MS Playing the AI Infrastructure Dip Where to Invest and Where We See Risk

Global Thematics Morgan Stanley & Co. LLC Global Insight

Dip: Where to Invest and Where

Morgan Stanley Asia (Singapore) Pte.+

Daniel K Blake Following large moves in AI infrastructure stocks, we revisit key Equity Strategist

debates and concerns (Chinese models, tokenmaxxing, political Morgan Stanley & Co. LLC pushback, "model lockdowns," physical bottlenecks) and ways to Robert S Kad play the dip. Bottom line: we're bullish on the "Intelligence Equity Analyst

Superhighway" but see key speed bumps ahead. Joseph Moore Equity Analyst Key Takeaways Brian Nowak, CFA Equity Analyst Our highest conviction view: the demand for compute is likely to significantly exceed supply for many years to come. Adam Wood Equity Analyst We disagree with the "tokenmaxxing" argument that Enterprise limits on token

spending will suppress AI revenue – we see the opposite playing out. Adam Jonas, CFA Equity Analyst Chinese LLM advances represent a real competitive threat to frontier model

developers, but only bolster our "Jevon's Paradox" bullishness on compute Morgan Stanley México, Casa de Bolsa, S.A. de C.V.+

demand. Nikolaj Lippmann Equity Strategist We are fundamentally bullish on the rate of improvement in AI capabilities, the

benefits of AI adoption, and associated capex. Morgan Stanley & Co. LLC

We list stocks to play each part of the AI infrastructure stack and explore David Arcaro, CFA Equity Analyst important potential Multipolar World policy dynamics with respect to AI

competition. Joe Laetsch, CFA Equity Analyst Given the significant selloff in AI Infrastructure Michael J. Cyprys, CFA, CPA stocks over the past few weeks, we assess the Equity Analyst

biggest investor concerns driving at least a portion

Cameron McVeigh, CFA of the weakness. To be clear, we believe a Equity Analyst meaningful driver of weakness has been technical

Concern #1: Reduction in "Tokenmaxxing" may limit revenue for LLM developers. Morgan Stanley does and seeks to do business with The issue here is the belief that companies will place compute budget limits on their companies covered in Morgan Stanley Research. As a result, investors should be aware that the firm may have a conflict of employees, as we have seen in a few high-profile examples. The data suggest that interest that could affect the objectivity of Morgan Stanley Research. Investors should consider Morgan Stanley this is not a valid concern. Research as only a single factor in making their investment (i) Ramp estimates find the current median token spend by Enterprise users is decision. extremely low (<$11/month), For analyst certification and other important disclosures, refer to the Disclosure Section, located at the end of this (ii) the economics of AI usage are deeply in the money (see our $55 labor cost report. saving/ $2-3 token cost math here), and += Analysts employed by non-U.S. affiliates are not registered with FINRA, may not be associated persons of the member (iii) Enterprise failure to adopt the most advantageous AI capabilities will lead to and may not be subject to FINRA restrictions on large competitive disadvantages, a dynamic we expect to become increasingly clear communications with a subject company, public appearances and trading securities held by a research analyst account.

Josh Baer, CFA over time. Equity Analyst

Concern #2: Chinese open-weight models may become a greater competitive Morgan Stanley Asia (Singapore) Pte.+ threat to American frontier LLMs, which in turn could drive lower spend on Jonathan F Garner compute to train LLMs. While reserving judgment on the capabilities of recently Equity Strategist released Chinese LLMs, we would cite the following caveats: (i) these models may Morgan Stanley Taiwan Limited+ actually be fairly compute intensive to achieve a given targeted result, implying that Charlie Chan the price of a token is not equal to the value achieved by token usage (see analysis Equity Analyst here), and (ii) we expect to see a range of LLM advances that drive more efficiencies,

which will sustain the drive for better AI results at lower cost (e.g., increasing the Morgan Stanley & Co. International plc+

usage of "orchestration layers" that allocate work to the most…