For the past three years, the artificial intelligence industry has been engaged in what may be the largest infrastructure arms race in technology history. Microsoft has announced plans to spend more than $80 billion in fiscal year 2025 on AI-enabled data centers. Meta expects to spend between $60 billion and $65 billion in capital expenditures, primarily to support AI infrastructure. Alphabet plans to invest approximately $75 billion in AI and cloud infrastructure. Amazon’s projected capital expenditure for 2025 is expected to exceed $100 billion, with the majority directed toward AWS and generative AI initiatives.
Collectively, Big Tech is deploying well over $300 billion annually to build the computational backbone of artificial intelligence.
The prevailing assumption behind these investments is simple: the companies with the largest infrastructure footprints and the most advanced models will dominate the AI era.
But history suggests something different.
Technology Revolutions Are Rarely Won by the Builders Alone
The internet did not create the most value for companies manufacturing networking equipment. Smartphones did not create the most value for semiconductor suppliers. Cloud computing did not make server manufacturers the most valuable companies in the world.
Instead, value consistently migrated toward companies that controlled:
- Distribution
- User relationships
- Ecosystems
- Interfaces
- Developer platforms
Google became synonymous with search because it owned the consumer gateway to information. Microsoft built one of the world’s largest businesses by controlling the operating system layer of personal computing. Apple created one of the most valuable companies in history by owning the smartphone experience, despite outsourcing the manufacturing of many of its components.
Technology history repeatedly demonstrates that owning the interface often matters more than owning the underlying infrastructure.
Artificial intelligence may be following the same pattern.
The Great Commoditization of Large Language Models
Just two years ago, frontier AI models exhibited significant capability gaps. Today, those gaps are rapidly narrowing.
Modern large language models—including GPT, Claude, Gemini, and other frontier systems—can all:
- Generate long-form content
- Write and debug code
- Analyze documents
- Perform reasoning tasks
- Conduct research
- Generate images and multimedia assets
- Support enterprise workflows
Independent benchmark studies continue to show performance improvements across all major providers, but differences increasingly exist at the margins rather than at an order-of-magnitude level.
From the perspective of the average user, artificial intelligence is becoming increasingly interchangeable.
Most people do not choose an AI assistant because it scored marginally higher on a benchmark test. They choose based on:
- Accessibility
- Cost
- Convenience
- Integration into their workflow
- Trust and reliability
As models become more capable, intelligence itself risks becoming commoditized.
And when technology becomes abundant, value typically shifts elsewhere.
Distribution Is Becoming the Ultimate Moat
Artificial intelligence may ultimately become a distribution game.
According to industry estimates, there are now more than 2.3 billion active Apple devices worldwide. The iPhone alone has surpassed 1 billion active users, creating one of the largest and most engaged technology ecosystems ever assembled.
This distribution advantage matters enormously.
Every day, billions of interactions occur across:
- iPhones
- Macs
- iPads
- Apple Watches
- AirPods
- Vision Pro devices
Increasingly, these interactions are becoming AI-assisted.
The critical insight is that users are beginning to care less about which model powers the experience and more about how seamlessly intelligence integrates into their lives.
When someone summarizes an email, rewrites a message, generates notes, or asks an assistant a question, they rarely think about which data center processed the request.
They simply expect the experience to work.
This is where platform ownership becomes extraordinarily powerful.
Apple May Have Chosen Optionality Over Ownership
Unlike its peers, Apple has largely avoided participating in the expensive race to build frontier-scale AI infrastructure.
Instead, Apple appears to be pursuing a different strategy: rent intelligence while owning the customer experience.
Reports indicate that Apple has entered significant agreements with external model providers, including Google and OpenAI, to integrate advanced AI capabilities into its ecosystem.
From a capital allocation perspective, this strategy is remarkably efficient.
Rather than deploying tens of billions of dollars annually into infrastructure, Apple retains flexibility.
If one model improves, it can adopt it.
If another provider becomes superior, it can switch.
If economics change, it can rebalance.
And if its own models eventually become competitive, it can integrate them natively.
This approach creates something that is increasingly rare in technology: strategic optionality.
The Future of AI Is Ambient and Invisible
The next phase of artificial intelligence will not involve users opening dedicated AI applications every time they need assistance.
AI is becoming ambient.
It will increasingly exist everywhere:
Inside messaging applications.
Inside productivity suites.
Inside search experiences.
Inside operating systems.
Inside cameras and photos.
Inside voice interfaces.
Inside enterprise workflows.
Over time, intelligence itself becomes invisible.
The experience becomes the product.
This shift fundamentally changes where value accrues.
Owning the model may remain important.
But owning the moments where humans interact with intelligence may become even more valuable.
Why the Last Mile Matters More Than Ever
The “last mile” refers to the final layer between technology and the end user.
Historically, this layer has generated extraordinary economic value.
In e-commerce, platforms own the last mile.
In smartphones, operating systems own the last mile.
In cloud computing, developer platforms own the last mile.
In artificial intelligence, the last mile is likely to include:
- Devices
- Operating systems
- Productivity applications
- Communication channels
- Enterprise workflows
- Personal context and memory
Apple already controls many of these layers.
This does not guarantee that Apple wins the AI race.
No company has ever won a technology transition by default.
However, it does suggest that the market may be underestimating a critical strategic reality:
The future of artificial intelligence may not belong exclusively to the companies spending the most money building intelligence.
It may belong to the companies that control where intelligence is experienced.
The New AI Value Equation
The AI value chain increasingly looks like this:
Infrastructure → Models → Platforms → Experiences → Users
History consistently shows that as technologies mature, infrastructure becomes cheaper, capabilities become more standardized, and experiences become the primary source of differentiation.
This does not diminish the importance of foundation models.
Without frontier models, there is no intelligence to distribute.
But if intelligence becomes abundant—and history suggests that it eventually will—the scarce asset becomes something else entirely.
Distribution.
Trust.
Context.
Ecosystems.
The last mile.
The biggest winners of the AI era may not necessarily be the companies with the most GPUs or the largest data centers.
They may be the companies that own the interfaces through which billions of people experience intelligence every day.
And by that measure, Apple may not be behind in artificial intelligence at all.
It may simply be playing a different game.
