Alibaba AI Chip Matches Nvidia H20, Boosting China's Domestic Tech

Alibaba Group Holding’s semiconductor design unit, T-Head, has developed an artificial intelligence chip with capabilities matching Nvidia’s H20 graphics processing unit (GPU), according to a China Central Television (CCTV) report. This development offers fresh evidence that Chinese developers are designing advanced chips intended to replace imports.

Credit: T-Head

The CCTV report, aired Tuesday, showcased T-Head’s PPU, an application-specific integrated circuit, during a performance benchmark where it was compared with Nvidia’s H20 and A800 GPUs. This comparison occurred during Premier Li Qiang’s visit to a China Unicom data centre in northwestern Qinghai province.

China Unicom briefed Premier Li on the use of mainland-developed chips within its infrastructure. This marked the first time a state broadcast highlighted Alibaba’s proficiency in semiconductor design.

Alibaba’s Hong Kong-listed shares closed 5.28% higher at HK$161.60 on Wednesday, following market attention to the CCTV report. Footage from CCTV displayed a chart comparing several locally designed AI accelerators with Nvidia’s two China-tailored GPUs, modified to comply with US tech export restrictions.

Credit: T-Head

T-Head’s PPU card, which included 96 gigabytes of high-bandwidth memory per unit, matched Nvidia’s H20 and surpassed Huawei Technologies’ Ascend 910B. The Alibaba PPU card also featured a chip-to-chip bandwidth of 700 gigabytes per second, a high-speed Peripheral Component Interconnect Express standard, and a 400-watt power consumption, which was lower than the H20’s requirement.

Another chart in the broadcast detailed China Unicom’s contracts with four domestic chip providers, totalling 22,832 cards that provided 3,579 petaflops of computing speed. T-Head’s PPU cards accounted for 16,384 units of these contracts, supplying 1,945 petaflops.

The other suppliers to China Unicom included MetaX, a GPU start-up that collaborates with the Chinese Academy of Sciences; Biren Technology, a Shanghai-based AI chip designer; and Zhonghao Xinying Technology. Zhonghao Xinying was founded by former Google tensor processing unit engineer Yanggong Yifan.

Other chip suppliers mentioned in the CCTV report as potential partners were Tecorigin, a company based in Wuxi in eastern Jiangsu province; Moore Threads Technology; and Tencent Holdings-backed Enflame. China’s growing number of mainland AI chip suppliers lends credence to Alibaba CEO Eddie Wu Yongming’s statement during the firm’s earnings call last month.

Wu had assured that the company prepared "backup plans" to secure AI chip supplies. This was in response to US restrictions and heightened geopolitical tensions. The rapid expansion of AI development projects has significantly increased China’s demand for high-performance computing.

This demand has prompted local governments and telecoms network operators to invest in large-scale data centres for AI-related projects. Premier Li visited China Unicom’s Sanjiangyuan Green Energy Intelligent Computing Centre, valued at 2.77 billion yuan (USD 389 million).

This centre, which broke ground in Aug. 2024, spans 5.3 hectares. It will be constructed in four phases, with a total capacity exceeding 20,000 petaflops upon completion, according to a report by the local Qinghai Daily.

---

• Alibaba's T-Head unit developed an AI chip, PPU, with capabilities matching Nvidia’s H20 GPU.

• A CCTV report showcased the PPU’s performance, indicating China’s progress in advanced domestic chip design.

• The PPU card matches Nvidia H20 performance and surpasses Huawei's Ascend 910B in certain metrics.

Source: SCMP

Alibaba Launches Open-Source AI Agent, Citing Efficiency and Performance Against OpenAI

Alibaba Group Holding has unveiled a new open-source artificial intelligence agent, which it states matches the performance of OpenAI’s flagship Deep Research tool while offering greater efficiency. The company describes its deep research agent as a “leading open-source deep research” tool.

Credit: ALIBABA

Alibaba stated its agent showed "incredible efficiency" compared with US proprietary tools because it has only 30 billion parameters. This is significantly fewer than the estimated parameter counts of the models that power US deep research agents.

