Arm used September 8, 2026 to unveil its most aggressive mobile graphics pitch yet: the C2-Ultra CPU core and the Mali G2-Ultra NX GPU, bundled into a new reference platform called CSS for Mobile 2. The company’s pitch is blunt. Android phones, it says, are close to running games with “desktop-class” visuals inside a power budget that fits in your pocket. That’s a bold claim for a chip that hasn’t shipped in a single flagship phone outside China yet.
The bigger story isn’t the silicon itself. It’s who hasn’t said yes to it. Qualcomm and MediaTek, the two companies that design the chips inside most of the world’s Android phones, have not announced products built on CSS for Mobile 2. That gap between Arm’s announcement and its licensees’ silence is the real news here, and it says as much about the state of the Android chip market as the specs themselves.
What Arm Actually Announced on September 8
Arm’s announcement centers on two new CPU cores, C2-Ultra and C2-Pro, which sit inside a new C2 CPU cluster. C2-Ultra is positioned as the top-end core in that cluster, aimed at flagship-tier workloads, while C2-Pro handles the tier below it. Both are part of the company’s next-generation smartphone reference design, CSS for Mobile 2.
Paired with the CPU cluster is the Mali G2-Ultra NX GPU, which Arm describes as its first “AI-native” Mali design. The distinction matters. Instead of treating AI workloads as something handled by a separate neural processing unit, Arm has built neural acceleration into the graphics pipeline itself, which is the same architectural direction Qualcomm has taken with its own GPU roadmap. CSS for Mobile 2 wraps the CPU cluster and the GPU into a single licensable package that chipmakers can drop into a system-on-chip design rather than assembling the pieces themselves.
The Performance Numbers Arm Is Standing Behind
Arm’s official claims are specific, and they’re all measured against last year’s C1-Ultra core rather than against any Qualcomm or Apple silicon. The company says C2-Ultra delivers up to 15% higher single-thread performance than C1-Ultra, with multi-thread performance up 12% over the prior generation. For AI workloads specifically, Arm claims up to 1.7x higher performance across current AI models, a jump it credits largely to a CSS for Mobile 2 configuration that carries two SME2 (Scalable Matrix Extension 2) units, double the SME2 capacity of the C1-Ultra configuration it replaces.
On the graphics side, Arm says the Mali G2-Ultra NX GPU can push desktop-class visuals within a 1-watt power budget, a figure the company leans on heavily in its marketing. Coverage of the launch has also cited a 14% overall GPU performance gain in certain benchmark comparisons against the prior generation, alongside AI features that let the GPU generate intermediate frames for smoother motion and upscale lower-resolution game output to sharper visuals. One report on the launch also cites a 38% power reduction tied to the new CPU cluster versus its predecessor, though the exact test conditions behind that figure haven’t been independently published.
How the AI Gains Actually Break Down
The SME2 doubling is the detail worth sitting with. Scalable Matrix Extension units handle the matrix math that underpins most on-device AI inference, from camera processing to the small language models phone makers now bake into their assistants. Doubling that capacity is less about gaming frame rates and more about Arm’s bet that phones will run heavier AI workloads locally instead of routing them to the cloud. Arm frames the gaming boost and the AI boost as two branches of the same investment, since the neural accelerators tucked into the GPU’s shader cores serve both jobs at once.
CSS for Mobile 2 Explained: A Platform, Not a Product
It helps to be precise about what Arm sold on September 8. CSS for Mobile 2 is not a chip you can buy. It’s a reference platform, a bundle of IP that Arm licenses to chipmakers, who then combine it with their own modem, imaging, and connectivity blocks to build an actual system-on-chip. That’s the same model Arm used with its prior-generation Lumex CSS platform, which paired the C1-Ultra CPU cluster with the Mali G1-Ultra GPU roughly a year earlier.
The practical effect is a lag between Arm’s announcement and any phone landing on a store shelf. Chipmakers still have to tape out silicon, validate it, and hand finished chips to phone brands, a process that typically runs many months even once a licensee commits. That lag is exactly why the identity of Arm’s launch partners matters more than the announcement date.
