Is the AI Investment Cycle Shifting Toward Memory Suppliers?

This article is adapted from Greed & Fear, a weekly newsletter written by Christopher Wood, Global Head of Equity Strategy at Jefferies. Wood has consistently ranked among the top equity strategists in major broker surveys across Asia over the past two decades.

AI-related capital spending continues to rise, but investor attention is shifting toward companies supplying critical components—particularly memory.

In a recent edition of Greed & Fear, Chris Wood argues that the multi-year AI investment cycle has entered a new phase, with pricing power moving from hyperscaler cloud providers to memory producers. SK Hynix and Micron have been key beneficiaries, with market capitalizations rising alongside contract prices for advanced memory.

That shift has come as cloud leaders such as Microsoft and Amazon Web Services continue to spend heavily on AI infrastructure but have seen weaker equity performance in recent months. These companies face growing pressure to demonstrate returns on what has become a multi-year investment cycle, even as some report year-over-year growth in AI-related revenue.

AI Infrastructure Economics Shift Toward Memory

High-bandwidth memory (HBM)—critical for training and running complex, data-intensive models—is in high demand and tight supply. Jefferies estimates that memory prices rose roughly 50 percent last quarter, reinforcing supplier leverage across the AI supply chain.

That pricing power carries significant cost implications. Building new fabrication facilities now requires capital outlays in the tens of billions of dollars, and Jefferies notes that some memory manufacturers are asking customers to help finance that expansion in return for assured supply.

Wood notes this departs from the long-standing expectation that large chip buyers would exert downward pressure on supplier margins.

Over time, memory producers such as SK Hynix are likely to face the same investor questions that have weighed on hyperscalers in recent quarters: when these AI infrastructure investments will begin to generate returns and, ultimately, profitability.

Wood also cautions that as the AI investment cycle matures, it may come to resemble other capital-intensive industries, where returns tend to normalize over time.

AI Growth Meets Energy Limits

Jefferies notes that alongside growing enthusiasm for AI memory providers, concerns persist around the energy demands of these technologies and whether the U.S. can meet them.

In October, OpenAI published an open letter urging the U.S. government to set an ambitious goal of adding 100 gigawatts of new power capacity per year.[1] Jefferies has previously noted that the current U.S. power grid—roughly 70 percent of which is more than 25 years old and showing signs of deterioration—may struggle to store and transmit that energy, even if generation capacity expands.

There is broad consensus that significant investment in domestic energy infrastructure will be required to support continued AI growth, from hyperscalers to component suppliers.

In some cases, technology companies are responding by securing power supplies directly, rather than relying solely on existing grid capacity. This week, Google announced a $4.75 billion deal to acquire Intersect, a wind and solar developer. The acquisition would make Google the first hyperscaler to own a power company.[2]

The Next Phase of the AI Cycle

While memory suppliers have emerged as near-term beneficiaries, it remains difficult to identify a clear, long-term winner in the next phase of the AI investment cycle.

The dominant theme is likely to be sustained investor pressure—on companies throughout the stack—to show that today’s record capital spending can translate into profitable growth. That scrutiny is already evident among hyperscalers and is likely to extend to critical component and AI infrastructure suppliers over time.

[1] https://ig.ft.com/ai-power/

[2] https://www.wsj.com/business/energy-oil/google-is-spending-big-to-build-a-lead-in-the-ai-energy-race-a8b5734a?gaa_at=eafs&gaa_n=AWEtsqdDhcrlcM196f5C8RAp-iIurGYIiZK9Q9FQ4aXS-1_XO1NVlbV8cC0jQ21M-gg%3D&gaa_ts=6982583d&gaa_sig=kZB-Cx1mnIqZJ8h-haE5QF2aS-394PZnxbNNkoodvMZTFEK7GzMzfe2zwyOl7t65KyGZsgIsVKx8LehDCD4XrA%3D%3D