Last updated: July 28, 2026 · By Vishal Swami, Founder & Lead AI Reviewer, AISagely
Apple Will 'Watch Everything Burn' When the AI Bubble Bursts is the read AI critic Ed Zitron gave MacRumors on July 27, 2026, and the spending numbers back him up. Apple spent $12.7 billion on capital expenditures in all of fiscal 2025, a fraction of what Amazon, Alphabet, and Meta committed to AI infrastructure that same year, so Zitron's argument is that Apple has the least to lose if the spending never turns into profit.
Short answer: Ed Zitron argues Apple can sit out an AI bust because it spent only $12.7 billion on capex in fiscal 2025, versus roughly $288 billion combined at Amazon, Alphabet, and Meta. Less exposure means less to lose if AI spending doesn't pay off — but it's also why Apple Intelligence still trails ChatGPT and Gemini on everyday tasks today.
I test AI tools daily, including the ones built into an iPhone, so this headline landed differently for me than it probably did for most readers. It isn't really a story about Apple. It's a story about what happens to your own AI subscriptions and workflows if the companies bankrolling them can't make the math work — and Apple's low-spend position is a useful yardstick for judging your own exposure.
What Ed Zitron actually said
Zitron, who writes the Where's Your Ed At newsletter and hosts the Better Offline podcast, told MacRumors that if the AI infrastructure boom collapses, "I think they will sit on the sidelines and watch everything burn." His case isn't that Apple is smarter than Microsoft, Google, Amazon, and Meta — it's that Apple never joined their data-center arms race, so it has almost nothing to unwind if the spending doesn't produce returns. He also said Apple could use a downturn to make selective acquisitions once asset prices drop, rather than being forced to write down its own infrastructure.
The scale gap is the whole argument. Those four rivals have spent more than $1 trillion in combined capital expenditures since 2022, and Zitron estimates they'd need roughly $1.5 trillion in new profit to justify it. Apple's fiscal 2025 capex was $12.7 billion, up 35% year over year but still dwarfed by what its rivals had already committed for the same year: Amazon's 2025 capex forecast sat at $125 billion, Alphabet's expected spend was $92 billion, and Meta had committed $71 billion specifically to AI chips and infrastructure, according to reporting confirmed in late October 2025. Microsoft doesn't publish a single annual AI figure the same way, but it spent $34.9 billion in the September 2025 quarter alone, nearly three times Apple's entire year.
The bubble concern isn't hypothetical, either. OpenAI's leaked 2025 financials, independently verified by the Financial Times and reported by Fortune in June 2026, showed the company lost $20.92 billion from operations on $13.07 billion in revenue. Zitron's broader point is that consumer AI pricing is subsidized so heavily that a single user can burn hundreds of dollars in compute on a $20-a-month ChatGPT plan, the kind of math that only survives as long as investors keep funding the gap.
Apple isn't AI-free, though. It reportedly pays Google roughly $1 billion a year to license Gemini for the next generation of Siri, and Zitron describes Apple Intelligence itself as "barely-functional." In my testing across iPhone, iPad, and Mac since launch, that tracks — Apple's writing tools and notification summaries are usually the first thing I bail out of in favor of ChatGPT or Gemini for anything beyond a quick rewrite. Apple is treating AI as a commodity to license, not a product to build, and Zitron's read is that this restraint is exactly what protects it.
What you need before you check your own exposure
You don't need a trading terminal to run this same check on your own AI spending. Pull up a list of every AI subscription and API bill you pay for this month, and if the vendor is public, glance at its most recent earnings call or capex disclosure. Ten minutes is enough. The goal isn't predicting the stock market, it's figuring out which of your own AI tools look like Apple, spending conservatively and hard to dislodge if funding tightens, and which look more like the hyperscaler bets Zitron is describing: priced as if 2026-level growth continues forever.
Step-by-step: how to bubble-proof your own AI toolkit
1. List what you're actually paying for
Write down every AI tool with a recurring charge: ChatGPT Plus, Claude Pro, a Midjourney plan, an API key billed by usage. Most people underestimate this list until they check their card statement.
2. Check whether the price covers what you use
If a tool's monthly fee is far below what the underlying model actually costs to run, that gap is being paid by someone else's investor money. That's fine today, but it's the first price to move if funding gets tighter.
3. Find your single points of failure
Apple's roughly $1 billion-a-year Gemini deal is a single point of failure dressed up as a partnership. Ask the same question about your own stack: if your main AI vendor raised prices threefold tomorrow, do you have a fallback, or does your workflow just stop?
4. Keep one free or low-cost fallback per task
You don't need a backup for everything, just for the one or two tools your work actually depends on. My free AI tools roundup covers where you can test an alternative before you need it in a hurry.
5. Watch capex and earnings calls, not just product launches
A flashy product demo tells you nothing about whether the company behind it can keep the lights on at current prices. A capex number or a leaked income statement, like OpenAI's, tells you a lot more.
