In a groundbreaking leap that could upend the way we think about memory performance, a research team at Fudan University has developed the world’s fastest non-volatile flash memory — and it’s not just fast. It’s lightning. Dubbed PoX, this next-gen chip clocks in at 400 picoseconds per bit write, or 25 billion writes per second, leaving today’s memory tech in the dust.
That’s not a typo. It’s nearly 10,000× faster than current flash memory and even enters speed territory previously reserved for volatile types like SRAM and DRAM, which, unlike PoX, lose data without power.
🚀 Smashing the Bottleneck Holding AI Back
Modern AI hardware is a data-hungry beast. It’s not arithmetic that eats up time and energy — it’s memory. Shuffling terabytes of parameters every second is the real bottleneck in inference and training. Flash memory was never fast enough to keep up… until now.
Led by Prof. Zhou Peng, the Fudan team rewrote the rules of flash memory by integrating two-dimensional Dirac graphene in place of traditional silicon, achieving ballistic charge transport and bypassing the limitations of classical physics.
By optimizing the so-called Gaussian length of the channel, they pulled off a feat called two-dimensional super-injection — a near-unlimited charge surge into the storage layer. In simpler terms? Memory that writes faster than your brain can blink — and retains it without any power.
“This pushes non-volatile memory to its theoretical limit,” Zhou said. “We believe it could be the key to unlocking high-speed AI hardware.”
🔋 Power-Efficient and Persistent — The Holy Grail
Co-author Liu Chunsen described the jump in relatable terms: going from a USB stick that writes 1,000 times per second to a chip that does one billion in a blink. And because PoX is non-volatile, it can store data indefinitely with no standby power, making it a dream for edge AI and low-power devices.
The implications? No more bloated SRAM caches, instant-on laptops and phones, entire databases running from persistent RAM, and a massive slash in energy consumption — all while maintaining brutal speed.
🌍 Strategic & Industrial Earthquake
Flash memory is a cornerstone of global semiconductor dominance, and China’s PoX breakthrough adds an original twist that might tip the scales. If mass-produced, PoX could redefine AI chip design, eliminate layers of memory hierarchy, and massively reduce costs and power use.
It also marks another strategic win for China’s semiconductor ambitions, strengthening domestic control over critical tech as geopolitical tensions reshape supply chains.
No commercial partners have been announced yet, but with Chinese fabs already racing to integrate 2D materials into CMOS, it’s only a matter of time before PoX hits the foundry floor.
Bottom Line
This isn’t just a research win. PoX could change everything — from how AI hardware is built, to how we store and move data, to how nations position themselves in the semiconductor race.
And if it delivers at scale?
Say goodbye to memory bottlenecks. And hello to a whole new class of AI-native, ultra-green computing.
