Apple has been talking to bankers and semiconductor startups about potential acquisitions as it scrambles to build AI server chips powerful enough to handle next-generation workloads, including a revamped Siri. The effort, revealed by The Information, comes as the company’s internal M2 Ultra-based servers struggle with large AI models and its next-in-line chip, codenamed Baltra, has slipped past its planned 2026 debut. For a company that historically avoids splashy deals, the shift signals just how critical the AI infrastructure race has become.

The M2 Ultra Was Never Meant for This

Apple designed the M2 Ultra to power high-end Macs, not the unrelenting demands of a data center AI cluster. The chip’s unified memory architecture—impressive for a workstation—hits a wall when running models like Google’s Gemini, which Apple leaned on for an overhauled Siri. According to The Information’s sources, Apple tried to get Gemini running on its own M2 Ultra servers but failed. The hardware simply wasn’t up to the task. As a result, Apple turned to Nvidia GPUs inside Google Cloud, a workaround that carries its own strategic sting given the historically frosty relationship between Apple and Nvidia.

That dependency is exactly what Apple wants to break. The Baltra chip, originally expected in 2026, was supposed to be the answer. Its delay—first reported by The Information—has forced Apple’s hand. The company is now signaling it may shop for outside help rather than build everything from scratch. The playbook isn’t entirely new: Apple’s 2008 purchase of PA Semi for $278 million ultimately gave it the powerhouse silicon team behind the A-series and M-series. This time, the stakes and the price tags could be much higher.

A Pivot to M&A: From Small Tuck-Ins to Bigger Plays

Apple’s M&A doctrine has long been “small and surgical.” The $3 billion Beats deal in 2014 was an outlier—until January 2026, when Apple dropped nearly $2 billion on Q.ai, an Israeli startup that interprets speech from facial micromovements. That purchase, Apple’s second-largest ever, was aimed at AI capabilities, not chip design. But it cracked open the door to larger deals.

Now, according to The Information, Apple’s hunt specifically targets AI chip companies. The company has gone as far as discussing potential acquisitions with bankers and approaching semiconductor startups to gauge their willingness to sell. CFO Kevan Parekh further signalled a strategic shift in Apple’s Q2 2026 earnings call by announcing the company would move away from its long-held “net cash neutral” policy. That policy prioritized returning cash to shareholders; scrapping it frees up tens of billions for big-ticket M&A.

On a parallel track, Apple deepened its chip-making ties with Broadcom. A July 6 regulatory filing revealed the two companies expanded their technology collaboration through 2031, covering multiple generations of custom ASICs. Broadcom didn’t specify the chips, but the deal underscores Apple’s willingness to go beyond its Cupertino walls. Apple’s own newsroom announcement framed the Broadcom expansion as a U.S. manufacturing win, committing to billions more in domestic chip production.

What Apple’s Moves Mean for Windows Users and IT Pros

There is no Windows patch, compatibility update, or enterprise action tied to this news. The impact is indirect but worth parsing for two audiences.

For home users and Windows enthusiasts: Apple’s server headaches don’t affect your PC. Windows 11, Copilot+ PCs, and the upcoming AI features in Microsoft’s pipeline all rely on a combination of local NPUs and cloud services backed by Azure’s massive Nvidia GPU clusters. That ecosystem isn’t flipping because Apple can’t get its own server hardware to perform. What you might see, however, is an even faster cadence of AI innovation as every major player throws money at custom silicon. More competition could eventually mean better AI features on all platforms.

For IT administrators and infrastructure teams: The immediate takeaway is supply chain vigilance. If Apple starts buying mid-sized chip design firms, the pool of third-party AI accelerator startups shrinks. That could mean fewer alternatives to Nvidia’s dominant CUDA ecosystem and potentially higher licensing costs if you’re evaluating non-GPU AI hardware. Talent competition will intensify too—chip design engineers are scarce, and a newly acquisitive Apple (on top of Microsoft, Google, and Amazon) will drive up salaries and attrition. If you run on-prem AI workloads, keep an eye on component lead times and second-source strategies. The M2 Ultra’s struggles are also a reminder that even custom silicon giants can misread data center workloads; don’t assume every in-house chip will be a panacea.

The Backstory: How Apple Got Here

Apple’s chip ambitions were born from a 2008 tuck-in acquisition. PA Semi gave the company a team that would later deliver the A4, then the A-series dynasty, and eventually the M1 that stunned the PC industry. But those processors were always tuned for power efficiency in battery-constrained devices. Data center AI is a different beast: it demands massive memory bandwidth, fast interconnects, and cluster-level software ecosystems—all areas where Nvidia’s CUDA and the hyperscalers’ custom ASICs currently reign.

When Apple launched Apple Intelligence at WWDC 2024, it leaned heavily on on-device processing with its Neural Engine. The cloud side, intended to power more complex requests, ran on M2 Ultra servers in Apple’s data centers. That worked for smaller models. But as large language models grew, the M2 Ultra fell behind. Bloomberg has reported that Apple is working on an M5 Ultra upgrade and a far more capable M7 Ultra, the latter with up to 1.5 TB of memory, targeting 2029. But even that timeline suggests a multi-year gap where Apple may not have a homegrown server chip that can match Nvidia’s latest.

Microsoft, by contrast, has invested heavily in Azure’s Nvidia GPU fleet while also developing its own Athena AI chip. Google’s TPUs power Gemini. Amazon has Trainium. Apple is playing catch-up in a space it hadn’t prioritized—until now.

What You Should Do Now

For the average Windows user: Nothing. But if you care about the broader tech landscape, Apple’s struggles are a signal that the AI hardware race is far from settled. The winner eventually determines which AI services get the fastest, cheapest infrastructure.

For IT decision-makers: Treat Apple’s moves as a leading indicator. If your organization is planning an on-premises AI deployment in the next 12–24 months, lock in supply agreements for Nvidia or AMD GPUs sooner rather than later. Monitor chip M&A news; a big Apple acquisition could tighten the talent market and disrupt startup roadmaps you might be evaluating. Also, watch how Apple’s Broadcom partnership evolves—it could produce an AI ASIC that competes with Nvidia’s H100 series down the road, introducing a new option for cloud providers and, eventually, enterprise gear.

The Outlook: More Deals, More Competition

The net cash neutral policy change was the real tell. Apple sitting on over $150 billion in cash and marketable securities is now explicitly open to spending it on growth, not just buybacks. That could mean a string of chip-related acquisitions in the next 18 months, from companies specializing in high-bandwidth memory interconnects to RISC-V AI accelerators. The Baltra delay isn’t just a footnote; it’s the catalyst that may force Apple to buy the expertise it needs. For Windows shops, the story is one more reason to stay nimble in an AI infrastructure market that grows more crowded—and more competitive—by the quarter.