Key Highlights
- Discussions underway between Apple and Caltech-backed PrismML regarding on-device AI compression capabilities
- The startup compressed Alibaba’s Qwen 3.6 model with 27 billion parameters from 54 GB to less than 4 GB
- PrismML’s approach keeps all 27 billion parameters simultaneously active on iPhone 17 Pro hardware
- Current enhanced Siri functionality depends on cloud-based Google Gemini models powered by Nvidia infrastructure
- Khosla Ventures led PrismML’s $16.25 million seed funding round this year
Apple (AAPL) stock climbed 0.44% on Wednesday following reports that the iPhone maker has entered discussions with PrismML, an artificial intelligence startup that originated from the California Institute of Technology.
According to The Information’s initial report, the negotiations focus on PrismML’s breakthrough capability to operate a 27-billion-parameter artificial intelligence model entirely on an iPhone 17 Pro device — eliminating the need for cloud connectivity.
PrismML successfully reduced Alibaba’s open-source Qwen 3.6 model size from approximately 54 gigabytes to under 4 gigabytes. This represents a compression rate exceeding 90%.
The distinction from conventional compression techniques lies in maintained performance. While typical compressed models lose accuracy when downsized for device operation, PrismML maintains it hasn’t sacrificed quality.
The company accomplishes this through ultra-compact 1-bit and ternary weight structures, delivering memory requirement reductions up to 14x while executing up to 8x faster than conventional models.
Traditional on-device AI implementations utilize sparse architecture, activating only portions of parameters simultaneously. PrismML’s technology maintains all 27 billion parameters in simultaneous operation.
This enables the device-based model to execute sophisticated operations including reasoning, autonomous agent functions, and software development tasks, per the startup’s claims.
The model’s open-source release is scheduled for next Tuesday.
Addressing Apple’s Cloud Infrastructure Challenge
Currently, Apple’s cutting-edge Siri capabilities — unveiled at June’s WWDC — depend on Google’s Gemini models operating on Nvidia processors within Google Cloud infrastructure.
Apple has deployed one on-device model featuring 20 billion parameters, though only 1 to 4 billion remain active simultaneously within its sparse configuration.
PrismML’s full-parameter methodology represents a significant advancement. This would enable Apple to execute more demanding AI operations completely on-device, reducing cloud infrastructure expenses while maintaining user data locally.
Background on PrismML
PrismML was established by Babak Hassibi, an electrical engineering professor at Caltech. The underlying mathematical research originated from university-based development.
Caltech maintains patent ownership for the foundational technology while granting exclusive licensing rights to PrismML.
The company secured $16.25 million in seed funding this year, with Khosla Ventures leading the round — the identical investment firm that provided OpenAI’s initial venture capital.
Neither Apple nor PrismML has issued public statements regarding the potential acquisition discussions.
As competitors including Microsoft, Amazon, and Meta invest hundreds of billions constructing data center capacity, Apple has maintained its commitment to device-based AI processing.
Acquiring PrismML would represent Apple’s most ambitious advancement of this strategic direction.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
