This week, the wolf of "AI disrupting everything" has finally arrived.
The market has finally realized that AI disruption is no longer a distant threat.
On February 14, according to Wind Chaser Trading Desk, Morgan Stanley stated in its latest research report that as AI models advance in a nonlinear and accelerated manner, the market’s pricing of disruption risk is beginning to show a domino effect:
Just a month ago, the market believed about 4% of the MSCI Europe index weight faced AI disruption risk; a week ago this ratio rose to 7%; and as of February 13, this number has jumped to 24% (including the banking sector).
The report points out that Morgan Stanley sees a threshold being crossed with the latest AI models—GPT-5.2 has reached or surpassed human expert levels on 71% of professional tasks—requiring investors to reexamine their asset allocation logic.
Morgan Stanley has shifted its stance from neutral to cautious regarding cyclical stocks relative to defensive stocks, noting the European credit market offers cheap downside hedging opportunities, with a focus on utilities, semiconductors, defense, and tobacco as the most resilient safe havens.
The firm underscores the need to reassess which assets cannot be "replicated" by AI—these will become value anchors in the new era. In an age when intelligence and labor can be infinitely replicated, true value will return to things that cannot be duplicated—physical assets, regulatory barriers, network effects, human experiences, proprietary data.
Astonishing Leap in AI Capabilities: 71% of Professional Tasks Conquered
Humans are not good at understanding nonlinear change, while the progress in AI models is precisely a typical nonlinear acceleration.
Morgan Stanley notes the data show stunning speed of advancement: Grok 4, launched in July 2025, scored 24% on the GDPVal test, meaning the model reaches human expert level in 24% of real professional tasks; but in just five months, by December 12, 2025, GPT-5.2's score has soared to 71%.
What is GDPVal? It’s an indicator measuring how AI models perform in real-world knowledge work, covering actual tasks handled by experienced professionals in various industries. OpenAI’s research shows that cutting-edge models complete these tasks about 100 times faster, at roughly 1/100th the cost of industry experts.
The report stresses that the coming breakthroughs are even more astounding. If the scale law for Large Language Models (LLMs) holds in 2026—as Morgan Stanley believes likely—several cutting-edge US LLMs are expected to launch in H1 2026 with capabilities far exceeding current models. The reason is simple: the five major US LLM developers are training their next-generation models with about 10 times the compute power of current models.
Domino Effect of Disruption Risk: From Software to Banks
The speed of market recognition change is equally astonishing.
Morgan Stanley’s tracking shows the market at first began to question software industry revenue growth over the coming years, but soon this concern spread like dominoes to broader economic disruption risks—competition landscape changes, job impacts, deflationary pressures, and more.
This recalls the evolution of market psychology at the start of COVID-19 in early 2020: In January, concerns were limited to demand and supply chain risks; in February, they spread to travel and leisure, industry, banks, etc.; by March, it turned into a broad market sell-off, ultimately triggering major policy action.
Currently, Morgan Stanley estimates about 10% of the MSCI Europe index weight (excluding banks) is seen by the market as facing substantive AI disruption concerns, rising to 24% if banks are included. Concerns about the banking sector are relatively new, mainly centered on broader economic deflation and employment issues, and (to a lesser extent) AI-related deposit competition.
Notably, these "disruption stocks debated in the market" have fallen from a peak 2025 P/E ratio of 24 times to today’s 16.4 times. But Morgan Stanley warns that referencing "undisputed disruption stocks" (which have dropped from 24.7 times to 11.1 times), valuations may have further downside room.
Who Can Survive the Age of AI?
In the face of this disruption storm, Morgan Stanley offers an assessment framework using five dimensions to judge sector and stock resilience:
AI exposure: Is it a disrupted party, "disruption subject debated by the market," enabler, or protected party?
Business type: Does it provide services, physical assets, commodities, or computing power?
Cyclicality: Cyclical stock, defensive stock, or other?
Investor positioning: Current positions level?
Stock momentum: Fundamental overlay factors
Based on this framework, Morgan Stanley finds the most resilient sectors are, in order: utilities, semiconductors, defense, tobacco, and personal/home care products.