The Adoption Gap: Why Qualcomm and MediaTek Stayed Quiet
Here’s the part that separates this launch from a typical spec-sheet update. Qualcomm and MediaTek, which together design the silicon inside the overwhelming majority of non-Apple flagship phones, have not announced any CSS for Mobile 2-based product. That’s a notable omission for a platform Arm is pitching as the future of Android gaming and on-device AI.
Part of the explanation is architectural independence. Qualcomm designs its own Adreno graphics cores rather than licensing Arm’s Mali GPU line, and it has already pushed its own AI-native GPU concept, described in industry coverage as Adreno Neural Fusion, down a similar path of embedding neural acceleration directly into shader hardware. Seen that way, Mali G2-Ultra NX reads less like a platform Qualcomm might adopt wholesale and more like Arm’s answer to a fight Qualcomm already started. MediaTek’s position is murkier since it has historically mixed Arm CPU cores with a range of GPU suppliers, and its silence so far leaves open whether it’s evaluating CSS for Mobile 2, waiting on a later revision, or planning its own path entirely.
None of this means CSS for Mobile 2 will fail to find buyers. Samsung’s Exynos division and a handful of Chinese chipmakers have licensed Arm’s full platforms before. But a launch that leans this hard on the phrase “desktop-class” invites scrutiny of exactly which phones will carry the label, and as of this week that list has exactly one confirmed entry.
One Phone, One Country: China’s Early Lead
According to launch-week coverage from Notebookcheck, a single phone built on CSS for Mobile 2 with the Mali G2-Ultra NX GPU is already on sale, and it’s sold exclusively in mainland China. The chipmaker behind that specific device hasn’t been named in the reporting that’s surfaced so far, but its existence confirms Arm’s platform has cleared the gap from announcement to shipping silicon faster than most reference designs manage.
That’s a meaningfully different launch pattern than Arm’s Western-market-facing releases usually follow. It also lines up with a broader trend: Chinese SoC vendors have moved faster on Arm’s newest CPU and GPU IP over the past two product cycles than Qualcomm or MediaTek, partly because domestic phone brands are under more pressure to differentiate on specs given tighter access to some overseas components.
What “Weeks Away” Really Means for Global Buyers
Coverage tied to the announcement suggests additional CSS for Mobile 2 devices are expected within weeks, which points to a window of late September into October 2026 for the next wave. But “additional devices” and “a global flagship from a top-five Android brand” are different claims. Until Qualcomm, MediaTek, or Samsung’s Exynos team confirms a product, buyers outside China should treat “desktop-class Android gaming” as a feature they’ll likely wait until 2027 to hold in their hands.
Generational Leap: C2-Ultra and Mali G2-Ultra NX vs the Prior Platform
The table below lays out Arm’s own comparison points for CSS for Mobile 2 against the CSS for Mobile platform it replaces, which paired the C1-Ultra CPU cluster with the Mali G1-Ultra GPU.
| Metric | Prior Generation (C1-Ultra / Mali G1-Ultra) | CSS for Mobile 2 (C2-Ultra / Mali G2-Ultra NX) |
|---|---|---|
| Single-thread CPU performance | Baseline | Up to 15% higher |
| Multi-thread CPU performance | Baseline | Up to 12% higher |
| AI model performance | Baseline (1x SME2 unit) | Up to 1.7x higher (2x SME2 units) |
| GPU performance (reported) | Baseline | Up to 14% higher |
| GPU architecture | Traditional shader pipeline | “AI-native” shaders with embedded neural accelerators |
| Target graphics power envelope | Not disclosed by Arm at this detail | Desktop-class visuals within a 1W budget (Arm claim) |
| CPU power draw | Baseline | Up to 38% lower (per one launch report) |
Mali G2-Ultra NX vs the Field: A Qualitative Comparison
Arm has not published, and no outlet covering the launch has produced, direct frame-rate or benchmark comparisons between Mali G2-Ultra NX and Qualcomm’s Adreno GPUs or Apple’s A-series and M-series graphics. What exists instead is a feature-level comparison, since all three companies are now converging on the same idea: put AI acceleration inside the graphics pipeline rather than bolting it on as a separate chip.
