Under the hood of the AI revolution: a new golden age for hardware manufacturing
The stock market surge so far in 2026 has been driven mainly by the technology giants (Alphabet, Amazon, Microsoft, Meta), whose total investment budgets now exceed $750 billion – $80 billion more than expected at the start of earnings season and 83% higher than spending in 2025. The surge in AI capital expenditures (capex) shows no signs of slowing down.
The surge in investment expectations is driving a similar rise in profit expectations for AI infrastructure companies, which is improving the broader market’s earnings outlook and leading to upward revisions in their estimates for the S&P 500’s earnings per share (EPS).
Trends in earnings forecasts for chip sector companies by year-end
While the stock market celebrates, a serious contradiction looms in the background: the pace of software demand and capital investment has far outstripped the physical capacity of hardware manufacturers. This gap has forced companies like OpenAI and Anthropic to limit their services due to “computational hunger.”
The hierarchy of critical bottlenecks
The industry is no longer struggling with a shortage of a single chip, but with a complex, interdependent system of constraints:
- HBM (High Bandwidth Memory) – the main obstacle: modern AI models require faster data access than ever before. Memory manufacturers (SK Hynix, Micron, Samsung) have sold out all their capacity through the end of 2026. Although software algorithms (e.g., Google TurboQuant) are attempting to reduce memory requirements, the physical shortage is expected to persist for another three years.
- GPUs and manufacturing constraints: although Nvidia dominates, the bottleneck here is no longer just design, but TSMC’s “CoWoS” packaging technology. The most advanced plants are running at full capacity, and building a new factory takes 2–3 years, so supply cannot immediately keep pace with market surges.
- CPU – the unexpected comeback: while CPUs were previously thought to have been pushed into the background in the AI era, the new type of “agent-based” AI systems (which plan, reason, and execute tasks) require much more processor coordination. While a chatbot required one CPU for every 12 GPUs, this ratio has shifted to 1:1 for agents, giving Intel a massive boost.
- Physical and political barriers: beyond chips, the construction of data centers is also hitting roadblocks. Political resistance is growing in the U.S. and Europe due to the massive power consumption, which is delaying projects. Power supply and cooling technologies (liquid cooling) have become the new critical elements of infrastructure.
Currently, hardware manufacturers are calling the shots. Since they are unwilling to risk “overbuilding,” the shortage may persist. This creates a peculiar situation: the world’s richest companies (Microsoft, Google) are lining up for chips, while limited supply has caused rental fees for older models (e.g., Nvidia H100) to rise by 30%. While cloud service providers have tripled their spending, hardware manufacturers (such as TSMC or memory manufacturers) are more cautious: they have increased their investments by only 50%, fearing that overcapacity will lead to losses later on. This contrast between “hardware caution” and “software euphoria” could mean that the pace of AI development in 2026 will be determined not by the genius of engineers, but by the speed at which factories are built and the number of silicon wafers produced.
Capital Offence AI Supply Chain, Capital Expenditures
Return Download
This is a marketing communication. Making a well-informed investment decision requires obtaining detailed information. Please read the Key Information Document, the official prospectus, and the management regulations available at the distribution points of the Fund and on the website of the Fund Manager (www.vigam.hu) for detailed information regarding the Fund’s investment policy, distribution costs, and the possible risks of investing. Costs related to the distribution of the investment fund (purchase, holding, sale) can be found in the Fund’s management regulations and at the distribution points. Past performance is not a reliable indicator of future returns. Future returns from the investment may be subject to taxation, and tax and duty information relating to individual financial instruments and transactions can only be accurately assessed based on the individual circumstances of each investor, which may change in the future. It is the investor’s responsibility to obtain information regarding tax obligations.
The data contained in this information material are provided for informational purposes only and do not constitute investment advice, an offer, or investment consulting. VIG Investment Fund Management Hungary Ltd. accepts no liability for investment decisions made based on this information or for their consequences. The license number of the Fund Manager for alternative investment fund management (AIFM) is: H-EN-III-6/2015. The license number of the Fund Manager for UCITS fund management (collective portfolio management) is: H-EN-III-101/2016.