‘Peak Semiconductors?’ The Answer May Depend on Next Week’s Big Tech Earnings Rally

[Financial News] As Morgan Stanley’s view that semiconductor stocks may have already peaked spreads, analysts said Google, Microsoft, Meta Platforms and Amazon.com, Inc., which begin reporting earnings next week, could raise their AI-related capital expenditure targets from previous levels. The outlook reflects expectations that a huge market for enterprise AI agents is opening up. Because Big Tech investment plans are directly tied to forecasts for memory chip demand, earnings reports from these companies through the end of the month are expected to serve as a key gauge for judging the semiconductor cycle and whether semiconductor stocks have reached a peak.
On the 13th, Hana Securities Research Center said the key points to watch in Big Tech earnings reports from next week through the end of the month are capital expenditure guidance, the pace of monetization from AI investments, and future business strategies. It also forecast that Big Tech companies are likely to expand capex guidance further in response to market demand for AI. Google (Alphabet) will report on the 22nd, followed by Microsoft (MS) and Meta on the 29th, and Amazon on the 30th.

Kang Jae-gu, an analyst at Hana Securities, said that once corporate AI adoption becomes widespread, mega-platform companies such as Google and Amazon could have room to increase capital spending to meet enterprise demand. He added that “an even bigger market is opening up” in AI. Amazon is issuing more than $25 billion in corporate bonds to fund AI infrastructure investment and other needs. Its capex this year is expected to approach $200 billion. Last month, NVIDIA, the world’s No. 1 AI chipmaker, and recently listed SpaceX each issued $25 billion in dollar-denominated bonds.
The wave of debt-fueled investment by Big Tech is based on optimism that the AI market is opening up. According to Google Cloud’s 2026 State of Infrastructure in the Agentic AI Era, the commercialization of agentic AI has already pushed inference AI demand past training AI demand, and 83% of respondents said they need to upgrade their existing infrastructure to run production-grade agentic AI. However, some say Big Tech’s AI and AI computing services will remain unavoidable because of challenges in securing AI accelerators, as well as data transfer and engineering costs.
Kang said recent reports that Meta is seeking to enter the AI infrastructure business and that SpaceX is pursuing an AI computing rental business support the view that the capex cycle for mega-platform companies could last longer than the market expects. He added that NVIDIA’s decision, starting with its earnings report for February to April this year, to break down data center results into categories such as hyperscalers including Big Tech, AI cloud, and industrial and enterprise customers also shows that demand is rising not only from Big Tech but from general companies as well.
Still, some point out that there is not yet clear evidence or a concrete strategy showing that Big Tech’s AI investments will translate into revenue. Morgan Stanley’s recent warning about a peak in memory semiconductors suggests that the pace of improvement in earnings estimates for chipmakers and in memory chip price increases is nearing a high point. If AI investment growth among Big Tech or hyperscalers slows, memory chip companies’ earnings momentum would also have to weaken.
By contrast, major domestic brokerages say memory demand remains solid. In a report released on the 9th, Kim Dong-won, an analyst at KB Securities, said DRAM and NAND flash memory production capacity in 2027 will rise only 7% and 4%, respectively, from a year earlier on a wafer basis, while demand is expected to grow 17% and 19%. He said the memory supply shortage will deepen next year compared with this year. He also projected that global AI investment, a barometer of memory chip demand, will expand nearly threefold from $390 billion in 2025 to $1.1 trillion in 2027.
Kim Sun-woo, an analyst at Meritz Securities, also said memory supply is still far short of keeping up with demand growth at least through the fourth quarter of next year. He added that “the memory industry is still not even in the middle of its cycle.”
ehcho@fnnews.com Jo Eun-hyo Reporter