In a recent LinkedIn post, Linas Beliūnas discusses the often-overlooked role Apple is playing in the artificial intelligence infrastructure landscape, challenging the narrative that the tech giant is lagging behind in AI development.
Beliūnas highlights a perceived irony: while Apple has been criticized for a lack of prominent AI initiatives compared to competitors like OpenAI, Google, and Microsoft, it is emerging as a significant player in AI infrastructure.
“Tech loves irony: Apple, the company everyone says is ‘behind in AI,’ is quietly becoming one of the most important AI infrastructure players on the planet 😳”
Challenging the AI Narrative
The prevailing story, as Beliūnas outlines, was that Apple lacked frontier models, wasn’t participating in the AGI race, and didn’t showcase flashy AI demos or build massive GPU clusters. This perception led many to view Apple as irrelevant in the AI space.
However, Beliūnas points to a growing trend among developers who are repurposing Mac Minis into AI clusters, capable of running workloads previously requiring dedicated data centers. This development, he suggests, is not driven by hype but by a focus on systems, efficiency, and thoughtful design.
Apple’s Strategic Focus on Local Intelligence
According to Linas Beliūnas, Apple’s strategy has always been centered on local intelligence rather than cloud-based AI. This approach is underpinned by several key hardware and software optimizations:
- Unified memory architecture
- Exceptional memory bandwidth per dollar
- Ubiquitous Neural Engines
- Integrated hardware and OS design
Beliūnas argues that this design philosophy has inadvertently opened up a new reality for AI deployment, where inference is primarily memory-bound, not FLOP-bound, and where the practical use case often involves a batch size of one. He emphasizes that most AI workloads do not necessitate hyperscale computing resources.
“While everyone chased brute force, Apple made efficient compute quietly excellent.”
The Unfolding AI Reality
The implications of Apple’s approach are becoming increasingly apparent, Beliūnas notes:
- Mac Minis are increasingly being used as alternatives to GPUs for inference tasks.
- Large AI models can now run locally, offline, and privately.
- Power consumption for these local setups is significantly lower, measured in hundreds of watts.
- Costs are dramatically decreasing, making advanced AI capabilities accessible to individual developers and smaller operations.
Beliūnas references the observation that even prominent figures like Andrej Karpathy have noted the suitability of Mac Minis for these tasks, particularly when considering memory bandwidth costs over raw FLOPS. He adds:
“The closest we have today to Jarvis from Iron Man 🤖”
The ‘Home AI Brain’ and Classic Apple Strategy
The emergence of always-on personal AI agents is further transforming Mac Minis into what Beliūnas terms “home AI brains.” These machines offer stable, silent operation and provide users with ownership of their intelligence, rather than renting it from the cloud.
This development, Beliūnas concludes, is characteristic of Apple’s strategic approach: ignoring prevailing narratives and excelling within specific constraints. He posits that Apple was not late to the AI game but was pursuing a fundamentally different path all along.
“The company accused of doing nothing in AI is quietly becoming one of the most practical AI infrastructure players in the world.”
Ultimately, Beliūnas’s analysis suggests that Apple’s focus on efficient, integrated systems has positioned it as a key, albeit understated, enabler of practical, accessible AI.
📝 About This Content
This article is based on insights shared by Linas Beliūnas on LinkedIn.
📅 Originally posted on January 25, 2026 | View original post on LinkedIn →