Morgan Stanley, previously known as the 'semiconductor grim reaper' for repeated pessimistic forecasts on Korean semiconductors, has issued an optimistic outlook on the artificial intelligence industry. According to Morgan Stanley Research, AI infrastructure-related stocks including semiconductor manufacturers and large language model developers fell an average of approximately 7% in the one-month period ending May 24 local time, driven by investor concerns over whether massive investment spending can generate sufficient returns. Despite the recent stock decline, Morgan Stanley assesses that AI industry fundamentals remain robust, with Stephen Byrd, Global Head of Thematic & Sustainability Research at Morgan Stanley, stating that AI infrastructure will become an 'intelligence highway' providing significant benefits to the global economy over time. The firm's current AI optimism contrasts sharply with its historical stance on semiconductors, having issued bearish reports in 2017, 2021 (titled 'Memory, Winter is Coming'), and 2024 that warned of industry downturns and recommended reducing exposure to memory chip makers including Samsung Electronics, SK Hynix, and Micron.
Morgan Stanley Calls AI Stock Decline a Temporary Speed Bump
Morgan Stanley attributes the recent AI-related stock price decline to investor skepticism about whether massive investment expenditures can generate sufficient returns. Stephen Byrd, Global Head of Thematic & Sustainability Research at Morgan Stanley, stated, "AI infrastructure will become an 'intelligence highway' that provides significant benefits to the global economy over time. We are optimistic about this intelligence highway, but there will be several speed bumps ahead."
Token Usage Per Employee Remains Low With Growth Potential
Morgan Stanley dismissed concerns that companies might limit AI token usage per employee to reduce costs as excessive worry. Byrd stated, "Looking at the data, average token consumption per employee is actually at quite a low level. There is room for token usage to increase substantially over the coming months and years."
China Open-Weight Models May Increase Compute Demand via Jevons Paradox
Morgan Stanley analyzed that the spread of open-weight AI models from China, while potentially posing competitive threats, could actually expand AI compute demand. Open-weight models publicly release AI knowledge parameters called 'weights,' allowing users to directly modify and operate them, achieving high performance with significantly lower costs and compute resources. Morgan Stanley assessed that open-weight model proliferation will increase AI efficiency and lower compute costs, enabling more companies and users to adopt AI. The firm explained this could result in 'Jevons' Paradox,' where overall compute demand increases. Byrd stated, "The movement by Chinese and US LLM developers to pursue efficiency reinforces our baseline forecast that AI compute demand will significantly exceed supply."
US Big 5 Tech CAPEX Projected to Reach $1.4 Trillion by 2028
Morgan Stanley projected that capital expenditures by the US Big 5 technology companies will increase from approximately $800 billion this year to $1.2 trillion in 2027 and $1.4 trillion in 2028.
![US Hyperscaler Investment Projections]()
US Hyperscaler Investment Projections [Source: Morgan Stanley]
Morgan Stanley's Track Record on Semiconductor Pessimism
Morgan Stanley has been called the 'semiconductor grim reaper' in the Korean market for repeatedly issuing pessimistic forecasts on Korean semiconductors. The firm warned of potential industry slowdowns in semiconductor-related reports in 2017, 2021, and 2024. In 2021, Morgan Stanley caused significant market impact with a report titled 'Memory, Winter is Coming.' Recently, the firm recommended reducing exposure to memory semiconductor-related stocks including Samsung Electronics, SK Hynix, and Micron.
FAQ
Why did AI-related stocks fall approximately 7% in the period ending May 24 local time?
Morgan Stanley Research reported that AI infrastructure-related stocks including semiconductor manufacturers and large language model developers fell an average of approximately 7% in the one-month period ending May 24 local time due to investor concerns over whether massive investment spending can generate sufficient returns despite strong AI industry fundamentals.
What is Morgan Stanley's projection for US Big 5 tech company capital expenditures through 2028?
Morgan Stanley projected that capital expenditures by the US Big 5 technology companies will increase from approximately $800 billion this year to $1.2 trillion in 2027 and $1.4 trillion in 2028, reflecting continued investment in AI infrastructure.
How does Morgan Stanley view China's open-weight AI models affecting compute demand?
Morgan Stanley assessed that the spread of open-weight AI models from China could paradoxically increase overall AI compute demand through 'Jevons' Paradox'—as AI efficiency increases and compute costs decrease, more companies and users will adopt AI, resulting in higher total demand that reinforces the firm's forecast that AI compute demand will significantly exceed supply.
