Qualcomm is shoving AI cores straight into its next premium Snapdragon GPU.
According to 9to5Google, Qualcomm has announced its next-generation Adreno Neural Fusion GPU, following its earlier outline of the Oryon CPU for its next flagship Snapdragon chip.
The new Snapdragon system-on-chip will be the first from Qualcomm to put dedicated AI GPU cores into the graphics pipeline. These Matrix Cores sit inside each of the three 1.45 GHz GPU slices in the new Adreno design.
Qualcomm said the Matrix Cores provide dedicated local hardware to “run advanced AI models directly inside the graphics pipeline”.
The Adreno GPU will also include a dedicated 18 MB High-Performance Memory cache. That does not sound enormous, but Qualcomm claims it gives the GPU low-latency memory for tile-based rendering, frame buffers, and compute workloads.
The company says the setup keeps more work within the graphics subsystem rather than constantly shuttling data to and from system memory. Less data traffic means lower latency, better efficiency and fewer excuses for a hot slab of glass in your hand.
“The trick is keeping the work close to the silicon. Adreno Matrix Cores handle AI models right where the graphics happen, and the 18MB of Adreno High Performance Memory holds the heavy data, tiles, frame buffers, and compute on-chip, so it isn’t constantly shuttling back and forth to system memory. Less movement means lower latency and better efficiency, which translates into gaming sessions that last longer,” Qualcomm said.
Qualcomm reckons the Adreno Neural Fusion GPU can deliver a 40 per cent power improvement when the technology is switched on. That is the sort of claim which sounds useful, provided real phones behave like the slide deck.
Unity and Unreal Engine have already integrated Qualcomm’s Neural Fusion stack. That gives developers a head start on using the new GPU features without waiting for the usual ecosystem faffing about.
Qualcomm said the promised result is better graphics through “upscaled images, stable frame rates, and improved efficiency through AI-enhanced rendering built directly into the Snapdragon graphics pipeline”. The company will say more about the chip’s neural processing unit at Snapdragon Summit, scheduled for 22-24 September 2026.
