Apple’s upcoming iPhone price hike and rising Mac costs may seem like routine adjustments, but they mask a deeper issue: the AI industry is burning through billions in subsidies while failing to deliver measurable value. Ed Zitron, a long-time critic of the AI boom, points out that companies like OpenAI lost $20.9 billion in 2025 despite $13.07 billion in revenue, all while forcing users to pay for services they can’t track or control. You’re not just paying for AI, you’re subsidizing a broken model that’s already costing businesses millions.

This article outlines the risks of the AI bubble burst and what Apple and other companies must do to avoid financial disaster. We’ll show you how to spot the cracks in AI’s economic foundation and what steps you can take to protect your bottom line before it’s too late.

The Hidden Cost of AI: Why the Bubble Might Burst Soon

The AI industry is built on a flawed economic model, and the consequences are already visible. Companies like OpenAI lost $20.9 billion in 2025 while generating only $13.07 billion in revenue, revealing a system that subsidizes unprofitable services. Users pay monthly fees but end up covering costs that are impossible to track or control. The model assumes customers will accept vague rate limits and hidden token usage, but this doesn’t align with how software is traditionally sold. As Ed Zitron notes, the AI boom is not just overhyped, it’s fundamentally broken. Businesses that ignore this risk will face financial strain when the bubble inevitably bursts.

The Broken Economics of AI Subscription Models

Subscription models hide real token costs from users

Monthly fees give users a false sense of control. In reality, companies like OpenAI and Anthropic charge per million tokens, but they hide these costs behind vague rate limits. Users pay a flat fee but end up subsidizing massive, untracked usage. As Ed Zitron puts it, customers are paying for services they can’t measure or manage.

AI companies subsidize usage to keep users engaged

To maintain engagement, AI firms give users far more tokens than their subscription justifies. A $20-a-month plan can burn hundreds of dollars in token costs, yet users rarely see the real price. This model assumes customers won’t notice or care, but it’s a gamble that’s already backfiring as businesses push back against hidden expenses.

Enterprise customers struggle with unpredictable AI costs

Uber’s experience shows the risks. The company spent its entire annual token budget in a single quarter, and its COO admitted it was getting “harder to justify” the cost. With no clear way to tie AI spending to actual outcomes, enterprises face a growing burden of unpredictable, hard-to-justify expenses.

Real-World Examples of AI Cost Overruns

Uber spent its entire annual token budget in a quarter

Uber’s experience with AI illustrates the scale of the problem. The company spent its entire annual token budget in just a quarter, revealing how difficult it is to track and manage AI costs. This kind of overspending isn’t just a one-off, it’s a systemic issue in how AI is billed and consumed.

GitHub Copilot’s shift to token-based billing highlights the problem

GitHub Copilot’s move to token-based billing in June 2026 shows how AI vendors are trying to align costs with usage. But this also exposes the underlying issue: users are paying for something they can’t easily measure or control. The shift doesn’t solve the problem, it just makes it more visible.

OpenAI’s losses reveal the unsustainable nature of current AI economics

OpenAI lost $20.9 billion in 2025 despite $13.07 billion in revenue, proving that the current AI business model is not viable in the long term. These losses aren’t just a result of high costs, they’re a direct outcome of a flawed economic structure that can’t be sustained without massive subsidies.

What AI Tools Actually Deliver vs. What Businesses Expect

AI tools often slow down workflows instead of speeding them up

Many AI tools introduce friction into existing workflows rather than eliminating it. Instead of streamlining tasks, they require users to retrain, reconfigure, and revalidate processes. This is especially true in manufacturing and operations, where manual checks are still needed to verify AI outputs. The result? More time spent managing AI than actually getting work done.

AI-generated code is frequently of poor quality

AI tools like GitHub Copilot have shown that while they can generate code, the output is often sloppy, inefficient, or outright incorrect. Developers end up spending more time debugging AI-generated code than writing it. This isn’t just a minor inconvenience, it’s a direct hit to productivity and quality assurance.

Limited differentiation between AI platforms reduces value

Most AI platforms offer similar core functions: generate, summarize, search. The lack of differentiation means businesses are paying premium prices for tools that don’t deliver unique value. When every AI vendor offers the same basic features, it becomes nearly impossible to justify the cost of adoption.

What Apple and Businesses Should Do Now

Evaluate AI tools based on real value, not just hype

Don’t be seduced by flashy demos or vague promises. Look at how AI tools integrate into existing workflows and whether they reduce manual effort. Tools that add friction or require constant revalidation are not worth the cost. As Ed Zitron notes, many AI tools slow down workflows instead of speeding them up.

Negotiate better terms with AI providers to control costs

Push for transparent billing and token usage limits. Companies like Uber found themselves spending entire annual budgets in a quarter, showing how easy it is to overspend without clear controls. Negotiate contracts that align usage with actual value delivered, not just token consumption.

Invest in AI only when the ROI is clear and measurable

Hold off on AI adoption until you can tie outcomes to specific metrics. If you can’t measure how an AI tool improves quality, reduces errors, or cuts time spent on tasks, it’s not worth the risk. The AI bubble burst is not a hypothetical, it’s a growing concern for businesses that can’t track or justify their AI spending.

The Road Ahead: Preparing for AI’s Next Chapter

AI may become more cost-effective as the bubble deflates

As the AI bubble bursts, the industry will likely see a shift toward more transparent and sustainable pricing models. Companies that survive will be those that align costs with actual usage, reducing the hidden token burn that has plagued the sector. This could mean lower long-term costs for businesses that are willing to wait for the market to stabilize.

Only the most practical and useful AI tools will survive

The AI tools that remain relevant will be those that deliver clear, measurable value. Tools that slow down workflows or produce low-quality outputs, like some AI-generated code, will be abandoned. Businesses should focus on tools that integrate smoothly into existing processes and reduce manual effort, not those that add complexity.

Apple and other companies must be ready to pivot quickly

Apple and other major players need to reassess their AI strategies and be prepared to shift course if current models fail to deliver. As Ed Zitron noted, the AI boom is not just a technological shift, it’s an economic one. Companies that can adapt to more practical, cost-effective AI solutions will be the ones that thrive when the bubble bursts.

Source: macrumors.com