Haimacloud releases hybrid portable console GEAR based on Snapdragon 865 and cloud RTX GPUs

At the Apsara Cloud 2026 conference, Alibaba Cloud presented its agent cloud strategy, which is based on three scenarios: model, management, and context. These scenarios correspond to three levels: AI Native Cloud, Agent Native Cloud, and Context Engine, according to the company's press release from September 22.

This review focuses primarily on the products of the agent platform and cloud infrastructure, rather than the roadmaps for training base models, which are covered in other materials.

At the management level, Alibaba Cloud announced AgentCore—an enterprise platform designed for creating, launching, and administering AI agents throughout their entire lifecycle. According to the company's materials, this platform allows organizations to gather tools, control human-agent interaction, track performance, and implement security at the operational level. This ensures the integration of diverse agents into critical business systems under a single controlled foundation. Additionally, the Agent Security Center was introduced, which ensures security and compliance throughout the entire lifecycle with real-time threat detection.

Chinese trade media commentary on the technical director of Alibaba Cloud indicated the vendor's claims that AgentCore can improve task completion rates and reduce total cost of ownership, although these figures remain company claims.

Regarding the model serving subsystem, Alibaba Cloud announced the launch of a next-generation Parallel File System (CPFS). This system provides throughput of hundreds of terabytes per second and one hundred million input/output operations per second, which, according to the company, can reduce enterprise AI storage costs by approximately 69 percent. In terms of internal storage, it was noted that the scale of one file system reached about 100 Petabytes, five times the previous figure. Furthermore, positive effects for the training process were noted, such as faster job startup and higher utilization of peak accelerators.

The HPN 8.0 Pro network architecture was separately presented, representing Alibaba Cloud's latest patented AI architecture. It is capable of providing 100 petabits of throughput with ultra-low latency. One cluster can support over 130,000 network ports at 800G speed, while redundancy ensures that failures in optical transceivers or communication lines do not interrupt service. The impact of updates and network failures has been reduced from 50 percent to 25 percent compared to the previous generation.

Collectively, the combination of AgentCore, next-generation CPFS, and HPN 8.0 Pro forms a specific product triad of cloud solutions that Alibaba Cloud demonstrated for scalable agent workloads, independent of chip announcements or large language models made during the same week of the conference.

Qualcomm has unveiled two new high-performance processors intended for Android smartphones: the Snapdragon 8 Elite Gen 6 and the Snapdragon 8 Elite Extreme Gen 6. These chips were presented during the Snapdragon Summit 2026 and mark a shift in the company's strategy, which now offers two options within its main line of mobile platforms.

Both models are manufactured using the 2-nanometer (nm) process and incorporate the new generation of the Oryon CPU, confirming recent rumors. The internal configuration includes two high-performance cores operating up to 5 GHz and six performance cores reaching 4 GHz.

A central point of innovation is the focus on processing artificial intelligence (AI) directly on the device. The company designed these components to execute AI tasks more sophisticatedly locally, eliminating, in many cases, the dependence on cloud processing.

Improvements in Artificial Intelligence Processing

The two chips come equipped with an updated version of the Hexagon NPU, the unit dedicated to AI processing. A notable advancement is the Hexagon Element Accelerator, developed specifically to optimize workloads involving AI models.

In the specific case of the Snapdragon 8 Elite Extreme Gen 6, the NPU's shared memory is 50% superior to the previous generation. This capacity allows for the local execution of larger AI models, including those with a Mixture of Experts (MoE) architecture exceeding 30 billion parameters.

Qualcomm also emphasizes that the new processors were designed to reduce the need for external memory access during certain AI operations, thereby aiming to reduce both latency and power consumption.

Additionally, AI task processing is performed directly on the GPU. The Extreme model incorporates Adreno Matrix cores, which are dedicated to handling AI workloads within the graphics pipeline itself.

Generational Leap in GPU and Exclusive Features

Qualcomm classifies the new Adreno GPU as the company's biggest generational leap to date. In both chips, the architecture has been redesigned to elevate graphic performance and enhance energy efficiency.

However, the Extreme model brings exclusive functionalities, such as Adreno Neural Fusion. This technology uses AI to enhance features like supersampling and frame generation in compatible games.

By directing these functions to the graphics pipeline, Qualcomm claims that latency and energy expenditure can be reduced, preventing certain operations from needing to be transferred to other parts of the processor.

The Extreme also features three GPU cores dedicated to AI and high-performance Adreno memory. The company cited titles such as Honkai: Star Rail, Diablo Immortal, and War Thunder: Mobile as examples of games that can benefit from AI-enhanced rendering capabilities.

Both processors include the Qualcomm Spectra ISP, responsible for managing image processing resources, digital stabilization, color treatment, and motion detection.

Video Capabilities and Comparative Performance

The Snapdragon 8 Elite Extreme Gen 6 has extra video recording capabilities, offering support for 8K capture at 60 frames per second, in addition to advanced professional codecs like Advanced Professional Video (APV). This model also supports slow-motion recording at 4K at 240 fps.

According to data released by Qualcomm, the Snapdragon 8 Elite Gen 6 presents 10% more CPU performance and 37% greater energy efficiency compared to the Snapdragon 8 Elite Gen 5. For the Snapdragon 8 Elite Extreme Gen 6, the CPU performance increase reaches 13% over the previous generation.

The new Adreno GPU demonstrates a gain of up to 35% in graphic performance and 40% in energy efficiency compared to the Snapdragon 8 Elite Gen 5. In the Extreme model, this graphic gain reaches 44%.

Regarding the NPU, the Snapdragon 8 Elite Gen 6 offers 14% more performance and 20% more AI performance per watt. For the Extreme, the gains are 35% in NPU performance and 33% in AI performance per watt, all these figures based on measurements provided by Qualcomm itself.

Availability and Initial Models

The first devices officially announced with these new processors are from Motorola and Xiaomi. The Motorola Signature 27 and the Xiaomi 18 Pro Max will be equipped with the Snapdragon 8 Elite Extreme Gen 6, while the Xiaomi 18 Pro will feature the Snapdragon 8 Elite Gen 6.

Although other models have not yet been confirmed, TechRadar listed potential candidates, such as the Galaxy S27 series (including S27 Ultra and S27 Plus), as well as the Galaxy Z Fold 9 and Z Flip 9. The speculative list also includes models like Xiaomi 18 Ultra, Oppo Find X10 Ultra, Find X10 Pro Max and Find X10, Honor Magic 10 Pro and Magic V7, along with RedMagic 12 Pro and 12S Pro.

Qualcomm guaranteed that the devices equipped with the two new chips will begin rolling out in the global market in the coming weeks.