A $30 billion Broadcom partnership and a $2 billion AI startup acquisition signal Apple's scramble to close a massive gap in machine learning performance.
Apple’s in-house silicon has earned a reputation for punching above its weight. But when it comes to the raw computational muscle needed for advanced AI workloads, the company’s chips are getting thoroughly outclassed, and Apple knows it.
The M2 Ultra, which Apple has deployed for some server-side AI processing, delivers roughly 31.6 TOPS (tera operations per second) in machine learning performance. For context, Nvidia’s RTX 4090 hits 1,321 TOPS. That’s not a gap. That’s a canyon with a river at the bottom.
Broadcom deal and the $30 billion bet on custom silicon #
Apple has responded with its wallet. The company announced a multiyear partnership with Broadcom worth over $30 billion, focused on custom silicon components and wireless connectivity technologies manufactured in the United States.
The deal, unveiled around July 8, 2026, represents one of Apple’s most ambitious semiconductor commitments to date. It’s designed to bolster domestic chip production at a time when geopolitical tensions continue to make global supply chains look fragile.
The company has also engaged in partnerships with Intel to advance US chip design and manufacturing, along with explorations involving Nvidia and Google for Siri-related workloads.
The Q.ai acquisition and Apple’s AI shopping spree #
Beyond silicon partnerships, Apple has been writing checks for AI talent. In January 2026, the company acquired Q.ai, an Israeli AI startup, for approximately $2 billion. The deal is aimed at enhancing Siri’s capabilities, specifically in non-verbal communication through improved facial expression analysis.
The Q.ai purchase ranks among Apple’s largest acquisitions ever. Apple has completed more than 100 acquisitions to date, with a pronounced recent emphasis on AI and machine learning companies.
Why the M2 Ultra gap matters more than you think #
The performance disparity between Apple’s M2 Ultra and Nvidia’s hardware isn’t just an engineering footnote. It has real implications for Apple’s ability to compete in AI-powered services.
Apple’s 31.6 TOPS versus Nvidia’s 1,321 TOPS means Apple would need roughly 42 M2 Ultra chips to match what a single RTX 4090 can do in machine learning tasks.
What this means for investors #
First, the $30 billion Broadcom deal is enormous even by Apple’s standards. It locks both companies into a long-term collaboration that could reshape how custom AI chips are designed and manufactured in the US.
Second, Apple’s willingness to spend $2 billion on a single AI acquisition suggests the company’s M&A budget for AI is expanding significantly. With more than 100 acquisitions already completed and a clear strategic focus on machine learning, the pipeline of potential targets in the AI startup ecosystem could see valuation inflation as Apple competes with Google, Microsoft, and others for the same talent and technology.
Third, the exploration of partnerships with Nvidia and Google for Siri workloads raises an interesting question about Apple’s competitive posture. Historically, Apple has been allergic to depending on rivals for core product capabilities. If it’s now willing to lean on Nvidia’s compute and Google’s AI infrastructure, that signals either pragmatism or desperation, and the distinction matters for how you value the stock.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our