“I was told there would be disinflation.”
The AI jokes just write themselves on some days. Thursday was one of those days.
I won’t bury the lede. Apple’s raising prices. Globally and more or less across the board, although the iPhone, the watch and AirPods were spared. For now.
You can thank AI, and specifically the memory shortage. “The rapid expansion of AI data centers has created an extraordinary surge in demand for memory and storage,” Apple said Thursday, explaining the price hikes. “We’ve never seen a component price increase this much, this quickly.”
This is the furthest thing from surprising to industry watchers. And Tim Cook tipped the price hikes in a Wall Street Journal interview earlier this month. “Unfortunately, price increases are unavoidable,” he said, adding that although Apple’s done everything it can to “mitigate” what he described as “huge” component cost increases, “the situation has become unsustainable.”
And, so, you’ll pay more for your gadgets so that shareholders like me, who haven’t upgraded their iPhone in half a decade, won’t have to suffer any margin contraction. I appreciate your sacrifice.
Jokes aside (and some of you probably don’t think that’s especially funny), Apple’s last two quarterly reports were notable for what didn’t show up: Evidence that surging memory costs are impacting the company’s operating results.
Now, the question’s whether consumers will balk at the price hikes to the detriment of Apple’s top line. Or keep buying, satisfied that Cook’s being honest to say he “shielded customers” as long as he could.
“We know this is not welcome news,” Apple sighed, in the Thursday statement quoted above. “And we are working tirelessly to find solutions.”
I don’t doubt that for a second. I realize Apple has its critics, but this isn’t a company that’s out to gouge people. Indeed, most management teams wouldn’t have adopted such an overtly apologetic cadence while raising prices.
As a habitual consumer of luxury items — and Apple’s higher-end products absolutely meet the definition of luxury items — I can personally attest that in some cases, there’s no announcement at all, let alone an explanation or any apology tour.
The market’s reaction was to punish the stock, which I think’s wrong. Not in a normative sense of the term, but rather in the sense that I doubt most people planning to buy, for example, a 15-inch MacBook Air for $1,299 are going to eschew that purchase at the new price, $1,499. And exactly no one pondering one of Apple’s highest-end Macs cares about a couple hundred bucks.
Also consider there are no “substitutes” for Apple products. Not really, and not if we’re being totally honest. Anyone who thinks there are — substitutes — doesn’t buy Apple devices in the first place. No one’s going to defect over a $200 price hike.
Let me dwell on that for a beat. An 18-year-old off to college and looking forward to a new MacBook isn’t going to happily accept anything other than a MacBook. Let’s be real here. I realize some of you are anti-Apple or anti-Mac and that you have your reasons, but nobody gets giddy for a Microsoft Surface. And nobody ever said, “This Dell XPS makes me feel like a kid at Christmas again.”
Besides, when it comes to sales, what matters most is the iPhone. Prices for those are stable. And anecdotally at least, AirPods have the most “ubiquity potential,” if you will, among products that aren’t the phone. Prices for those are stable too. It’s not an accident that Apple’s holding the line on prices for its two key products.
If you ask me, the big story here isn’t the cost of any individual gadget, nor even that the world’s most important consumer products company is raising prices due to an unprecedented surge in component costs.
Rather, the story is the ironic juxtaposition between AI as a disinflation savior (through the longer run productivity growth channel) and the AI buildout as a new, inflationary supply shock.
“With the backdrop of easing inflationary pressures from the energy sector, the reflationary angst has rotated toward the AI buildout,” BMO’s Ian Lyngen said Thursday, just hours before Apple’s official announcement.
“High technology prices throughout the US market would undoubtedly contribute to goods inflation, even if it is unclear whether the latest inflationary impetus would be sufficient to prevent falling energy costs from containing aggregate consumer prices during the next few months,” BMO’s US rates team went on, in the same note.
This comes at a time when the rebuilding of trade barriers already threatens to reverse decades of goods deflation. Cheap “stuff,” particularly falling prices for consumer electronics, was the door prize given to displaced manufacturing workers in Western economies during the hyper-globalization era. A lot of those workers came to see that as a raw deal, and disaffection among that demographic goes a long way towards explaining the rise of populism in the West.
Now, between the tariffs and the component supply shock created by the AI buildout, prices for items we’re used to seeing get cheaper could begin to rise instead, removing a (arguably the) key disinflationary tailwind that DM central bankers relied on to justify low rates for the better part of three decades.
Maybe AI-related productivity gains will offset this in time, but it’s a brave soul who makes policy today based on what might happen tomorrow, particularly if today’s policy decisions are at odds with what the here and now macro reality seems to demand.
As Lyngen put it, “It goes without saying that a near-term inflation spike associated with the AI buildout runs counter to the prevailing narrative that AI will be disinflationary over time and contain wages.”
Now you see why I was so keen on noting that Kevin Warsh’s antipathy to wordy FOMC statements didn’t keep him from adding a reference to AI-driven productivity gains in the new, slimmed-down Fed communications.
As I put it last week, drawing attention to the contrast between Warsh’s derisive attitude towards forward guidance and his faith in an unproven macro narrative, “It’s foolish to speculate publicly about the trajectory of rates but it’s prudent, apparently, to reference a disinflationary productivity renaissance that hasn’t happened yet.”
