Apple is reportedly investigating several semiconductor startups for potential acquisition as it seeks to bolster its artificial intelligence capabilities. This strategic move aims to lessen Apple's dependence on external hardware, particularly Nvidia's powerful GPUs, for its most demanding AI workloads.
Reducing Reliance on External AI Hardware
Currently, Apple utilizes its own data centers for some AI processing, but for heavy-duty tasks, it relies significantly on Nvidia GPUs hosted on Google Cloud. This external dependency creates performance limitations and a desire for greater control over its AI infrastructure.
The company is said to be facing challenges with its current in-house AI server infrastructure, which uses M2 Ultra chips. While powerful, these chips are not specialized enough to handle the most demanding AI tasks as effectively as dedicated AI accelerators.
Strategic Acquisitions and Internal Development
According to reports, Apple's exploration includes semiconductor startups specializing in AI chip design. This is not Apple's first foray into acquiring AI talent; earlier this year, it reportedly acquired Israeli AI startup Q.ai for nearly $2 billion and is also in discussions with PrismML.
Beyond acquisitions, Apple has internal projects underway. It is rumored to be developing its own AI server chip, codenamed "Baltra," though its anticipated launch this year has been delayed. Furthermore, Bloomberg reports that Apple is building an AI server chip based on its M5 Ultra, with an upcoming M7 Ultra chip expected to offer significant AI performance gains, potentially supporting up to 1.5TB of memory. However, a server variant of the M7 Ultra is not expected until 2029.
Focus on On-Device AI and Model Compression
A key area of interest for Apple is AI startups specializing in model compression. This technology is crucial for enabling advanced AI models to run more efficiently directly on devices like the iPhone, enhancing privacy and responsiveness. For instance, Gemini-powered Siri is reported to handle demanding AI requests using Nvidia GPUs hosted on Google Cloud, highlighting the need for more efficient on-device processing.
By acquiring companies with expertise in AI chip design and model optimization, Apple aims to integrate more powerful and efficient AI capabilities directly into its ecosystem, from its data centers to the hands of its users.