The new agent achieved industry-leading scores across various advanced benchmarks, including Humanity’s Last Exam, according to a graphic released by Alibaba. This challenging set of academic questions tests the limits of existing AI systems.

Alibaba's agent scored 32.9% on this benchmark, surpassing OpenAI’s Deep Research, which achieved 26.6% earlier this year. Adina Yakefu, a machine learning community manager at open-source platform Hugging Face, described Alibaba’s self-reported benchmark scores as “amazing.”

The agent has been integrated into Alibaba’s maps application, Amap, and its AI-powered legal research tool, Tongyi FaRui. Users of Amap can leverage the agent’s web retrieval capabilities to plan multi-day trips.

Tongyi FaRui has been updated with the agent’s research functions, enhancing its ability to retrieve case law with verified citations, according to a blog post on Tuesday by Alibaba’s AI search development team, Tongyi Lab.

Credit: OpenAI

Deep research agents are AI tools designed to perform complex, multi-step web retrieval tasks. OpenAI launched the first such agent, Deep Research, integrating it into ChatGPT in Feb. Other major US technology organisations, including Google DeepMind, have also introduced similar tools.

Parameters are variables that encode an AI model’s “intelligence” and are adjusted during the training process. Generally, a higher number of parameters indicates a more powerful model, but it also requires more computational resources to train and operate.

The strength and efficiency of Alibaba’s agent stem from its innovative data curation pipeline, which produced “very high-quality” synthetic training data, according to Tan Sijun, an AI researcher at the Sky Computing Lab of University of California, Berkeley.

Synthetic training data is generated by AI systems instead of being sourced from the real world. As real-world data becomes increasingly scarce, AI companies are turning to synthetic data to train new systems.

Alibaba’s synthetic data solution was applied throughout the entire training pipeline and incorporated a new technique enabling a “data flywheel.” This process reuses data generated during training to enhance the model without human intervention.

Its developers wrote that this approach “ensures both exceptional data quality and massive scalability, breaking through the upper limits of model capabilities.” They noted this training pipeline has yet to be validated on base models significantly larger than 30 billion parameters.

The company also stated that the agent’s context length of 128,000 tokens remains a limitation for many complex research tasks that require long inputs.

---

• Alibaba Group Holding released an open-source AI agent that rivals OpenAI’s Deep Research tool in performance.

• The new agent claims "incredible efficiency," operating with 30 billion parameters, fewer than estimated counts for US competitors.

• It achieved a 32.9% score on Humanity’s Last Exam, surpassing OpenAI’s 26.6% score obtained earlier in the year.

Source: SCMP

Cupra Tavascan Review: One Of The Sportiest EVs We Have Ever Tried!

The Cupra Tavascan is the brand’s first all-electric SUV coupé, blending bold design with advanced technology to deliver a driving experience that feels both futuristic and emotional. Positioned as a performance-oriented EV, it aims to stand out in a crowded electric SUV market by combining distinctive styling, a luxurious and immersive interior, and a suite of smart features that elevate both safety and convenience.

On the outside, the Cupra Tavascan is a striking SUV-coupé that commands attention with its athletic stance and sculpted proportions. Its sharp character lines flow seamlessly across the body, giving it a sense of movement even while standing still.

The sleek profile is complemented by a coast-to-coast rear light bar that incorporates the glowing Cupra emblem, a signature touch that reinforces its electric identity. Built on Volkswagen’s MEB platform, the Tavascan combines form and function in a way that feels both futuristic and unmistakably Cupra.

The headlights are among its most defining features. Designed in the brand’s triangle motif, these “eye-like” light clusters give the car a fierce, almost jewel-like expression. Beyond aesthetics, they provide strong illumination with advanced LED technology, ensuring both visibility and presence on the road.

Interior

Stepping inside the Tavascan reveals a cabin that blends luxury with sustainability. High-quality materials such as plant-based textiles and recycled microfibers are used extensively, creating an atmosphere that feels premium without compromising on environmental responsibility.

