--- title: "How to View the Major AI Correction: Morgan Stanley's 120-Page In-Depth Analysis" type: "News" locale: "en" url: "https://longbridge.com/en/news/294593827.md" description: "Morgan Stanley released a 120-page research report interpreting the correction in the AI sector, arguing that it is primarily driven by technical factors rather than a deterioration in basic factors. Addressing market concerns about the reversal of the \"Tokenmaxxing\" narrative, Morgan Stanley pointed out that the baseline for corporate AI usage is low, ROI exceeds 10x, and profit margins will improve with GPU generational evolution, leaving significant room for token price reductions and sustained demand release" datetime: "2026-08-02T03:06:52.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/294593827.md) - [en](https://longbridge.com/en/news/294593827.md) - [zh-HK](https://longbridge.com/zh-HK/news/294593827.md) --- # How to View the Major AI Correction: Morgan Stanley's 120-Page In-Depth Analysis Since late June, the global AI sector has experienced a significant correction. Is this merely a brief technical correction, or a signal that the cycle has peaked? This is a question worth its weight in gold. On July 27, Morgan Stanley published a comprehensive 120-page report titled "Playing the AI Infrastructure Dip" to explore this issue. The report offers a clear judgment: the primary drivers of the correction are technical factors, including the digestion of crowded positions, deleveraging of margin financing, and the reversal of momentum factors, rather than a breakdown in fundamental logic. I carefully read this report and gained valuable insights. Here is an interpretation of its main content. **(1) Capital Market Concern #1: Reversal of the Tokenmaxxing Narrative?** The first negative signal recently watched by the market is the reversal of the "Tokenmaxxing" narrative. Many companies have begun to impose budget caps on employees' AI token usage, sparking doubts about the sustainability of revenue growth for large model companies. This concern is unfounded. First, the current baseline for employee token spending is extremely low, while the ROI is exceptionally high. Morgan Stanley's research into numerous enterprise-level AI application scenarios shows that a single AI call saves approximately $55 in labor costs on average, while the average cost to complete an enterprise-level task through Agent collaboration is only $2–5. This results in an ROI of over 10x. **For tools with an ROI multiple exceeding 10x, adoption is not a matter of budget decisions but of core competitiveness.** Companies that do not actively deploy AI capabilities will face increasingly significant competitive disadvantages. Second, GPU generational evolution will improve data center profit margins, meaning that token prices can be significantly reduced in the future without harming profitability, thereby further releasing demand. Calculations from Morgan Stanley's Intelligence Factory model show that the net profit margin for token sales in Blackwell-based data centers is approximately 58%. When Rubin and Feynman generation GPUs are deployed, profit margins will rise to approximately 80% and 90%, respectively. **This means Hyperscalers can lower token pricing by approximately 75% while maintaining profit margins.** There is sufficient room for the cost curve to shift downward, allowing for both price reductions and profitability. **(2) Capital Market Concern #2: Does the Kimi Moment Falsify the Rationality of Capex?** The release of Kimi K3 has led capital markets to re-examine a core assumption: If China can train frontier models with comparable performance at a lower cost, does the annual AI Capex of over $1 trillion by US Hyperscalers face a downward revision in returns? Morgan Stanley believes that the extreme pursuit of efficiency by large model companies in China and the US will not weaken computing power demand. On the contrary, it reinforces the structural judgment that "demand far exceeds supply." In the 19th century, economist William Stanley Jevons observed that Watt's improvement of the steam engine significantly increased coal combustion efficiency, yet total coal consumption in the UK soared. This was because the steam engine became economically viable and was deployed in far more factories and mines than before. **Thus, he proposed the famous Jevons Paradox: When the efficiency of using a resource improves, the total consumption of that resource increases rather than decreases.** The Jevons Paradox also applies to the current wave of the AI revolution: Improved computing efficiency lowers the unit cost of tokens, and lower costs mean more scenarios, more users, and higher-frequency calls, ultimately driving up total computing consumption. Morgan Stanley cited a set of data to quantify the severity of this supply-demand imbalance: Google executives recently stated that the company may need to double its computing power every six months, achieving a 1,000-fold increase within five years. However, from the supply side, NVIDIA's AI chip sales CAGR for 2025–2028 is approximately 140%. Even extrapolating at this rate for five years, the cumulative delivered computing power would account for less than 10% of Google's single-company demand forecast. In other words, even if the world's largest computing power supplier produces at its highest historical growth rate, it can only cover a small fraction of a single client's needs. **(3) Capital Market Concern #3: Do Supply-Side Constraints Constitute a Hard Ceiling?** The third concern in capital markets is whether physical world constraints will prevent computing infrastructure from being delivered on demand, even if demand-side certainty is sufficient. **Morgan Stanley categorizes physical world constraints as the "3Ps": People, Power, and Politics.