Apple’s Price Hikes Signal AI-Driven Cost Surge
Rising hardware prices reflect how AI infrastructure is straining supply chains and fueling inflation across consumer tech.
This article originally appeared on Weijin Research on Huxiu on June 26, 2026. Original Chinese title: 「苹果为AI放出了通胀压力」. It has been translated and adapted for an English-speaking audience.
For decades, one of Silicon Valley’s most successful narratives has been selling the world on a future of abundance. Moore’s Law is the physical embodiment of that creed, and the consumer electronics industry has served as its most faithful evangelist for “technological deflation.” The one blemish was that professional services remained stubbornly constrained by Baumol’s cost disease. So people hoped this wave of AI would finally drag that sector into the deflationary orbit as well.
Yet reality veered in a different, unsettling direction first. The promise of technological deflation has not materialized; instead, asset inflation and inflation in factors of production have arrived ahead of it.
This summer, Apple raised prices. Macs, iPads, HomePods, and even the Vision Pro have all been adjusted upward, with price hikes reaching 20 percent. The iPhone price has held steady for now, but the market widely expects it is only a matter of time. The next-generation iPhone needs more than 12GB of RAM to support the full Apple Intelligence experience, and it must compete with AI servers for production capacity. Bloomberg predicts that the average selling price of iPhones will rise 12 percent this year.
For decades, Apple has wielded the strongest supply-chain management in the global consumer electronics industry and has been one of the most profitable device makers. When Lei Jun and Yu Chengdong each took their turn complaining to the cameras about rising costs, the market believed Apple would absorb the increases to hold its share. But now Apple is publicly saying it can no longer hold the line.
Outgoing CEO Tim Cook described the cost shock as a “once-in-a-century flood.” The supply-chain master lamented that in his more than forty-year career, he had never seen such a rapid price spike. Elon Musk, another supply-chain management virtuoso, quickly chimed in, calling it the most violent price revolt he had ever witnessed.
This sends a signal to the entire consumer electronics industry. Research firm Counterpoint expects other brands to follow Apple’s lead, raising prices on some products, trimming discounts on entry-level models, or shifting their product focus further toward the premium segment.
The source of this cost tsunami lies in the near-insatiable demand curve of AI data centers. For decades, consumer electronics dictated the rhythm of the semiconductor industry. The entire supply chain revolved around billions of consumers. The most advanced process nodes usually landed in Apple’s phones first. Now, the massive capital expenditure on AI infrastructure is devouring the production inputs that consumer electronics depend on, pushing cost pressures all the way down to ordinary consumers’ pockets.
Apple is searching for a foundry alternative to TSMC and has been in talks with Intel for over a year. But Intel, in turn, is tied up with Elon Musk’s TeraFab project.
Right now, everything must make way for AI. Packaging DRAM and NAND into HBM and HBF and selling them to compute-starved hyperscale cloud providers and large-model companies offers far more market upside. More and more DDR5 is being stacked into HBM. Even DDR5 itself has found new use cases in the pre-fill stage of AI inference.
Market research firm TechInsights reports that prices for DRAM memory and NAND storage chips have quadrupled over the past 12 months. This showed up in Micron’s latest earnings: quarterly revenue rose 346% year-on-year, and gross margin surged from 39% to 84.9%.
Micron executives judge that tight supply conditions will persist through 2027 and beyond. Morgan Stanley predicts that during this period, consumer electronics could face a wafer supply shortfall of as much as 15%.
The computing boom has spread from GPUs across the board to memory, hard drives, and CPUs. Shortages and price increases have followed. Whether in the US, South Korea, Japan, or China, capital-market hotspots are moving from chips to semiconductor equipment and energy equipment, then on to critical materials, components, and even assembly. Other assets have fallen out of favor.
The bigger shock has yet to arrive. The US has announced data-center spending, but only a small fraction has actually been deployed. The funds OpenAI and Anthropic are expected to raise in their upcoming IPOs could further fuel the AI infrastructure wave. Big tech’s cash flows are shrinking fast. SpaceX is issuing debt just to keep its AI infrastructure buildout going. Even deep-pocketed Nvidia and Google have turned to large-scale debt financing. By 2032, US investment in AI infrastructure is projected to reach roughly $8 trillion, enough to buy every piece of real estate in New York City five times over.