Morgan Stanley says European utility companies occupy almost all the top 20 spots in anti-disruption rankings. Their common features: providing physical infrastructure that AI cannot replicate, belonging to defensive industries, and currently being relatively underweighted.
Conversely, software, business services, media and entertainment, travel and leisure, and other service-intensive industries, as well as transportation, diversified finance, and banks, are considered to face the greatest pressure from the spread of disruption risk.
Eight Major Asset Types That Cannot Be Replicated by AI
Meanwhile, Morgan Stanley stresses that once AI reaches transformative levels, asset types that cannot be "replicated" by AI will rise in value. This is the key framework for understanding future asset allocation:
A. Physical scarcity: Real estate, energy and power assets, transportation infrastructure, data centers, mined metals, water resources, casino licenses in limited jurisdictions, theme park land, cruise ports and dock rights, spectrum licenses, fiber cable networks, etc.
B. AI adopters with pricing power: Ability to prove pricing power is increasingly important.
C. Unique luxury goods, real estate, and services.
D. Network effects: Large tech platforms, online marketplaces, healthcare businesses with patient relationships.
E. Genuine, unique human experiences: Media businesses with strong brands, sports assets/teams, music, and other performances that value human elements.
F. Regulatory scarcity: Businesses holding various licenses, approvals, and protected franchises.
G. Proprietary data and brands: AI adopters owning proprietary datasets and IP libraries.
H. Series of semiconductor assets: Such as leading-edge process, ASML’s EUV lithography, TSMC’s manufacturing expertise, rare earth chip processing.
Credit Market: Cheap Downside Protection
Though AI disruption concerns have begun to affect parts of the credit market, especially leveraged loans, European investment-grade spreads still hover near post-global financial crisis lows. Even as equity implied volatility keeps rising, credit volatility remains abnormally subdued.
However, if AI disruption concerns spill into more sectors (plus expected acceleration of issuance), it may start to challenge the resilience of the credit market.
Morgan Stanley believes the credit options market gives investors excellent entry points to prepare for widening spreads. With Europe’s relatively low tech exposure, overall yield still high, policy support, and resilient growth, these hedging tools’ cost-effectiveness is especially remarkable.
Compute Demand Gap: An Invisible Supply Crisis
On the other side of AI disruption is a frenzy for compute infrastructure demand. Multiple data points show compute demand is growing far faster than current supply forecasts:
- Google executives recently stated the company may need to double compute every six months, "reaching 1000 times in 4-5 years." In comparison, Morgan Stanley forecasts Nvidia compute sales CAGR of around 210% for 2025-2028; over five years, cumulative compute is about 300 times—far below Google’s stated need for 1000 times plus.
- OpenRouter data shows from late November 2024 to late November 2025, weekly average token demand grew by over 2200%. Token use is a direct proxy for compute demand.
- More critically, the compute intensity of a single LLM query is rising rapidly. Research institute METR points out that the average "work" AI performs for each customer query doubles every seven months.
According to the report, even if customer count stays the same, this growth means compute demand increases will far exceed Nvidia’s roughly 120% CAGR prediction.
Morgan Stanley notes this supply-demand mismatch is already showing up in the market:
CoreWeave can renew leases for older generation Nvidia GPUs (Hopper) at 95% of original price, far above price levels implied by economic use depreciation over time for chips;
Google's "power shell" rental deals for Anthropic and FluidStack led to about 18.5% unlevered ROIC for Bitcoin miner Hut8, equivalent to about a 300% premium for power access.
~~~~~~~~~~~~~~~~~~~~~~~~
The above highlights are from Wind Chaser Trading Desk.
For more detailed interpretation, including real-time decoding and frontline research, please join the [Wind Chaser Trading Desk · Annual Membership]
Risk Warning and DisclaimerThe market involves risks and investment should be made cautiously. This article does not constitute personal investment advice, nor does it consider the specific investment objectives, financial status, or needs of individual users. Users should evaluate whether any opinions, views, or conclusions in this article are appropriate to their own circumstances. All investments made based on this information are at their own risk.