30 trillion RMB in knowledge work, 30 trillion RMB in consumer market -- Navigating the era of AI reasoning

Morgan Stanley believes that the artificial intelligence industry is moving from the infrastructure construction stage into the "inference era." This technological evolution will digitize and reshape the global knowledge work market of $20 to $30 trillion and the consumer market of nearly $30 trillion, thereby driving an unprecedented sector rotation in the capital market.

According to a report by Zhuifeng Trading Platform on September 7, a research report titled "The Morgan Stanley AI Guidebook: Navigating the Age of Inference," led by Morgan Stanley's Brian Nowak , points out that the flow of funds in the technology cycle is reaching a critical turning point, as the growth rate of data center capital expenditures by hyperscale cloud service providers is expected to peak in 2027 and slow significantly in 2028.

(Capital expenditures by hyperscale cloud computing vendors are projected to reach approximately $1.5 trillion and $1.6 trillion by 2027 and 2028, respectively.)

Investor funds are expected to flow out of the semiconductor and hardware sectors and shift in large quantities to the software and AI enablers. Tech giants with complete ecosystems and free cash flow, such as Amazon, Google, Microsoft, and META, will see a new round of valuation expansion.

(The AI business cycle is expected to expand to the software and services sector and companies applying artificial intelligence.)

The report emphasizes that AI applications in both enterprise and consumer sectors are experiencing rapid growth . As of the second quarter of 2026, approximately 25% of companies in the S&P 500 have begun to quantify the financial benefits of generative AI.

At the same time, the AI infrastructure credit financing market has shown strong resilience. Since the beginning of this year, global AI-related debt issuance has reached approximately $450 billion. The robust operating cash flow of hyperscale cloud service providers is fully capable of covering future debt needs, eliminating market concerns about financing bottlenecks.

During this transition period, market focus will shift entirely to AI ROI, the competitive landscape of open-source models, and the increasingly prominent power and regulatory bottlenecks . Finding a balance between the surge in computing power and physical limitations will determine the ultimate destination of trillions of dollars in inference spending in the next phase.

Capital flows reach a turning point: hardware cooling down, software and enablers rising.

Morgan Stanley believes that the capital expenditure trajectory of hyperscale cloud service providers is a key indicator that determines market capital rotation.

Data center capital expenditures are projected to grow by 60% year-on-year, reaching a total of $1.5 trillion by 2027. However, due to physical constraints such as labor, materials, and electricity, as well as the upfront deployment of some capacity from 2027 to 2029, the growth rate of capital expenditures is expected to slow significantly to about 12% in 2028.

This slowdown in growth, coupled with the accelerated growth in software revenue, signals the imminent arrival of a typical multi-year technology cycle of capital rotation. This aligns closely with the patterns observed in the mobile internet era: after the initial boom in hardware and semiconductors, value will shift to applications and software services.

(Relative stock performance during the mobile internet era)

While hardware stocks are unlikely to experience a precipitous drop due to their relatively low valuations, software and enablers will gain greater opportunities for expansion as better-than-expected earnings move up the value chain .

With the implementation of capital expenditures, the total global computing power capacity will experience exponential growth. It is projected that by 2028, the total computing power capacity will surge from 35 GW in 2025 to approximately 145 GW.

(Morgan Stanley's projected path to reach approximately 145GW of computing power by 2028)

Custom ASIC chips will account for 34% of new computing power in 2025, up to 66% in 2028, with Google’s TPU and Amazon’s Trainium leading this structural shift .

(In the coming years, the share of incremental production capacity for ASICs will continue to grow.)

Penetrating a 60 trillion yuan market: Technology and finance industries lead the way

With the advent of the reasoning era, the real challenge facing generative AI is its commercialization.

The market faces a global opportunity for the digitalization of knowledge work worth $20 to $30 trillion, while the consumer market, including retail, travel, autonomous driving, food delivery and advertising, also holds an untapped potential of about $30 trillion.

