Apple Avoiding the $650 Billion AI Server Trap Could Be a 'Brilliant' Move [Analyst]

Tech giants are currently pouring hundreds of billions of dollars into artificial intelligence data centers, but Apple has chosen a very different path that industry analyst Horace Dediu suggests could prove to be either the most brilliant or the most misguided corporate moves in history.



Amazon, Google, Microsoft, and Meta are projected to spend a combined $650 billion this year alone on AI data center infrastructure. Hyperscalers are now directing roughly 94 percent of their operating cash flows toward these projects. To sustain the expansion, the five largest tech firms raised $121 billion in bonds in 2025 alone, and Morgan Stanley projects they could accumulate as much as $1.5 trillion in tech debt over the coming years. Meanwhile, Apple has kept its capital budget at a comparatively modest $14 billion, fluctuating primarily with standard hardware tooling cycles rather than large-scale server farm construction.


Dediu believes the industry may be overinvesting in AI infrastructure just as the models themselves are becoming easier to replicate. Instead of building massive new data centers, Apple chose to partner with Google and license access to Gemini for an estimated $1 billion per year. The arrangement also gives Apple room to change course if a stronger or cheaper model appears later. This hybrid strategy could let the company benefit from rapid advances in AI without committing tens of billions to building its own infrastructure.

Rather than relying entirely on cloud processing, Apple is also leaning heavily on on-device computing across its hardware ecosystem. The company now has an installed base of more than 2 billion active devices. The recent launch of the M5 processor family underscores this hardware-first approach. Equipped with a 16-core Neural Engine and Neural Accelerators integrated directly into every GPU core, the new silicon delivers up to four times the AI performance of the previous generation. According to Dediu, this allows Macs and iPads to run models with tens of billions of parameters locally, reducing the need for large-scale cloud infrastructure for many everyday AI tasks.

The financial contrast between these strategies is significant. While several competitors have reduced stock buybacks to fund their AI expansion, Apple spent $90.7 billion repurchasing shares during its last fiscal year. Dediu notes that the company appears to be betting that the long-term winners of the generative AI transition may not be the companies building the infrastructure, but those that control the hardware platforms and customer relationships where AI services are ultimately delivered.

Please download the iClarified app or follow iClarified on X, Facebook, YouTube, and RSS for more Apple AI updates.