| Vendor / Platform | AI-native GPU approach | Confirmed CSS for Mobile 2-class product | Public launch status |
|---|---|---|---|
| Arm (Mali G2-Ultra NX) | Neural accelerators embedded in shader cores | Yes (Arm’s own IP) | Announced Sept. 8, 2026; one China-exclusive phone shipping |
| Qualcomm (Snapdragon / Adreno) | Custom Adreno GPU with its own neural rendering direction | Not announced | No public commitment to Arm’s platform |
| MediaTek (Dimensity) | Historically mixes Arm CPU IP with varied GPU suppliers | Not announced | No public commitment to Arm’s platform |
| Apple (A-series / M-series) | Proprietary GPU, not Arm-licensed | Not applicable | Not part of Arm’s licensee ecosystem |
Read plainly, that table shows an industry converging on the same architectural idea from three different starting points, with Arm the only one of the four willing to put a number and a launch date on the record so far.
A Short History of Arm’s Mobile GPU Push
Arm’s move toward AI-embedded graphics didn’t start this month. The company’s prior CSS for Mobile platform, built around the C1-Ultra CPU cluster and the Mali G1-Ultra GPU, laid the groundwork roughly a year earlier and established the CSS licensing model that CSS for Mobile 2 now extends. Before that, Arm’s mobile GPU roadmap moved in smaller, single-digit-percentage steps for several generations, the kind of incremental gains that rarely made headlines outside chip-enthusiast forums.
What’s changed is the framing. Where past Mali launches sold themselves on raw triangle throughput and power efficiency, Arm is now selling an entire experience category, “desktop-class gaming,” borrowed directly from PC marketing language around ray tracing and AI upscaling. That shift mirrors what Nvidia did with DLSS on PC GPUs: reframe a generational hardware bump as an experience unlocked by software running on new silicon. Arm is betting the same playbook works on a phone screen.
What This Means for Android Flagship Pricing and Timing
Arm’s announcement lands at an awkward moment for the broader chip supply chain. Memory pricing has been climbing across the industry through 2026, and any chipmaker building a CSS for Mobile 2 device has to absorb both the licensing cost of the new platform and whatever premium memory suppliers are charging this quarter. That combination makes it more likely that the first non-Chinese phones carrying this silicon land as premium, higher-margin flagships rather than mid-range volume sellers, at least for the first product cycle.
There’s also a knock-on effect for game developers targeting Android. A GPU architecture built around AI upscaling and frame generation only pays off if studios actually build for it, the same adoption curve PC gaming went through with DLSS and FSR over several years. Engine support matters more than raw silicon here. If Unreal Engine and Unity ship first-class support for Arm’s neural rendering features quickly, adoption among mobile studios could move faster than the hardware rollout itself. If that support lags, Mali G2-Ultra NX risks becoming a spec-sheet bullet point most games never actually exercise.
Game Developers and the Unreal Engine Angle
Mobile studios have spent the past several years optimizing around a fairly narrow band of Arm CPU IP capability, since fragmentation across hundreds of Android chip variants makes it risky to lean too hard on any single vendor’s features. A widely discussed Hacker News thread on the launch raised the same fragmentation concern, with commenters noting that Arm’s GPU IP is now the dominant graphics option across Android excluding Qualcomm’s own Adreno line. Arm’s neural upscaling and frame-generation tools change that calculus only if enough of the installed base can run them. With adoption currently sitting at one phone in one country, most studios have little incentive to build Mali G2-Ultra NX-specific code paths yet. That could shift quickly if Samsung, Qualcomm, or a major Chinese brand like Xiaomi, Oppo, or Vivo commits to the platform for a 2027 flagship, since Android game engines tend to chase whichever hardware baseline reaches the widest audience fastest.
Five Predictions for the Next Two Quarters
- Expect at least one additional Chinese SoC vendor to confirm a CSS for Mobile 2 design before the end of 2026, extending Arm’s early lead in that market.