6. Re-run the check every earnings season
Big tech reports quarterly. Set a recurring reminder to skim the AI capex headlines for whichever vendors you actually pay, rather than reacting only when a story like this one goes viral.
Example prompts you can copy
- Audit your subscriptions: "Here's a list of my AI subscriptions and what I use each for [paste list]. Which ones would I miss least if the price tripled, and which have no real substitute?"
- Stress-test a vendor dependency: "I rely on [tool] for [task]. List two realistic alternatives I could switch to within a week if the price or terms changed."
- Separate hype from utility: "Explain what would actually break in my workflow if [AI tool] shut down tomorrow, versus what I'd just find annoying."
- Track the money, not the marketing: "Summarize [company]'s most recent earnings call or financial disclosure in three sentences, focused only on AI spending and revenue, no marketing language."
Common mistakes to avoid
The mistake I see most is treating "free" or deeply discounted AI features as a permanent price rather than a promotional one; subsidized pricing is a choice a company makes while it's chasing growth, not a guarantee. Second is assuming Apple's caution proves Apple is right. Being under-invested protects you from a bust, but it's also why Apple Intelligence still ships behind ChatGPT and Gemini on raw capability. Third is vendor concentration: putting an entire workflow behind one API with no fallback, then being surprised when a price change or outage stops your work cold. Fourth is overreacting the other way, canceling every AI tool the moment a "bubble" headline runs, instead of just checking which specific tools you actually depend on.
Apple vs. the hyperscalers: 2025 AI capital spending compared
| Company | FY2025 capital spending | Scale vs. Apple | What it signals |
|---|---|---|---|
| Apple | $12.7B (all capex, +35% YoY) | 1x (baseline) | Buys AI as a commodity; least exposed if it busts |
| Amazon | $125B (2025 forecast) | ~10x | Deepest infrastructure bet of the four |
| Alphabet | $92B (expected) | ~7x | Backed by a highly profitable ads business |
| Meta | $71B (AI chips & infrastructure) | ~6x | Zuckerberg's AI bet is the least hedged of the four |
| Microsoft | $34.9B (September 2025 quarter alone) | ~3x per quarter | OpenAI's largest backer and largest creditor |
Tools that make this easier
You don't need to predict a bubble to make better AI-tool decisions, you just need somewhere honest to check a tool before you depend on it. My AI tool ratings and AI tool reviews hubs cover where each major assistant is genuinely strong versus where it's still guessing, which is the same question you should ask before building a workflow around any one vendor. If you're deciding whether a small business should lean on AI at all given the pricing uncertainty above, my best AI tool for small business guide covers lower-risk starting points. For the cross-checking habit this whole approach depends on, my guides to using Claude and using Gemini walk through getting a genuinely independent second opinion instead of leaning on a single vendor. If you're specifically weighing Apple's own AI stack, my guide to using ChatGPT on iPhone covers the fallback I actually use when Apple Intelligence comes up short. And if budget is the reason you're exposed to one vendor's pricing in the first place, my free AI tools roundup is where to start swapping in a cushion.
My take
Zitron's framing is a good headline, but the underlying bet is one any AI-tool user can copy at a much smaller scale: don't build anything you can't afford to lose the day the subsidy disappears. Apple's $12.7 billion looks timid next to $288 billion at three of its rivals, and that restraint is genuinely why it has less downside if the industry's spending doesn't pay off. It's also why Apple Intelligence is still the AI feature I trust the least on my own phone. Both things are true at once, and the same trade-off is sitting inside your own AI subscriptions right now, whether you've priced it in or not.
Frequently Asked Questions
Is Apple really immune if the AI bubble bursts?
Not immune, but less exposed. Apple's $12.7 billion in fiscal 2025 capex is a fraction of what Amazon, Alphabet, and Meta spent, so it has far less infrastructure to write down if AI spending doesn't produce returns. It would still feel a slowdown in AI-driven hardware demand and its Google Gemini licensing costs.
What did Ed Zitron actually say about Apple?
Told to MacRumors on July 27, 2026, Zitron said that if the AI spending boom collapses, "I think they will sit on the sidelines and watch everything burn," adding that Apple could use the downturn to make selective acquisitions rather than absorb losses of its own.
How much is Apple spending on AI compared to Microsoft and Google?
Apple's total fiscal 2025 capex was $12.7 billion. In the same period, Amazon forecast $125 billion, Alphabet expected $92 billion, Meta committed $71 billion to AI chips and infrastructure, and Microsoft alone spent $34.9 billion in a single quarter.
Is the AI bubble actually going to burst?
No one can say for certain, but the financial strain is real and documented: OpenAI's leaked 2025 financials showed a $20.92 billion operating loss on $13.07 billion in revenue, verified by the Financial Times. That doesn't guarantee a collapse, but it shows the current pricing isn't yet self-sustaining.
What should I do with my own AI subscriptions if this is true?
List what you pay for, check whether any single vendor is a point of failure, and keep one free or low-cost fallback ready for the tool you rely on most. That's the same protection Apple's low-spend position gives it, just applied to a personal budget instead of a balance sheet.