Smartphone cameras have improved dramatically over the past decade, but the gains are increasingly coming from software, not glass. Physics — not ambition — is now the real bottleneck, and manufacturers are running out of room to grow sensors inside phones that are also expected to stay thin, light, and affordable.
Sensor Sizes Have Hit a Physical Wall
Scatter plot showing increasing main camera sensor sizes for Apple and Samsung smartphones from 2016 to 2025.
The biggest sensor in a modern iPhone measures roughly 9.8×7.3mm, compared to a full-frame sensor's 24×36mm — over 12 times larger. That gap matters because a larger sensor collects far more light, directly translating to better dynamic range and low-light performance. But phone sensors can't keep growing indefinitely: a bigger sensor demands a bigger, thicker lens assembly, which conflicts with the industry's push for slim, pocketable designs.
Manufacturers have pushed sensor sizes up steadily since 2016, but that trend line is flattening as camera bumps already dominate phone designs. Smaller pixels also introduce their own problems — they're far more sensitive to microscopic optical misalignment, since a 1-micrometer error matters much more on a ~1μm pixel than on a ~4μm pixel found in dedicated cameras.
Computational Photography Papers Over the Cracks
This is where computational photography stepped in — using algorithms, multiple exposures, and machine learning to synthesize image quality the hardware alone can't deliver. Techniques like HDR merging, night-mode multi-frame stacking, and simulated bokeh let phones fake qualities — like shallow depth of field — that would otherwise require a large sensor and wide aperture lens. As one MIT Media Lab paper puts it, computational photography "combines plentiful computing, digital sensors, modern optics, actuators, probes and smart lights to escape the limitations of traditional film cameras"
.
But this software layer has its own ceiling. Photography critic and analyst writing for PetaPixel frames it through "photographic bandwidth" — the information-theoretic limit on how much optical data a given lens and sensor can absorb under specific lighting conditions. No amount of processing can invent detail that was never captured; a tiny lens still lets in a fraction of the light a dedicated camera does, especially in dim environments
.
AI Photography: Genuine Leap or Diminishing Returns?
AI-driven imaging has pushed further than classical computational photography by learning what scenes "should" look like rather than just merging frames. Some observers argue this is making sensor size "increasingly irrelevant," as AI models fill in detail, suppress noise, and enhance dynamic range beyond what raw hardware captures. Google's Pixel line and Samsung's Galaxy AI features exemplify this — leaning on neural processing to compensate for smaller optics. Yet critics counter that this substitutes real information with plausible-looking inference. Overprocessed skies, waxy skin tones, and AI-hallucinated moon detail are frequent complaints, and depth-of-field simulation still can't fully replicate the natural optical falloff of a large sensor and wide aperture. AI photography extends what's possible, but it's a workaround for physical limits — not a replacement for them.
Physical Lens Limitations Nobody's Solved Yet
Beyond sensor size, smartphone lenses face constraints computation can't touch. Thin phone bodies prevent lenses with broadly variable focal lengths, which is why phone makers stack two, three, or four fixed-focal-length cameras (main, ultra-wide, telephoto) instead of one true zoom lens. Small, fixed apertures also mean diffraction blur becomes a real concern — light bending slightly as it passes through a tiny opening, softening fine detail in ways no algorithm can fully undo.
Periscope-style folded telephoto lenses, like those popularized by OnePlus's Hasselblad-branded modules, were one workaround — bending the light path 90 degrees inside the phone body to fit a longer focal length without adding thickness. But even these have hit diminishing returns, since the sensor behind them is still tiny compared to what a longer true zoom lens would pair with on a dedicated camera.
What Comes Next for Mobile Imaging
The next wave of improvement looks less about single dramatic breakthroughs and more about combining specialized hardware with smarter software: variable aperture mechanisms (already used by Samsung and Huawei), multi-lens fusion that blends data from several sensors in real time, and dedicated AI silicon — like Google's Tensor chips — built specifically to accelerate on-device image processing without cloud latency. Expect incremental sensor gains (larger 1-inch-type sensors trickling down from flagship to mid-range) paired with increasingly aggressive AI reconstruction, rather than any single sensor size doubling the game overnight.
The honest takeaway for photography enthusiasts: smartphone cameras aren't stagnating, but they're shifting from a hardware race to a software and AI race, because the laws of physics — not engineering effort — are what's really capping how good a lens the size of a coin can ever be.