** People: Skilled trades required for data center construction (electricians, welders, plumbers) are in a state of structural shortage. Power: In some regions, the queue time for grid interconnection has extended to 5–7 years, becoming the largest single time bottleneck for data center commissioning. Politics: Data center construction is facing multi-level political resistance from local to federal levels, and the trend is undergoing a structural reversal. In recent years, states competed to offer generous incentives to attract data centers. However, the trend has reversed, with states beginning to pause, add conditions to, or directly revoke tax benefits for data centers. **Issues such as slowing the pace of data center development and protecting residents' electricity bills from the impact of data center infrastructure costs are increasingly becoming part of gubernatorial campaign platforms,** expected to become important voter issues in the November elections. Meanwhile, at the federal level, there are attempts to establish a national "data center tariff." The House of Representatives is deliberating the **Ratepayer Protection Act**, which is the first federal attempt to legislate the allocation of infrastructure construction costs. It requires state utility companies to consider creating "large load standards," making data centers pay for grid upgrades. Previously (in March), Amazon, Google, Meta, Microsoft, Oracle, xAI, and others signed the White House's Ratepayer Protection Pledge, voluntarily committing to protect existing consumers from the impact of data center infrastructure costs. **This bill will legalize this voluntary commitment, effectively forming a national surcharge system for data center electricity.** Morgan Stanley acknowledges the validity of this concern but characterizes it as "speed bumps" rather than structural barriers. As grid interconnection queues exceed five years in some regions and data centers face increasing pressure to "self-power," **on-site power generation is becoming a core solution.** **(4) Time to Power: The Undervalued Electricity Time Arbitrage** Morgan Stanley conducted a quantitative assessment of the electricity gap for US data centers. The conclusion is that US data center electricity demand from 2026 to 2028 will be approximately 68GW. After deducting facilities under construction (15GW) and contracted grid capacity (15GW), the potential gap reaches 38GW, with grid interconnection queues in some regions already reaching 5–7 years. In this context, Morgan Stanley believes that the time value of electricity access (Time to Power) is the area with the largest pricing deviation in the current market. The underlying logic is very clear: Data center commissioning is limited by power supply → Grid interconnection queues take 5–7 years → Alternative solutions capable of providing power within 1–3 years possess significant time arbitrage value. Morgan Stanley identified two core "de-bottlenecking" paths: - Bitcoin mining farms: These companies already possess substantial grid interconnection capacity and physical land, which can be directly converted for data center use, contributing 10–19GW. - Power generation solutions supporting rapid deployment: Offering a 1–3 year time advantage relative to grid interconnection, where gas turbines can contribute 15–20GW and fuel cells can contribute 5–8GW. Even incorporating "Time-to-Power" solutions such as gas turbines, fuel cells, direct supply from nuclear power plants, and Bitcoin farm conversions into probability-weighted calculations, there is still a net gap of approximately 1GW in the base case, expanding to 11GW in the pessimistic case. **In other words, the scarcest resource for Hyperscalers currently is not Capex, but physical space with available power, i.e., "Powered Shell."** Morgan Stanley's research suggests that the market has not fully priced in these Powered Shell Provider targets. To quantify this undervaluation, Morgan Stanley used traditional renewable energy PPAs as a benchmark for comparison, as detailed in the table below: Currently, the EV/Watt for these Powered Shell Providers is only in the $2–4 range, including companies such as TeraWulf, Cipher Mining, HUT 8, Riot Platforms, Applied Digital, and Galaxy Digital. Based on the reference of mature data center operators (such as Equinix and Digital Realty) at 20–25x EV/Watt, Morgan Stanley assigns a discounted target valuation of 15x EV/Watt to this group of transitioning companies. Risk Warning and Disclaimer The market carries risks; investment requires caution. This article does not constitute personal investment advice, nor does it consider the specific investment goals, financial status, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investment based on this content is at your own risk. ### Related Stocks - [MS.US](https://longbridge.com/en/quote/MS.US.md) - [GOOGL.US](https://longbridge.com/en/quote/GOOGL.US.md) - [GOOG.US](https://longbridge.com/en/quote/GOOG.US.md) - [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md) - [AMZN.US](https://longbridge.com/en/quote/AMZN.US.md) - [META.US](https://longbridge.com/en/quote/META.US.md) - [MSFT.US](https://longbridge.com/en/quote/MSFT.US.md) - [ORCL.US](https://longbridge.com/en/quote/ORCL.US.md) - [WULF.US](https://longbridge.com/en/quote/WULF.US.md) - [CIFR.US](https://longbridge.com/en/quote/CIFR.US.md) - [HUT.US](https://longbridge.com/en/quote/HUT.US.md) - [RIOT.US](https://longbridge.com/en/quote/RIOT.US.md) - [APLD.US](https://longbridge.com/en/quote/APLD.US.md) - [GLXY.US](https://longbridge.com/en/quote/GLXY.US.md) - [EQIX.US](https://longbridge.com/en/quote/EQIX.US.md) - [DLR.US](https://longbridge.com/en/quote/DLR.US.md) - [MS-O.US](https://longbridge.com/en/quote/MS-O.US.md) - [MS-Q.US](https://longbridge.com/en/quote/MS-Q.US.md) - [MS-E.US](https://longbridge.com/en/quote/MS-E.US.md) - [MS-I.US](https://longbridge.com/en/quote/MS-I.US.md) - [MS-L.US](https://longbridge.com/en/quote/MS-L.US.md) - [MS-P.US](https://longbridge.com/en/quote/MS-P.US.md) - [MS-A.US](https://longbridge.com/en/quote/MS-A.US.md) - [MS-F.US](https://longbridge.com/en/quote/MS-F.US.md) - [MS-K.US](https://longbridge.com/en/quote/MS-K.US.md) - [NVD.DE](https://longbridge.com/en/quote/NVD.DE.md) - [ORCL-D.US](https://longbridge.com/en/quote/ORCL-D.US.md) - [BRPHF.US](https://longbridge.com/en/quote/BRPHF.US.md) ## Related News & Research - [How the 'Nostradamus of AI' failed to spot a market crash](https://longbridge.com/en/news/294580418.md) - [Meet the Gen Z AI whiz at the center of a hedge fund meltdown](https://longbridge.com/en/news/294425537.md) - [Here are my 3 top artificial intelligence (AI) stocks to buy in August](https://longbridge.com/en/news/294601562.md) - [17:23 ETArmor Launches Sovereign AI: A whole-company AI work platform for regulated industries](https://longbridge.com/en/news/294418838.md) - [Official FO: AI Quant Agents go live with passive mode - Aug 2026](https://longbridge.com/en/news/294567741.md)