 710 server processor for cloud computing and has been developing chips specialized for AI workloads.
• Alibaba-branded chips are reportedly being used to train advanced large language models and other generative AI systems in Alibaba Cloud.

These chips are believed to be custom-designed to perform the huge amounts of parallel processing that deep-learning tasks require, potentially improving efficiency and reducing both costs and development times.

This move is part of Alibaba’s larger push to grow its cloud services and AI products, positioning itself as a global rival to Amazon Web Services and Microsoft Azure. By combining custom hardware with proprietary software, Alibaba aims to build an integrated ecosystem offering high performance at scale.

Baidu’s Kunlun Processors Gain Traction

• Baidu, whose strength lies in search and autonomous driving technologies, is doubling down on custom chips through its Kunlun processor line.
• The company unveiled the first generation of Kunlun AI chips in 2018 and has since followed up with improved iterations for both data-center and edge AI applications.
• The newest version of the Kunlun chips is reportedly being used to train Baidu’s sophisticated generative AI models, including its flagship Ernie Bot.

“Kunlun is the result of years of work on streamlining AI hardware,” said a Baidu spokesman in a previous briefing. “We want to accelerate AI processing power so that it fits the workloads we are seeing—from natural language understanding to computer vision and autonomous driving.”

Meeting Soaring AI Demand

The move toward developing in-house chips comes as companies across industries scramble to meet the soaring demand for AI capabilities. Training state-of-the-art models—whether for language, vision, or multimodal tasks—requires immense computational resources.

• High-end AI chips, sometimes called GPUs or AI accelerators, crunch complex matrix operations at great speed.
• By relying on chips of their own design, Alibaba and Baidu can exert greater control over supply chains and reduce reliance on overseas semiconductor technology.
• This strategy also allows them to fine-tune hardware for specific AI tasks, potentially lowering operating expenses by eliminating certain licensing and procurement fees.

The Impact on the Global Semiconductor Landscape

This development reflects a wider trend: Chinese companies are accelerating efforts to achieve semiconductor independence.

• China’s government has poured large sums into domestic chip design and fabrication in recent years, viewing it as a strategic priority for economic and national security.
• If Alibaba and Baidu succeed in scaling their in-house chips for AI training, the competitive landscape of global silicon companies could shift, with U.S. firms like Nvidia and AMD potentially losing market share in China.

However, manufacturing high-end chips remains a formidable challenge:

• Advanced semiconductor production still relies heavily on global supply chains for specialized equipment and materials.
• While Alibaba and Baidu can design their own processors, mass production often requires partnerships with foundries like Taiwan Semiconductor Manufacturing Company (TSMC) or other contract manufacturers.

A Spark for Innovation in AI

Beyond supply chain concerns, designing custom chips provides technological advantages:

• Specialized processors can be optimized for specific machine learning architectures, delivering greater energy efficiency and higher throughput.
• Faster training times and more advanced AI models could give Alibaba and Baidu an edge in applications like real-time translation, autonomous systems, and generative content creation.

Vertical integration—controlling both the hardware and the AI software stack—ensures a closer match between chip capabilities and model requirements. This synergy can lead to breakthroughs in performance and cost-effectiveness that off-the-shelf chips may not achieve.

Challenges and Next Steps

Despite the promise, Alibaba and Baidu face significant hurdles:

• Designing high-performance AI chips is hugely expensive, requiring top-tier expertise and years of research and testing.
• Fabrication at cutting-edge process nodes (e.g., 5 nm or below) is a complex and costly endeavor.
• Scaling production to meet the massive and growing AI computing demands across their services will require mass production and consistent quality.

Analysts emphasize that global collaboration remains crucial.

• Even as Chinese companies progress in design, the most advanced manufacturing technologies—such as extreme ultraviolet (EUV) lithography—are still dominated by a few foreign suppliers.
• Balancing these dependencies will be a key challenge for future bilateral AI chip programs.

New Age for China’s Artificial Intelligence Ambitions

Alibaba and Baidu’s embrace of homegrown chips reflects a broader story: China’s tech giants are determined to set their own course in the global AI race.

• Investing in domestic semiconductor technology allows them to address immediate supply constraints and ensure long-term innovation.
• With AI increasingly driving economic growth and national competitiveness, control over underlying hardware is now a strategic necessity.
• The success of Alibaba and Baidu may inspire other Chinese companies—and perhaps firms worldwide—to pursue vertical integration in AI hardware and software.