Google’s Tensor chips have often felt like they’re built to be “good enough” rather than genuinely exciting. The Tensor G5 fits that reputation: capable, reliable, and tuned heavily around Google’s software and AI goals—but rarely the kind of processor that turns heads in benchmark charts. By comparison, MediaTek’s Dimensity 9500 has been getting attention for pushing harder on modern CPU design choices and delivering stronger overall performance across tests.
That performance gap appears to be influencing Google’s next steps. Early details suggest the upcoming Tensor G6, internally known as “Malibu,” will adopt a couple of key ideas that helped rival chips pull ahead—especially on CPU design and task efficiency. The Tensor G6 is expected to arrive in the second half of 2026 and may be manufactured using TSMC’s 2nm process, a move that could provide meaningful gains in efficiency and speed if the rest of the design comes together.
A quick look at what separates Tensor G5 and Dimensity 9500 today helps explain why these rumored Tensor G6 changes matter.
On the CPU side, both chips use an 8-core setup, but their core choices and tuning differ significantly. Dimensity 9500 uses a high-clocked prime core and a modern mix of performance-focused cores, while Tensor G5 relies on an ARM Cortex-X4 prime core and a combination of Cortex-A725 and Cortex-A520 cores. Cache details for Tensor G5 haven’t been publicly detailed, but the broader takeaway is that MediaTek leaned harder into newer core designs and aggressive performance targets.
Graphics is another dividing line. Dimensity 9500 uses an ARM Mali-G1 Ultra MC12 GPU with improved ray-tracing performance and smooth high-frame-rate gaming potential, including 120fps gameplay scenarios. Tensor G5, meanwhile, uses an Imagination IMG DXT-48-1536 GPU and does not include ray-tracing support.
For AI, both companies lean into custom acceleration. Dimensity 9500 includes MediaTek’s NPU 990 for AI and machine learning workloads, while Tensor G5 uses Google’s custom TPU for AI tasks—an area where Google typically focuses much of its silicon identity.
Now to what’s being rumored for Tensor G6—and why it could be a notable shift for Pixel performance and efficiency.
Tensor G6 CPU: a reshaped 8-core strategy
The expected CPU layout for Tensor G6 is:
– 1 ARM Cortex-X930 “Super Core”
– 6 ARM Cortex-A730 performance “Big” cores
– 1 ARM Cortex-A5xx-series efficiency “Little” core
This is a clear change from Tensor G5’s 1+5+2 approach. The big idea: fewer efficiency cores, more performance cores. If accurate, it suggests Google is rethinking how it balances sustained performance, responsiveness, and power draw. Combined with an advanced 2nm manufacturing process, this configuration could deliver better real-world speed—especially in heavier multitasking and longer workloads where mid-cores matter.
Why fewer efficiency cores could help
One of the most interesting rumored lessons Google seems to be adopting is that efficiency cores aren’t always the magic solution they’re marketed as. By shifting from two small efficiency cores down to one and adding an extra “Big” core instead, Tensor G6 could improve performance headroom without relying so heavily on slower cores for everyday tasks.
In practice, this may boost the chip’s ability to stay smooth under load—like when you’re bouncing between apps, editing photos, using on-device AI features, or gaming—while still leaning on process improvements to keep battery impact under control.
A move toward newer ARM CPU cores
Another rumored shift is even more straightforward: using newer CPU cores closer to the cutting edge.
A major criticism of Tensor G5 is that its CPU core choices lag behind the latest available designs, which can contribute to noticeable benchmark and efficiency disadvantages compared to rivals that move faster. Tensor G6 is expected to feature the Cortex-X930, a core that hasn’t yet become widely available in commercial devices. If Google follows through, it could remove one of the biggest reasons Tensor chips have struggled to keep pace on raw CPU performance.
Tensor G6 GPU: a surprising step backward?
While the CPU story sounds more ambitious, the GPU rumor is puzzling. Tensor G6 is expected to use a 3-core Imagination IMG CXT GPU—described as older than the IMG DXT GPU used in Tensor G5.
If that holds true, it could limit gaming performance improvements and potentially widen the gap versus competitors prioritizing modern graphics features like ray tracing and higher-end GPU pipelines. It also raises questions about Google’s priorities: the company may be optimizing around power efficiency, cost, supply considerations, or specific workloads rather than peak graphics performance.
AI on Tensor G6: two-tier TPU approach
Tensor G6 is expected to continue using a custom TPU for major AI workloads, but it may add a second, smaller “nano-TPU” designed to handle simpler AI tasks more efficiently.
That kind of split could be meaningful for battery life and responsiveness, especially as phones increasingly run lightweight AI features constantly in the background. If Google can offload routine tasks to a low-power accelerator while reserving the main TPU for heavier operations, Pixel devices could feel faster in day-to-day AI features without paying as much in energy use.
Modem upgrade: a possible shift to MediaTek
Connectivity could also see a big change. Tensor G6 is expected to move away from a Samsung modem and adopt MediaTek’s M90 modem, with reported downlink speeds up to 12Gbps.
If accurate, this could translate into stronger real-world cellular performance depending on carrier support and implementation—potentially improving speed, stability, and efficiency, which have been pain points for some users on past generations.
What it all means for Tensor G6 and future Pixel phones
Based on these rumored details, Tensor G6 looks like a chip with clearer intent on CPU modernization and smarter AI efficiency—even if the GPU choice seems unexpectedly conservative. If Google truly pairs a more aggressive core configuration with the latest ARM designs and a TSMC 2nm manufacturing process, Tensor G6 could be positioned as a more competitive flagship processor than recent Tensor generations.
The big question will be whether Google’s decisions—especially around graphics—support the kind of performance jump Android users expect by 2026, or whether the focus remains centered on AI features and everyday smoothness over peak gaming and benchmark dominance.