Stacked area chart showing AI infrastructure’s share of nominal US GDP from 2015 to 2026, broken down by compute hardware (non-AI and AI-related), data center construction, and networking. The chart includes a 2015–2022 trend extrapolation and projects AI-related infrastructure to reach approximately 0.83% of GDP by 2026.
Before his appointment, current Fed Chair Warsh wrote that “AI will be a significant deflationary force.” The data, however, delivered a sharp rebuke. U.S. headline inflation in May climbed back above 4%. More factors of production are now inflating. Goldman Sachs economists expect consumer electricity prices to rise about 6% per year. Localized labor shortages are also pushing up costs in other sectors.
Perhaps neither the Fed nor Silicon Valley is wrong. As a general-purpose technology, AI is inherently disinflationary. The real problem, though, lies in timing — and in what the path to that destination looks like.
The worry is this: if API prices for the highest-intelligence-density tokens keep rising overall, and if completing a task requires ever more tokens for repeated reasoning, control, verification and correction, then AI’s advantage over professional services workers will disappear. Baumol’s cost disease would then not simply go away on its own. Economists at UBS have had to revise their forecasts. Even under the most aggressive timeline, AI is at least two to three years away from beginning to help bring inflation down.
The market mechanism has already started making small corrections on its own. SK Hynix plans to shift some HBM capacity back to standard DRAM, because gross margins for the two have now inverted. Silicon Valley giants are also taking a more sober view of the tokenmaxxing race.
But the scarcity problem hasn’t really been solved. Price increases will keep finding new ways to work their way back into the price system.
On one hand, AI scarcity is not just technological — it is also institutional. This wave of the AI revolution is unfolding in an era of highly fragmented supply chains, where global manufacturing capacity cannot flow efficiently.
In Jensen Huang’s “five-layer cake” theory, energy, chips and infrastructure are the physical scarcities. That, in turn, creates scarcity in the most cutting-edge model capabilities. It’s why Anthropic placed a high-premium compute order with SpaceX, setting Elon Musk thinking about construction projects in space.
On the other hand, many proponents of open-source models see America’s AI development model as far more “extensive” than the approach taken by Chinese AI companies, which constantly squeeze every drop of efficiency out of their compute. American firms seem to burn through expensive GPU and HBM resources with little restraint. That, in turn, makes the whole endeavor ever more capital-intensive.
A more accurate way to put it, however, is that the U.S. is pursuing a capital-driven, time-first development model. In an environment where capital is abundant, fundraising is easy, and the rewards for being first are enormous, it makes sense to throw more GPUs and HBM at the problem to shorten development cycles rather than sacrifice speed to improve resource utilization. For OpenAI and Anthropic, the timing of a model’s release is itself part of competitive advantage. As long as marginal gains are still there, the opportunity cost of slowing R&D to squeeze more from existing compute could far exceed the cost of buying new capacity.
Closed-source competition further reinforces this capital-heavy model. Because rivals’ model capabilities, training progress and next-generation product timelines are all shrouded in secrecy, no company wants to be the first to ease up on capital spending. Misjudging a competitor’s progress could mean losing not a few percentage points of profit but your entire market-leading position.
The rapid catch-up of Chinese open-source models further intensifies the logic of racing against the clock. From DeepSeek-R1 to GLM-5.2, Chinese open-source models have repeatedly defied expectations that the gap with the U.S. would keep widening. For OpenAI and Anthropic, if their lead is not large enough to constitute a generational gap, it becomes hard to maintain commercial pricing power over open-source alternatives based on a narrow performance edge alone.
Almost every general-purpose technology revolution has been accompanied by an episode of capital over-allocation. Railroads had it. The internet had it. Today’s AI is unlikely to be an exception.
Capital often arrives at the future before productivity does; scarcity, for now, has arrived ahead of abundance. Inflation in capital and factors of production may simply be the price that has to be paid to build tomorrow’s infrastructure in advance.
As Peter Thiel has worried, what is truly unsettling may be that AI has almost become the only technological direction still moving at high speed. Resources that were once shared across multiple technology frontiers — capital, talent, advanced manufacturing capacity, energy infrastructure — are increasingly being reorganized around this single general-purpose technology. It is an unbalanced structure of innovation.
Apple’s price increase this time may become the occasion for the market to rethink: how high a price it is willing to pay for such a future.