Apple is pivoting its hardware strategy to acquire AI chip designers to reduce its reliance on Nvidia. Driven by the fact that its internal “Baltra” server chips were unable to handle large-scale AI models like Google’s Gemini, the company is moving toward a strategy to secure high-performance silicon for its data centers.
For years, Apple favored medium-sized mergers and acquisitions, of around a few hundred million dollars, while being cautious about large-scale operations. Several signs indicate a shift in this approach. The AI arms race involves compute requirements. If the company cannot handle AI tasks in its own data centers, it relies on Nvidia hardware hosted on Google Cloud.
The Baltra Bottleneck and the Nvidia Dependency
Apple’s internal server AI chip, developed in-house under the code name “Baltra,” was initially scheduled for release this year. According to sources cited by The Information, its schedule has been delayed. Up until now, Apple’s chip design has focused on low-power semiconductors for devices like the iPhone, iPad, or Mac, rather than high-performance server chips.
When Apple engineers attempted to run the Gemini model—which powers the new version of Siri—on Apple’s own server infrastructure, the chips, designed for Mac-like workloads, proved incapable of processing large-scale AI models. To keep Siri functional, Apple chose to entrust part of these calculations to Nvidia GPUs hosted on Google Cloud.
This creates a dependency. Relying on Nvidia means accepting the supply chain. For a company that relies on its own data centers for a portion of its AI tasks, using hardware hosted on Google Cloud illustrates the limits of its current internal server chips.
Breaking the “Net Cash Neutral” Seal
The shift is also financial. CFO Kevan Parekh recently indicated that Apple no longer intended to stick to its “net cash neutral” policy, which consisted of maintaining a balance between cash and debt. The group did not detail the reasons for this change, but several observers see it as a maneuver to give itself more financial flexibility for large-scale acquisitions.
In January, Apple acquired Q.ai for about $2 billion, an Israeli artificial intelligence company specializing in speech interpretation from facial micromovements. It is its second-largest acquisition, behind the purchase of Beats Electronics for $3 billion in 2014. When a company that usually favors medium-sized mergers and acquisitions drops billions on an AI firm, it marks a shift.
- Historical Approach: Medium-sized mergers and acquisitions.
- New Mandate: Potential for large-scale acquisitions.
- The Catalyst: The internal server chips were insufficient to run AI models alone.
The Ternus-Srouji Power Shift
A leadership reshuffle slated for September could redraw the group’s merger and acquisition strategy. John Ternus, current head of hardware, is set to succeed Tim Cook as CEO, while Johny Srouji, who oversaw semiconductor development, should oversee all of Apple’s hardware engineering, beyond just chips. This change fuels expectations of more offensive investments and acquisitions by Apple in the fields of artificial intelligence and semiconductors.
This consolidation of power suggests that the “silicon-first” mentality is now the philosophy across hardware. Apple is essentially trying to build on the foundation of its 2008 acquisition of PA Semi for $278 million, which laid the groundwork for the development of its own Apple Silicon chips.
The PA Semi deal provided the blueprint for Apple Silicon. Now, Apple is looking toward potential acquisitions in the domains of artificial intelligence and semiconductors.
The Broader Chip War
Apple’s struggle highlights the difficulty of processing AI at scale. While Apple’s chip design has focused on low-power semiconductors for devices like the iPhone, iPad, or Mac, the company is now looking at acquiring AI chip manufacturers to reduce its dependence on Nvidia for the most demanding AI calculations.
By pursuing acquisitions of chip manufacturers, Apple is attempting to address the limits of its current infrastructure. If they can acquire the necessary technology, they can address the needs of their AI models.
This move also puts Apple in a position where it is looking to expand its capabilities in the fields of artificial intelligence and semiconductors.
The 30-Second Verdict: Apple’s internal server chip (Baltra) was insufficient to run models like Gemini alone. By signaling an end to its “net cash neutral” policy and elevating Johny Srouji, the company is moving toward a strategy of potential large-scale acquisitions to address its needs in AI and semiconductors.
The company is looking at acquisitions as a serious option to reduce its dependence on Nvidia for the most demanding AI calculations.