The surfaces are soft to the touch, while copper accents highlight the brand’s distinctive design language. This is not just a practical interior; it feels like a modern lounge designed around the driver.

What sets the Tavascan apart is its unique central spine — a structural element finished in Cupra’s trademark copper. This unconventional design anchors the cockpit while accentuating its driver-centric layout. Everything is positioned to create a sense of focus and control, making the driving experience feel personal and immersive.

The interior lighting completes the atmosphere. Ambient lighting extends seamlessly across the dashboard, through the door panels, and even into the rear seats. Subtle details like illuminated cupholders and perforated panels with backlighting add depth and warmth, ensuring every passenger shares in the immersive environment, not just those in the front.

Features

Technology in the Tavascan is designed to be both intelligent and intuitive. One of its standout innovations is the smart lighting system, which goes beyond simple illumination to act as a communication tool.

This light provides real-time notifications, signaling everything from incoming phone calls and navigation prompts to battery charging status and even brake warnings. This transforms light into a functional interface between the car and its driver.

The Tavascan also comes equipped with a comprehensive suite of driver assistance systems. Parking is simplified through park assist technologies that can guide the vehicle into tight spaces.

Lane assist and Electronic Stability Control work together to keep the car steady and safe, reducing the risk of drifting or loss of traction. Meanwhile, adaptive cruise control and a built-in speed limiter help manage speed intelligently, particularly on longer journeys or in changing speed zones.

For those who enjoy tailoring their drive, multiple customizable drive modes are available, from efficient range settings to performance-oriented options like Cupra mode. On highways, Travel Assist combines steering guidance with adaptive cruise control to deliver semi-autonomous driving, keeping the car centered in its lane and at a safe distance from traffic. Altogether, these features strike a balance between safety, convenience, and driver engagement.

Side Assist uses radar sensors to monitor blind spots and approaching vehicles, giving drivers extra confidence when changing lanes. Complementing this is the Exit Warning system, which alerts occupants to oncoming traffic or cyclists when opening a door, helping prevent accidents in busy urban environments. For longer journeys, a Drowsiness Monitor keeps track of driver behavior, detecting signs of fatigue and prompting timely rest breaks to ensure safer, more focused driving.

The personal voice assistant adds a layer of convenience and connectivity to every drive. Activated by natural voice commands, it allows drivers to control navigation, climate settings, media, and even certain vehicle functions without taking their hands off the wheel.

Battery

Powering the Tavascan is a 77 kWh lithium-ion battery, available in two configurations. The Endurance version uses a rear-wheel drive setup with 286 horsepower, offering a WLTP range of up to around 534 kilometers. It accelerates from 0 to 100 km/h in about 6.8 seconds, making it brisk without sacrificing efficiency. For those seeking more power, the VZ model adds a second motor to deliver all-wheel drive and a combined output of 340 horsepower. Range in this version is slightly lower, at around 499 kilometers, but acceleration improves to 5.5 seconds for the 0–100 km/h sprint.

Charging is equally convenient. With DC fast charging at up to 135 kW, the Tavascan can be charged from 10 to 80 percent in roughly half an hour, while an 11 kW AC charger will fully recharge it overnight. This flexibility ensures that whether at home or on the road, the Tavascan remains practical and ready for long trips.

Conclusion

The Cupra Tavascan is more than just another electric SUV; it is a statement of design and technology working in harmony. Its bold exterior and signature triangular headlights give it a distinctive identity, while the luxurious and ambient-lit interior creates an inviting space for every journey.

With advanced driver assistance features, smart lighting that actively communicates with the driver, and customizable drive modes, the Tavascan offers a blend of safety, convenience, and enjoyment behind the wheel. Coupled with a capable battery and strong charging performance, it delivers the range and performance to match its striking looks. For drivers seeking an electric SUV that dares to stand apart, the Cupra Tavascan is a compelling choice.