Of course, Warsh shouldn’t be expected to fret about the Apple price hikes. Nor any other price hikes for that matter. He’s worth $135 million on a poor day.
As someone who is definitely not an Apple fan, I can agree with your sentiment. The price hikes are unlikely to discourage buyers because their prices are not differentiated enough to cause discouraged buyers to defect to another product ecosystem. The reality is that all technology product manufacturers are going to have to raise their prices because everyone’s input costs are going up. Apple is smart enough to realize this and while they may be apologetic, they are certainly aware of what their consumers are willing to tolerate, and what pricing strategy best benefits their shareholders.
Is anyone concerned that anytime we talk about “containing wages” we are talking about death to the housing markets? The spending we’re doing may be for a new set of towels but not a new house.
Recently, Microsoft has joined Apple in deprecating perfectly good hardware to advance the software layer. Like you, I would be perfectly happy with my old iPhone 5s, but it was deprecated. And again, my iPhone 8 was deprecated and I now have an iPhone SEgen3 in my pocket. But Microsoft deprecated my Windows 10 PC workstation that has been running 3D software, trading platforms, dev and games since 2015 without a moment of lag. The support extension runs through October, but I replaced the PC a few months ago realizing component prices were rising and I was not saving time-value by putting the purchase off. The two most expensive parts were the DRAM (I need a lot of it…) and the Nvidia GTX card, which I required a 16gb version.
To augment today’s news and price action, I saw a Bloomberg article saying that OpenRouter traffic is expanding 5x in a year, with non-frontier model token usage gaining as well as the frontier leaders. Which begs two questions. When will cheaper tokens be good enough? Qwen will catch up to Claude Opus while the Administration kneecaps access to the true leading frontier models. So the desirability of OpenAI and Anthropic IPO shares is losing premium by the day. This may be different than the social media parallel of social media platform>internet connection since eventually a cheap token ‘thinks’ the same as an expensive one, and as in the article “You don’t need a PhD to collate receipts.” The token layer is being commoditize as we speak and we are watching ‘bottleneck’ stocks GEV, CAT, URI, PAVE, etc., rally hard. AI users are not locked in to an ecosystem like my iPhone use is. An iPhone SE is plenty good enough. I suspect the model labs are desperately looking for ways to gate and create ecosystems. Perhaps SaaS isn’t quite dead yet.
As long as hallucinations are an issue, you need a human in the loop. For use cases where false information is unacceptable, eg. medicine, this can mean the difference between a useable and unusable model, so $/token is less relevant. I asked Claude, and it said Claude has a 0% hallucination rate and frontier models are significantly ahead of open source, linking this to a test that showed a 35.9% hallucination rate for Claude (but it does seem like it was right the results are far worse for DeepSeek).
I’m still most comfortable with a burner phone.
I still rock a 1st generation iPhone SE (essentially an iPhone 6 in an iPhone 5 case). I love that it comfortably fits in my pocket.
Great article. I wonder how much the AI buildout is driving up the cost of the AI buildout? At some point, the competition for ever more expensive chips should crush the competition that simply can’t raise or generate the money to keep up anymore. That should eventually thin the field as the winners begin to snap-up the losers at bargain prices. That’s what eventually happened with the railroads.
GLM 5.2 is close enough to the Opus 4.8 benchmarks to make it a worthwhile open weight competitor.
The bigger concern for the frontier models should be the introduction of the Nvidia Spark RTX consumer grade chip with upwards of 192GB of unified memory. This provides a path to running very highly capable LLMs locally. How many people are going to need a subscription after that becomes widely adopted remains to be seen.
I’ve been using an apple for my entire professional career and would be loath to code on an xps. The amount of friction to use Linux or Windows is a total nonstarter. I or my employer would pay anything.
I was musing over another impact of AI-driven datacenter cost pressures. After reading about the heat wave in Europe forcing hospitals to postpone surgeries due to a lack of air conditioning equipment.
What a great headline it would be to read “Hospitals crowded out of cooling system market thanks to demand from datacenter construction”. Of course, I agree that datacenters are far more useful than some money-losing hospital.
I think domestically the interesting commentary is around government leaders telling farmers that they need to suffer through droughts while data centers have unlimited access to water.
Data centers do not inherently need any water to function. It seems data centers are today racing to find competitive advantages and if that means finding cheap power or water then so be it. IT is a paybook used by the aluminum industry for decades. I think once reality bites the business will evolve into self contained utilities and no water for the data centers and power plants feeding them.
“This comes at a time when the rebuilding of trade barriers already threatens to reverse decades of goods deflation. Cheap “stuff,” particularly falling prices for consumer electronics, was the door prize given to displaced manufacturing workers in Western economies during the hyper-globalization era. A lot of those workers came to see that as a raw deal, and disaffection among that demographic goes a long way towards explaining the rise of populism in the West.
Now, between the tariffs and the component supply shock created by the AI buildout, prices for items we’re used to seeing get cheaper could begin to rise instead, removing a (arguably the) key disinflationary tailwind that DM central bankers relied on to justify low rates for the better part of three decades.”
That’s it in a nutshell.
I read all the comments. Interesting to me to see how others see technology use. I am using Ubuntu. I have found ai helps me manage linux kernel.