RAM Prices Could Drop as Google's AI Technology Panics Hoarders

Over the past year, the world has faced a memory crisis. Memory prices, particularly for DRAM types such as DDR4 and DDR5, have surged rapidly due to supply shortages, as production is prioritised to meet the needs of artificial intelligence (AI) development.

However, there is a glimmer of hope from Google amid this memory crisis. The technology giant is reportedly developing AI technology that is more RAM-efficient. This breakthrough is causing panic among RAM hoarders, which could reduce the shortage and lead to a drop in memory prices.

TurboQuant was developed by the company’s research division, Google Research, with a primary focus on memory efficiency during the inference process—that is, when the AI model is run rather than trained.

This technology targets one of the main bottlenecks in modern AI systems: the limitation of ‘working memory’, particularly in the component known as the KV cache, which is temporary memory used by the model to process and remember data context.

TurboQuant’s operation relies on a technique called vector quantisation, a method of simplifying the representation of numerical data in vector form to make it more compact without losing important information.

With this approach, data that previously required large space can be compressed significantly while still maintaining the AI model’s accuracy.

Technically, TurboQuant relies on two main methods: PolarQuant and Quantisation-aware Joint Learning (QJL).

PolarQuant functions by changing the way data is represented to make it more efficient when stored in memory, without sacrificing the quality of computation results.

Meanwhile, QJL trains the AI model to be ‘aware’ that the processed data will be compressed, allowing the model to adapt and still produce accurate outputs even when working with compacted data.

With the combination of these two techniques, researchers claim that TurboQuant can save up to six times more memory usage compared to conventional methods.

This means AI models can ‘remember’ more information in much smaller space, while also reducing performance barriers due to memory limitations.

It is this capability that positions TurboQuant as a potential solution to the current RAM crisis.

In recent times, memory prices, especially DDR5, have risen sharply due to high demand from the AI industry.