Qualcomm has struck a partnership with Multiverse Computing, a startup specializing in quantum-inspired algorithms, to develop methods that reduce the size of artificial intelligence models while keeping their accuracy intact. The collaboration targets a persistent problem in AI deployment: large models that demand too much power and memory for edge devices like smartphones, cars, and industrial sensors.
Why model size matters
Today's most capable AI models, such as large language models and computer vision systems, often require cloud servers to run. That creates latency, raises costs, and burns through energy. Qualcomm, which makes chips for mobile and automotive devices, wants to bring those capabilities directly onto hardware that has limited battery and compute resources. The company says shrinking models without degrading performance is a key step toward that goal.
What the partnership brings
Multiverse Computing, based in Spain, uses techniques derived from quantum physics to optimize complex problems. Under the deal, the two companies will develop algorithms that compress AI models—reducing their memory footprint and computational load—while maintaining the same level of accuracy. The work could lower the cost of running AI inference on Qualcomm's Snapdragon and other platforms.
Energy and cost implications
Smaller models mean less data movement and fewer calculations per query, which directly cuts energy consumption. For companies deploying AI at scale, that translates into lower electricity bills and a smaller carbon footprint. For consumers, it could mean faster, more private AI features on phones and laptops because processing stays on the device rather than being sent to the cloud.
The companies have not disclosed a timeline for when the first results of the partnership will appear in commercial products. Qualcomm typically integrates new AI techniques into its chip reference designs before they reach device makers. The next major milestone will likely be a demonstration at a chip industry conference or a technical paper detailing the compression methods.
