Spending on AI infrastructure continues to grow — major technology companies will invest several trillion dollars in the coming years to satisfy user cravings for ChatGPT and Claude. Bloomberg reports this.

Huge expenses are driven not only by hyperscalers like Microsoft and Meta investing in data centers but also by rising component prices needed to build massive computing facilities.

The capital investments of American hyperscalers may exceed $1 trillion by 2027. Source: Bloomberg.## “Chipflation” and investments

“Chipflation” has become a problem not only for the artificial intelligence sector. A shortage is also emerging in traditional microchips. This leads to rising prices for smartphones, gaming consoles, and other electronics.

The popularity of programming applications and AI agent usage creates demand for hardware components that support workflows: GPUs, memory, and CPUs. The latter previously played a secondary role in the AI revolution but are now essential for LLM workloads.

Some of the most profitable tech companies and funded startups are competing to secure enough equipment, fearing to fall behind in the race for superintelligence.

The largest microchip manufacturer, TSMC, plans to invest a record $56 billion. Elon Musk is considering building his own chip manufacturing plant. The facility would cost him between $55 billion and $119 billion.

The situation is driving up the stocks of AI equipment suppliers, which are outperforming most of their customers.

Microchip manufacturers are outpacing giants in the AI field. Source: Bloomberg.

“Chip companies thrive because of everyone above them in the supply chain,” said James Covello, global head of equity analysis at Goldman Sachs.

Recent reports from hyperscalers indicate a painful blow from inflation. Microsoft expects rising component prices to increase annual capital expenditures by $25 billion, with total spending reaching a record $190 billion. Meta has raised its forecasted expense range by $10 billion.

“Memory Tax”

Tech corporations are increasingly facing a “memory tax,” as advanced AI accelerators require much more high-bandwidth storage.

The three largest DRAM suppliers — SK Hynix, Samsung Electronics, and Micron Technology — have become favorites on the stock market. Their combined market capitalization exceeds $2.8 trillion.

Demand from data centers has contributed to SK Hynix’s increased profitability. Source: Bloomberg. SemiAnalysis estimates that total spending on various types of memory will reach 30% in 2026. In 2024, the figure was 8%.

Source: SemiAnalysis. Large investments in chips motivate hyperscalers to find ways to reduce computing costs. One approach is to use alternative AI processors like AMD’s. Some are developing their own solutions — tensor processors from Alphabet, Trainium chips from Amazon, and Maia 200 from Microsoft.

Other innovations include Google’s TurboQuant compression technology, which reduces memory expenses.

The culprit of inflation

The hype around artificial intelligence causes smartphone, gaming console, and PC manufacturers to struggle with securing memory chip supplies, as manufacturers prioritize more profitable markets and long-term contracts.

Consumer electronics manufacturers must either pass the cost increases onto consumers or reduce device specifications. The third option is to accept lower profits.

Building semiconductor factories takes years, so quick responses to demand are unlikely.

Given the rising energy costs caused by energy-intensive data centers, AI will continue to significantly impact inflation for some time, Bloomberg summarized.

Recall that Samsung’s semiconductor division posted record profits in the first quarter, surpassing analyst expectations.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.