- Qualcomm is unlikely to license Mali G2-Ultra NX outright, but watch for Adreno feature announcements that mirror Arm’s neural-shader pitch within the next two Snapdragon generations.
- MediaTek is the swing vote to watch. A Dimensity chip built on CSS for Mobile 2 would be the strongest signal yet that Arm’s platform has real momentum outside China.
- Game engine support will lag hardware availability by at least two to three quarters, consistent with how DLSS and FSR adoption played out on PC.
- Pricing pressure from the ongoing memory shortage will push the first global CSS for Mobile 2 phones toward the premium tier rather than mid-range devices, at least through the first half of 2027.
The Open Questions That Could Make or Break This Launch
Arm’s marketing is ahead of its shipping hardware, which isn’t unusual for a platform license but does leave real gaps. Nobody outside Arm has independently verified the 1-watt desktop-class graphics claim under real gameplay conditions. Nobody has published a head-to-head benchmark against current Snapdragon or Apple silicon. And the identity of the Chinese chipmaker behind the first shipping phone remains unconfirmed in public reporting, which makes it hard to judge how representative that device is of what a global flagship version might deliver.
The biggest open question is simply whether Qualcomm and MediaTek’s silence is strategic patience or a signal they see limited value in Arm’s specific approach. Both companies have their own AI-graphics roadmaps in motion, and platform licensing decisions like this typically take one to two years to show up in shipping silicon. Arm won’t have a clear answer on broad Android adoption until well into 2027.
Frequently Asked Questions
What are Arm’s C2-Ultra and Mali G2-Ultra NX?
C2-Ultra is Arm’s new top-end CPU core for Android chips, part of a broader C2 CPU cluster that also includes the C2-Pro core. Mali G2-Ultra NX is Arm’s new GPU, described as its first “AI-native” Mali design because it embeds neural acceleration directly into its shader cores. Both are packaged together in a reference platform Arm calls CSS for Mobile 2.
Is CSS for Mobile 2 a chip I can buy?
No. It’s a licensable platform that chipmakers like Qualcomm, MediaTek, or Samsung’s Exynos team would need to build into an actual system-on-chip before it reaches a phone. Arm sells the IP. Phone brands buy finished chips from licensees, not from Arm directly.
Has Qualcomm or MediaTek confirmed using this platform?
No. As of this launch, neither company has announced a product built on CSS for Mobile 2. Qualcomm designs its own Adreno GPUs rather than licensing Mali, which makes broad adoption of the GPU portion less likely even if it licenses Arm’s CPU cores.
Is any phone with this chip available yet?
Yes, according to launch-week reporting, one phone built on CSS for Mobile 2 with the Mali G2-Ultra NX GPU is already on sale, but only in mainland China. The chipmaker behind that specific device has not been named in public coverage.
What does “desktop-class” graphics actually mean here?
It’s Arm’s marketing shorthand for AI-assisted rendering techniques, frame generation and upscaling, that let a phone GPU approximate visual quality closer to a gaming PC while staying inside a roughly 1-watt power budget, according to Arm’s own claims. It doesn’t mean the GPU matches PC hardware in raw computational throughput.
How does this compare to Qualcomm’s Adreno Neural Fusion?
Both approaches embed AI acceleration directly into the GPU’s shader hardware rather than routing AI tasks to a separate chip, and industry coverage has framed Mali G2-Ultra NX as Arm’s answer to Qualcomm’s direction. Neither company has published benchmarks comparing the two designs head-to-head.
When will global flagship phones with this chip ship?
Coverage tied to the announcement points to additional CSS for Mobile 2 devices arriving within weeks of the September 8 reveal, but that likely refers to further regional launches rather than a global flagship from a top Android brand. Broader adoption from Qualcomm, MediaTek, or Samsung remains unconfirmed and would more realistically land in 2027 product cycles.
Will this make Android games look like PC games?
Only if game engines and individual studios build support for the new neural rendering features, which historically takes several product cycles even after hardware ships. The bigger constraint right now is that only one confirmed phone runs this silicon, giving developers little reason to prioritize it yet.
