Alibaba is reportedly internally testing a new artificial intelligence inference chip manufactured domestically, marking the latest chapter in its seven-year semiconductor journey that continues to face market skepticism despite substantial investments. The move comes as the Chinese tech giant prepares to report first-quarter fiscal 2026 earnings and seeks to address gaps in large language model inference and cloud computing capacity.

The company has invested over 100 billion yuan (14 billion) in AI infrastructure and product development over the past four quarters, with portions allocated to chip research. In February, Alibaba announced plans to invest more than 380 billion yuan (53 billion) over three years to build cloud and AI hardware infrastructure, including semiconductor development.

The new chip represents a strategic shift from overseas foundries to domestic manufacturing, reflecting efforts to reduce supply chain dependencies amid ongoing US-China technology tensions. However, questions persist about whether Alibaba can achieve meaningful commercial scale and performance competitiveness in the high-barrier semiconductor market.

Industry observers remain cautious about the chip's potential impact, noting that previous Alibaba semiconductors have struggled to move beyond limited deployment despite impressive benchmark claims.

Platform Strategy Shows Mixed Results

Alibaba's semiconductor ambitions began in 2018 with the acquisition of C-Sky Microsystems and the establishment of T-Head Semiconductor under its Damo Academy research division. The initiative aimed to create a complete in-house development pipeline from architecture design to commercial application.

The company's first major product, the Hanguang 800 AI inference chip launched in 2019, claimed peak performance of 78,563 images per second on ResNet-50 tasks with 500 IPS/W efficiency—reportedly four times and 3.3 times better than industry competitors respectively. Alibaba demonstrated the chip's capabilities in Hangzhou's City Brain project, where four Hanguang 800 units allegedly replaced 40 traditional GPUs while reducing latency from 300ms to 150ms.

However, industry experts noted that such benchmark tests often involve targeted optimizations for specific network models, representing "point testing" results rather than performance across diverse, heterogeneous production workloads.

Pivot to Server Processors Reflects Cloud Focus

In 2021, Alibaba unveiled the Yitian 710, a 128-core Armv9 architecture processor targeting general-purpose server applications rather than AI inference. The shift reflected the company's strategy to closely integrate its semiconductor development with its cloud computing business as competition intensified in that sector.

Alibaba Cloud subsequently launched Elastic Compute Service instances based on Yitian processors, allowing customers to select these options through the control panel. While initial supply remained limited, the move demonstrated progression toward commercial deployment.

Nevertheless, external skepticism persisted. Beyond official demonstrations and select partnerships, the Hanguang 800 saw limited large-scale customer adoption. The Yitian 710, despite cloud instance availability, experienced slow progress in overall supply and customer adoption.

Manufacturing Shift Addresses Supply Chain Risks

The newly reported inference chip differs from previous products primarily in manufacturing approach and positioning. While the Hanguang 800 relied on overseas foundries, the latest chip utilizes domestic wafer fabrication facilities—a clear hedge against geopolitical risks and supply chain dependencies.

Critical questions remain about domestic foundry capacity, yield stability, and alignment with Alibaba's data center deployment timeline. These factors will determine whether the chip can transition from prototype testing to large-scale commercial availability.

Performance competitiveness presents another challenge. Even setting aside comparisons to Nvidia Corp., achieving AMD-level performance standards remains a long-term hurdle. While Yitian 710 demonstrated relatively strong power efficiency in certain database and web service tests, comprehensive performance across integer, floating-point, and virtualization workloads requires validation.

Ecosystem Development Remains Key Challenge

Industry insiders emphasize that hardware performance parity alone cannot ensure market success without mature software support infrastructure. Alibaba must address compatibility and migration costs for independent software vendors and mainstream frameworks—representing significant software-layer barriers to adoption.

Long-term success requires building comprehensive ecosystems beyond hardware design, including stable compilers, operator libraries, developer toolchains, and third-party application support. Nvidia's market dominance stems partly from sustained investment in comprehensive development ecosystems, particularly its CUDA platform.

According to Neil Shah, vice president at Counterpoint Research, Nvidia's leading position derives not only from hardware but also from its robust ecosystem, especially CUDA's seamless compatibility layers that enable developers to easily port existing code.

Alibaba has attempted ecosystem development through T-Head's open-source Xuantie IP and toolchains, but these efforts primarily target Internet of Things and edge computing applications rather than data center-scale AI computing ecosystems. The fundamental challenge lies in internet companies' focus on asset-light, rapid-iteration business models conflicting with semiconductors' long development cycles, high capital requirements, and elevated risks.

Some senior investors view Alibaba's semiconductor strategy as primarily defensive rather than commercially driven, noting that chips are unlikely to become independent profit centers in the near term. The company has not separately reported semiconductor business revenue and investments in financial statements, suggesting strategic rather than market-expansion positioning.

This defensive logic serves supply chain risk management more than technological breakthrough, with Alibaba's chips primarily valued for reducing cloud computing costs, improving supply stability, and providing strategic buffer amid intensifying US-China technology friction.