'It seemed to defy the laws of physics': The everlasting 'memory crystals' that could slash data centre emissions

Big data is becoming a big problem. In the face of rising emissions from data centres, researchers are turning to novel solutions for storage. Memory crystals and DNA are two frontrunners.

On a visit to Japan in 1999, researcher Peter Kazansky encountered a mysterious physical phenomenon that he now believes holds the answer to the future of data storage.

In Kyoto University's optoelectronics lab, scientists were experimenting with writing on glass using ultrafast, femtosecond lasers which emit a light pulse every quadrillionth of a second.

But they noticed something unusual about how light was travelling through the glass that had been lasered. Rayleigh scattering is a well-established effect which describes how white light bounces off small particles in all directions (explaining, among other things, why the sky appears blue). But in this case, the light wasn't bouncing as expected.

"It was difficult to explain," says Kazansky, a professor in optoelectronics at the University of Southampton in the UK, who was collaborating with the Kyoto University researchers. "We saw light scattering in a way that seemed to defy the laws of physics." 

The puzzling observation eventually sparked "a real eureka moment", says Kazansky. The researchers discovered hidden nanostructures within the silica glass that had been created by "micro-explosions" from the femtosecond lasers.

Imagine holding a thick rock of crystal up to the sun and seeing light ricocheting off it in many different directions. With the lasering technique, the Kyoto researchers had accidentally created tiny holes that had the same property.

Some 1,000 times smaller than the breadth of a human hair, these "whirlpools" of light are so tiny as to be indiscernible to the human eye. But it soon became clear to the scientists that they had a potentially transformative utility. "This was the first proof that we could use light to 'print' complex patterns inside transparent materials at a scale smaller than the wavelength of light," says Kazansky.   

Now, 27 years later, he hopes the discovery can help solve one of the most intractable problems of our information age: mass data storage. 

In the era of the internet, AI, smart homes and surveillance capitalism, there's one thing we just can't stop producing: data. We will collectively generate 394 trillion zettabytes of the stuff every year by 2028, predicts analyst company IDC (a zettabyte is a trillion gigabytes). Every time we do anything on the internet – watch a YouTube video, send an email, ask an AI chatbot a question – strings of data points bleep off into cyberspace. 

And while we might be inclined to think of data as lighter than air, whizzing intangibly through cables under the sea, or fizzing gently in "the Cloud" somewhere above our heads, it does in fact require intensive physical resources – the demand for which is now proving insatiable. 

The conundrum of where to put all this stuff is inspiring some novel solutions, including Kazansky's solution of etching data into glass with lasers. But other options, such as storing data in DNA, are also being pursued by scientists and companies such as Microsoft.

Data is processed and housed in data centres: gigantic alien structures densely stacked with 7ft (2m)-high racks of blinking servers. These humming boxes of hardware and cables are hungry for energy needed to power both their computing and the vast on-site cooling systems required to keep them from burning up. (A data centre is not a pleasant place to work: hot and deafeningly loud, it is best suited to those who can "endure a lot of pain", according to one 2025 New Yorker investigation.)

Globally, data centres now account for around 1.5% of the world's electricity demand, but their energy use is projected to double by 2030, when they are also on track to produce about 2.5 billion tonnes of carbon dioxide-equivalent emissions globally – equal to about 40% of the US's entire annual emissions. The recent generative AI boom has made matters worse, sharply increasing demand for high-performance computing systems which consume gargantuan amounts of power and expel intense clouds of heat.

The majority of energy used by data centres is expended on "hot data": data that needs to be available at our fingertips for quick access and regular updating. Think transferring money from your bank account or online documents you frequently edit. 

But most data in the world is not this kind of data; up to 80% of it is in fact "cold data" – data that no one needs any time soon, and when they do, they're prepared to wait minutes or........

© BBC