Apple looks to acquire chip companies to reduce dependence on Nvidia

Apple is seeking to acquire chip companies in order to reduce its dependence on Nvidia for the AI workloads it runs internally. That is what sources have told The Information. The focus is on specialised chips for Apple's own AI servers, on which the company runs large language models for services including Siri.

The immediate trigger is concrete: Apple's internal servers, built around the M2 Ultra chip, cannot adequately handle demanding AI models such as Google's Gemini. For those tasks, Apple currently relies on Nvidia GPUs, sourced via Google Cloud. At the same time, development of Apple's own AI server chip, referred to internally as 'Baltra' and originally expected in 2026, has been delayed, making the urgency for external solutions greater.

Alongside potential acquisitions, Apple is already working with chipmaker Broadcom on an AI server chip. That collaboration began in 2024 and now runs through to 2031. In that context, Apple has committed to a deal worth more than 30 billion dollars for the production of over fifteen billion chips manufactured in the United States.

Larger acquisitions than Apple is accustomed to

Apple is known for acquiring relatively small startups, typically for amounts in the hundreds of millions of dollars. That pattern now appears to be shifting. According to sources, the company is open to deals in the billions, a step change in scale that is unusual for Apple.

CFO Kevan Parekh has publicly indicated that Apple is no longer aiming for a so-called 'net cash neutral' position, in which cash reserves and debt are kept in balance. This gives the company financial room to make larger moves. As of 28 March 2026, Apple held 45.57 billion dollars in liquid assets.

A recent acquisition illustrates the new direction. In January 2026, Apple acquired Israeli company Q.ai for nearly two billion dollars. Q.ai specialises in interpreting speech through microscopically small movements in the face. It is the second-largest acquisition in Apple's history, after the purchase of Beats Electronics for three billion dollars in 2014.

Talks with PrismML on model compression

Apple has also held discussions with PrismML, a startup that has developed technology to compress large AI models so that they can run directly on an iPhone. Whether those talks have resulted in a deal is not known. The interest does fit Apple's broader strategy of keeping AI processing on the device itself as much as possible, rather than in the cloud.

Model compression is technically an active area of research. Reducing larger models without significant quality loss requires specialised knowledge and computational techniques. For Apple, which controls its own silicon from iPhone to Mac, bringing the software side of that capability in-house is a logical next step.

History of chip strategy: from PA Semi to proprietary processors

Apple's current position as a designer of its own chips has its roots in the acquisition of PA Semi in 2008, for 278 million dollars. That company provided the engineers and expertise that ultimately led to the A-series processors in the iPhone and later the M-series chips in the Mac. The acquisition is widely regarded in the chip industry as one of the most effective technology acquisitions of the past decade.

With its current search for chip companies, Apple appears to be looking to make a comparable move, but this time specifically targeting AI inference for server hardware. That is a different domain from the mobile chips where Apple is already strong: the requirements around memory bandwidth, energy consumption at data-centre scale, and parallel processing of large models differ substantially from what is needed for a smartphone or laptop.

What this means for the European and Dutch chip market

Apple's search for chip companies focused on AI inference touches a market in which European players are also active. Companies developing specialised chips or compression technology for AI workloads are seeing growing strategic demand from large tech companies. The Netherlands occupies a central position in the global chip production chain through ASML, but smaller Dutch and European chip companies and AI startups also operate in adjacent niches.

For investors and founders in this segment, the pattern Apple is demonstrating is relevant: large tech companies are willing to pay more for specialised expertise that strengthens their infrastructure position. That increases the opportunities for companies with distinctive technology at the intersection of chip design and AI modelling, even if they are relatively small.