Arm has announced a new mobile platform designed to run sophisticated AI agents directly on smartphones while delivering desktop-class gaming graphics, marking a significant shift toward processing artificial intelligence at the device level rather than in distant data centers. The Compute Subsystem (CSS) for Mobile 2, unveiled at Arm Everywhere China in Shanghai, introduces faster processors, doubled AI acceleration for graphics, and neural upscaling technology that reconstructs game visuals from minimal pixel data.

Why Is Arm Betting on Local AI Processing?

The smartphone industry is experiencing a fundamental transition. Rather than sending every user interaction to cloud servers and waiting for a response, Arm argues that AI workloads should increasingly run on devices themselves. This shift addresses real-world problems: latency delays, connectivity dependencies, and the enormous cost of running services at global scale.

"When every interaction has to travel to a datacenter and back, there's implications for latency, connectivity, the cost of delivering those services at enormous scale, so this is where we're seeing it becoming more distributed, and I think few doubt now that basically everything you can push to the edge you will," said Chris Bergey, Executive Vice President of Edge AI at Arm.

Chris Bergey, Executive Vice President of Edge AI at Arm

Agentic AI represents a new category of artificial intelligence that goes beyond simple voice recognition or image filtering. These systems interpret user objectives, coordinate multiple applications and models simultaneously, and execute sequences of actions without constant human guidance. Running this type of AI locally requires more than just a faster processor; it demands multiple high-performance cores working in parallel.

What Hardware Changes Enable This Shift?

The CSS for Mobile 2 introduces several architectural improvements designed specifically for local AI workloads. The flagship processor cluster includes a new top-tier core called the C2-Ultra, which Arm claims delivers up to 15 percent higher single-thread performance compared to last year's C1-Ultra.

Arm's chipmaker partners can customize the processor configuration, but the recommended setup includes:

• CPU Cores: Two C2-Ultra high-performance cores paired with six efficiency-focused C2-Pro cores for handling background tasks without draining battery
• AI Acceleration: Two Scalable Matrix Extensions (SME2) units, double the previous generation, enabling up to 70 percent speedup on small language models
• Memory Architecture: Private caches on individual processors plus a large shared L3 cache, all connected through the DynamIQ Shared Unit to minimize latency during concurrent workloads

These improvements are specifically tuned to keep multiple AI tasks responsive simultaneously, a critical requirement for agentic systems that must juggle several operations at once.

How Does Neural Graphics Transform Mobile Gaming?

Perhaps the most striking innovation in CSS for Mobile 2 is the Mali G2-Ultra NX GPU, which Arm describes as its first AI-native graphics processor. Rather than rendering every pixel in a game scene, this GPU uses neural networks to intelligently reconstruct visuals from partial data.

The technology relies on two complementary techniques. Neural Super Sampling upscales lower-resolution graphics, converting 540p images to 1080p quality. Neural Frame Rate Upscaling generates intermediate frames, effectively doubling the frame rate by having the neural hardware predict what pixels should appear between rendered frames. Combined, these techniques mean the GPU renders only one-eighth of the pixels in the final display sequence, with artificial intelligence reconstructing the rest.

"The G2-Ultra NX GPU is able to sustain a 30 frame per second gaming session on content with very rich and intricate ray traced scenery. And this is really enabled thanks to a 70 percent reduction in the overall ray tracing workload, which then translates to a 30 percent boost in performance," explained Deyan Lazarov, GPU Product Manager at Arm.

Deyan Lazarov, GPU Product Manager at Arm

The result is striking: desktop-quality graphics with full ray tracing (realistic lighting and shadows) running within a 1-watt power budget. Arm also demonstrated other games running at 60 frames per second using the same technology. For context, a 1-watt power draw is negligible compared to a smartphone's total battery consumption, meaning these visual improvements come essentially free in terms of battery life.

When Will These Chips Reach Consumers?

The CSS for Mobile 2 represents Arm's reference design, not a finished product. Chipmakers must now license these designs and manufacture actual silicon, a process that typically takes several months. Smartphone manufacturers will then integrate these chips into devices. Industry observers expect the first phones featuring this platform to arrive as early as next year.

To accelerate adoption, Arm has published its neural graphics software development kit and sample code on GitHub, supporting the cross-platform Vulkan graphics API. This gives game developers time to optimize their titles for the new hardware before devices hit the market.

The shift toward local AI processing represents a fundamental change in how smartphones will function. Rather than devices serving primarily as portals to cloud services, they will become increasingly autonomous, capable of running sophisticated AI agents and delivering rich experiences without constant internet connectivity. Arm's new platform signals that this transition is no longer theoretical; it is becoming hardware reality.