As AI demand surges, costs are also rising sharply! It is reported that Taiwan Semiconductor is planning a price increase of up to 10% in 2027, raising concerns about "chip inflation."

Zhitong
2026.07.21 10:43

According to media reports citing informed sources, Taiwan Semiconductor is in discussions with clients about plans to raise chip manufacturing prices by up to 10% in 2027, in order to cover the rising production costs caused by surging AI demand, conflicts in the Middle East, and supply chain disruptions. The increase, which will range from 5% to 10%, is set to take effect next year and will apply to both advanced and mature process chips. This move aims to address the pressures of soaring costs for materials, equipment, and electricity, and to support its accelerated expansion to meet the demand for AI chip supplies from clients such as NVIDIA and Apple

According to media reports citing informed sources, Taiwan Semiconductor Manufacturing Company (TSMC.US) has begun discussions with clients to plan a price increase of up to 10% for chip manufacturing by 2027, in order to cover the rising production costs. The report states that as a major chip foundry supplier for NVIDIA (NVDA.US) and Apple (AAPL.US), TSMC initiated related negotiations in June and finalized a basic price adjustment plan this month, with an increase ranging from 5% to 10%. These price adjustments are expected to take effect next year, covering both advanced process chips and mature process chips.

TSMC manufactures chips for many global tech giants, including Google (GOOGL.US), Amazon (AMZN.US), in addition to the aforementioned NVIDIA and Apple. For a long time, TSMC has resisted the severe price fluctuations seen in the memory chip industry and has consistently emphasized building long-term partnerships with clients to cope with the industry's cyclical volatility. Meanwhile, global supply chain disruptions caused by conflicts in the Middle East and surging demand in the AI sector are driving up production costs and increasing pressure on TSMC to accelerate capacity expansion globally.

Clients, including NVIDIA, have been urging TSMC to speed up production expansion due to concerns about bottlenecks in the supply of key components needed for AI accelerators and data centers. In response, TSMC has launched a large investment plan, including the Arizona project, which is considered the largest foreign direct investment project in U.S. history.

TSMC and other chip manufacturers are facing pressure from soaring costs in various production processes, including materials, equipment, and electricity costs. This month, TSMC raised its capital expenditure forecast for 2026, now expecting capital expenditures to reach $60 billion to $64 billion, up from the previous estimate of $52 billion to $56 billion. The company anticipates that the growth in artificial intelligence (AI) demand and rising costs for capacity expansion will drive increased investment, especially in advancing the $265 billion expansion plan in Arizona.

Reportedly, TSMC has postponed the price increase until 2027 to allow clients time to adjust. TSMC CEO C.C. Wei stated during an analyst Q&A after announcing better-than-expected financial results in July, "We will not suddenly raise prices. We obtain reasonable returns through our own value and ensure that our profits and gross margins are sufficient to support long-term sustainable expansion. This is beneficial not only for clients but also for TSMC; this is our business philosophy."

In a statement on Tuesday, TSMC said, "Our pricing strategy is strategic, not opportunistic. We will continue to work closely with our clients and demonstrate our value to them."

It is worth mentioning that reports of TSMC's price increase next year come amid rising concerns about "chip inflation." According to investment firm Susquehanna's latest report, the global semiconductor industry's delivery cycles further lengthened in June, and this trend remains significant even against the backdrop of rising prices Analyst Christopher Rolland pointed out that in June, the semiconductor industry's delivery times recorded the largest month-on-month increase since the current cycle, rising by 5 days to 19.4 weeks. More notably, the industry pricing saw the "largest month-on-month increase" in June, expanding by 5% compared to the previous month. The simultaneous acceleration in delivery times and price increases highlights that the chip supply-demand situation continues to tighten.

Additionally, the growth in delivery times in June was "broad-based"—about 81% of the companies covered reported stable or rising delivery times, and all distributors experienced growth. Delivery times for all product categories increased month-on-month, and analysts believe this "indicates that the upward cycle is now expanding beyond analog parts."

In contrast to the ongoing tight supply-demand fundamentals, U.S. chip stocks experienced significant volatility in July. The Philadelphia Semiconductor Index has fallen about 17% in July, although it is still up 65% year-to-date. The index dropped about 10% last week, marking the largest weekly decline in over a year, and has retreated more than 20% from its historical high in June, officially entering a technical bear market.

Behind the dramatic fluctuations in chip stock prices, the simultaneous acceleration in delivery times and price increases seems to indicate a deeper structural contradiction is emerging—"Chipflation." Julia Hermann, a global market strategist at New York Life Investment Management, recently warned that "Chipflation"—the soaring prices of AI-related logic chips and memory chips—will become the next headwind testing the resilience of AI trades.

In an interview, Julia Hermann pointed out, "Hyperscale cloud service providers are now caught in a dilemma. On one hand, there are rising input costs—chip prices are increasing, along with rising energy and utility costs; on the other hand, the realization of investment returns still takes years. We believe this environment will truly test market confidence—if investors still believe in the long-term potential of AI trades, they may tolerate short-term volatility and a slowdown in monetization."

Julia Hermann noted that one of the best indicators to observe memory chip inflation is the South Korean DRAM export price index. In past cycles, the year-on-year growth rate of memory chip prices peaked around 100%, while currently, the year-on-year increase in South Korean DRAM prices is as high as 370%. In her view, while the surge in chip prices is indeed a sign of strong demand, it is also a double-edged sword—persistently high prices will significantly raise the construction costs of AI infrastructure, which may, in turn, suppress or even end the current AI capital expenditure boom. Therefore, she is currently focusing on "quality" in the AI supply chain, which includes robust profitability, moderate earnings volatility, and sufficient interest coverage.

Recently, SK Hynix Chairman Chey Tae-won also issued a warning about chip inflation, stating that it is not normal for the memory market to maintain high prices for an extended period. He expects that global semiconductor demand will expand significantly next year (2027), with demand in the AI sector growing by 60% to 100% compared to this year, and overall semiconductor demand will also increase by at least 50% to 60%. However, the new supply next year is expected to be "almost zero," and the supply-demand gap may further widen In response to concerns that capacity expansion may prematurely end the current "super cycle," Cui Taiyuan provided a thought-provoking reply. He stated that current chip prices are at abnormally high levels and should naturally decline. If prices continue to rise and further exacerbate "chip inflation," the semiconductor industry will ultimately suffer backlash. However, he clearly stated that increasing supply and driving prices down does not mean that companies cannot be profitable