Morgan Stanley believes the artificial intelligence sector is making a decisive leap from the infrastructure buildout phase into the "Age of Inference." This technological evolution is set to digitize and transform the global knowledge work market, valued at between 20 and 30 trillion US dollars, alongside a consumer market approaching 30 trillion US dollars, thereby fueling a shift in capital markets on an unprecedented scale.

A recent research report led by Brian Nowak highlights that as hyperscale cloud providers' data center capital expenditure growth is projected to peak in 2027 and decelerate markedly by 2028, the flow of funds within the technology cycle is reaching a critical inflection point.

Investor capital is expected to gradually rotate away from the semiconductor and hardware layer, shifting decisively toward the software layer and AI enablers. Tech titans such as Amazon, Google, Microsoft, and Meta, distinguished by their complete ecosystems and robust free cash flow, are poised for a fresh wave of valuation expansion.

The report underscores that AI adoption across both enterprise and consumer segments is accelerating at a rapid pace. By the second quarter of 2026, roughly 25% of companies in the S&P 500 had begun quantifying the financial benefits derived from generative AI.

Meanwhile, the credit financing market for AI infrastructure demonstrates exceptional absorption capacity. Global AI-related debt issuance has reached approximately 450 billion US dollars this year, with the strong operating cash flows of hyperscale cloud providers fully capable of covering future debt obligations, alleviating concerns over financing bottlenecks.

During this transitional phase, market focus will pivot decisively toward AI return on investment, the competitive dynamics of open-source models, and the increasingly prominent constraints posed by power availability and regulatory oversight. Striking a balance between surging computational demand and physical limitations will ultimately determine the trajectory of trillions in inference-related spending in the next phase.

Capital Flows at an Inflection Point: Hardware Cools While Software and Enablers Rise

Morgan Stanley identifies the capital expenditure trajectory of hyperscale cloud providers as the primary indicator shaping market rotations. Capital spending on data centers is expected to surge 60% year-over-year by 2027, reaching a total of 1.5 trillion US dollars. However, due to physical constraints involving labor, materials, and electricity, combined with the early deployment of some capacity slated for 2027 through 2029, growth in capital expenditure during 2028 is anticipated to moderate sharply to approximately 12%.

This deceleration in growth, coupled with accelerating revenue within the software layer, signals the onset of a typical multi-year technology cycle rotation. This pattern aligns closely with the dynamics observed during the mobile internet era: after the initial explosion in hardware and semiconductors, value migrates toward applications and software services.

While hardware-layer stocks are unlikely to experience cliff-edge declines given their reasonable valuations, software and enabler companies are positioned to capture greater multiple expansion opportunities as earnings upside shifts up the stack. As capital expenditures materialize, global computational capacity is set to grow exponentially. By 2028, total compute capacity is projected to climb from 35 GW in 2025 to approximately 145 GW.

Within this expansion, custom ASIC chips are expected to increase their share of new compute capacity from 34% in 2025 to 66% by 2028, with Google's TPU and Amazon's Trainium leading this structural transformation.

Penetrating a 60-Trillion-Dollar Market: Technology and Financial Sectors Take the Lead

Transitioning into the inference age, the true test for generative AI lies in successful commercialization. The market faces a digitalization opportunity valued at 20 to 30 trillion US dollars within global knowledge work, alongside a consumer market encompassing retail, travel, autonomous driving, food delivery, and advertising that holds an estimated 30 trillion US dollars in untapped potential.

Adoption rates within the application layer are quickening. From a macroeconomic diffusion standpoint, the technology sector leads in implementing and quantifying the benefits of generative AI, followed closely by financial services, healthcare, and industrial sectors. Mentions of AI-related benefits in technology sector earnings calls have surged from 28% a year ago to 51% today.

Historical parallels suggest that current AI penetration rates may be significantly underestimated. In the second year of cloud computing's development (2014), public cloud spending accounted for 4% of total IT budgets. Applying the same adoption pace, enterprise AI spending could reach approximately 800 billion US dollars by 2027.

Given that AI does not require wholesale infrastructure migration and offers a shorter time-to-value, its actual diffusion speed is expected to outpace cloud computing. On the consumer front, platforms possessing vast first-party data and distribution channels will secure overwhelming competitive advantages.

Compute Returns and Open-Source Models: Reshaping Business Models

Addressing the market's intense focus on capital returns, analysis indicates that generative AI offers exceptionally attractive return on invested capital prospects, with ROIC expectations ranging between 25% and 50% across multiple monetization pathways. The highest returns are generated by companies operating model APIs on proprietary infrastructure, such as Meta, Google, and SpaceX, achieving returns around 46%. Even pure-play hyperscale GPU leasing operations can sustain a ROIC of roughly 30%.

At the model level, open-source models with lower serving costs will not diminish compute demand. Instead, they will serve as a critical catalyst for AI adoption across the entire economy. Open-source models will drive down average token prices, triggering the Jevons Paradox, where falling prices lead to an explosive surge in inference demand.

This competitive dynamic will compel frontier AI laboratories to continually innovate in pursuit of inference spending share, while simultaneously reinforcing the core value of hyperscale cloud providers' "model routing layers," such as AWS Bedrock, Microsoft Foundry, and Google Vertex AI. These cloud platforms maximize both computational efficiency and monetization potential by matching the most economical models to specific tasks.

Overcoming Infrastructure Bottlenecks: Credit Expansion and Power Breakthroughs

Underpinning this expansive inference market is uninterrupted financing and infrastructure development. Although AI-related debt issuance has surged to roughly 450 billion US dollars this year, high-quality hyperscale cloud providers still possess substantial headroom for additional debt issuance. Market adjustment mechanisms will primarily manifest through widened credit spreads.

More importantly, operating cash flows at Amazon, Google, Meta, and Microsoft are accelerating. These four giants are expected to generate operating cash flows of 980 billion and 1.2 trillion US dollars by 2027 and 2028, respectively, a scale seven to eight times the 290 billion US dollars in debt they would need to raise over the same period, demonstrating remarkable balance sheet resilience.

However, physical bottlenecks remain formidable. Political scrutiny at state and local government levels is intensifying over data centers' electricity consumption impacts and water resource usage. To counter grid connection delays, hyperscale data centers will increasingly adopt Behind-the-Meter on-site power generation solutions, which is expected to add approximately 3 billion US dollars in capital expenditure per GW.

This pressing need to trade time for power makes on-site power generation companies that directly benefit from this trend, along with providers of power shell assets, attractive long-term investment opportunities.