(A research report estimates that approximately $30 trillion in consumer spending needs further digitization.)

The adoption rate at the application layer is accelerating. Looking at the diffusion across the macroeconomy, the technology sector is leading the way in adopting and quantifying the benefits of generative AI, followed closely by the financial, healthcare, and industrial sectors.

(Application cases cover the entire economic field)

The percentage of tech companies mentioning AI revenue in earnings calls has risen sharply from 28% a year ago to 51% today.

Historical comparisons suggest that current AI penetration rates may be significantly underestimated. In 2014, the second year of cloud computing development, public cloud spending accounted for 4% of total IT budgets. Extrapolating at the same rate of adoption, enterprise AI spending is projected to reach approximately $800 billion by 2027.

(From a corporate perspective, it can be assumed that public cloud spending accounted for 4% of the IT budget in the second year, 2014.)

Because AI does not require wholesale infrastructure migration and offers shorter value conversion times, its actual diffusion rate is expected to be faster than that of cloud computing . On the consumer side, platforms with vast amounts of first-party data and distribution channels will hold a dominant position.

(The development of generative AI should be faster. Even if its penetration rate is only 4% in the second year, 2027, it means that the scale of spending on generative AI in the enterprise sector will reach approximately $800 billion.)

Return on Computing Power and Open Source Models: Reshaping Business Models

Regarding the highly anticipated issue of return on investment, analysis shows that the return on investment for generative AI is extremely attractive, with expected ROIC for multiple monetization paths ranging from 25% to 50%.

Among them, the highest returns are achieved by companies that run model APIs on their own infrastructure (such as META, Google, and SpaceX), with a return on investment of approximately 46%. Even pure hyperscale GPU leasing (IaaS) businesses can maintain an ROIC of around 30%.

(25%-50% return on invested capital for three generative AI frameworks)

At the model level, open-source models with lower service costs will not only not reduce the demand for computing power, but will instead become the key to promoting the popularization of AI across the entire economy .

The open-source model will depress the average token price, thus triggering the Jevons Paradox, which states that a drop in price leads to an explosive increase in inference demand .

This competitive dynamic will force cutting-edge AI labs to innovate continuously to compete for inference spending share, while further solidifying the core value of hyperscale cloud service providers’ “model scheduling layer” (such as AWS Bedrock, MSFT Foundry, GOOGL Vertex).

These cloud platforms maximize computational efficiency and monetization capabilities by matching the most economical model to different tasks.

Overcoming Infrastructure Bottlenecks: Credit Expansion and the Power Breakthrough

The cornerstone supporting this massive inference market is continuous funding and infrastructure expansion.

Despite a surge in AI-related debt issuance to approximately $450 billion this year, high-quality hyperscale cloud service providers still have significant room for debt issuance. Market adjustments will primarily manifest in the widening of credit spreads.

More importantly, the operating cash flow of Amazon, Google, Meta, and Microsoft is accelerating. It is projected that by 2027 and 2028, the operating cash flow of these four giants will reach $980 billion and $1.2 trillion respectively, a scale seven to eight times the $290 billion in debt they need to raise during the same period, demonstrating extremely strong balance sheet resilience.

However, physical bottlenecks remain significant. Political scrutiny of data centers by state and local governments regarding their impact on electricity costs and water consumption is intensifying.

To address grid connection delays, hyperscale data centers will increasingly adopt behind-the-meter on-site power generation solutions, which is expected to increase capital expenditure by approximately $3 billion per GW.

The rigid demand for time-for-electricity makes on-site power generation companies that directly benefit from this, as well as providers of power supply shell assets, attractive long-term investments.

(Power supply enclosures and racks remain the biggest drivers of computing power investment.)

~~~~~~~~~~~~~~~~~~~~~~~

The above content is from Zhuifeng Trading Platform .

For more detailed analysis, including real-time updates and firsthand research, please join the [ Trading Channel Annual Membership ].

Risk Warning and DisclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.