AI Trends

Fitch Says AI Revenue Growth Needs to Keep Pace With Investment Expansion, While Commercial Returns Remain to Be Proven

EastmoneyUpdated Nov 18 23:02

AI summary

Fitch noted that the pace of AI commercialization and monetization may lag behind capital investment, with related credit pressure potentially spreading from frontier model developers to computing service providers and the hardware supply chain. Global annualized AI revenue currently stands at about $230 billion and is expected to reach $1.05 trillion by 2030, which could support approximately $500 billion in sustainable capital expenditure. However, the five largest hyperscalers' combined capital expenditure is projected to exceed $1 trillion in 2027. AI-related bond issuance is also rising, with hyperscale technology companies issuing about $277 billion in bonds from the start of 2026 through November 11.

Why read this

AI score 71

Fitch juxtaposes AI revenue growth with the scale of capital expenditure, providing a quantitative benchmark for assessing the commercialization gap and the pathways through which credit risk may spread.

Topics

Tags

Body · AI translation

Article Body

International ratings agency Fitch said in a report released on November 17 that AI-related bond issuance is reshaping the global investment-grade corporate bond market. Bond issuance by hyperscale technology companies has risen from less than $50 billion annually between 2020 and 2024 to approximately $277 billion from the beginning of 2026 through November 11.

The report also projected that the combined capital expenditure of the five hyperscalers—Amazon, Alphabet, Meta, Oracle and Microsoft—will exceed $1 trillion in 2027. AI-related industries continue to account for a growing share of the investment-grade bond market, and increasing sector concentration could exacerbate market volatility.

Investment has surged, but can revenue keep pace?

“Revenue needs to grow a little faster to catch up with investment,” said Jeremy Hu, senior director of Asia-Pacific Credit Commentary and Research at Fitch Ratings. Speaking at the “Fitch on China 2026” media briefing on November 18, Hu said that assessing the outlook for AI investment requires looking beyond the scale of capital spending to whether computing capacity can be delivered on schedule, whether infrastructure can meet demand and whether the investment can ultimately generate corresponding returns.

Cindy Huang, senior director of Asia-Pacific Corporate Ratings at Fitch Ratings, said the main credit risk currently facing the AI value chain is that commercialization and monetization may proceed more slowly than capital investment. If revenue growth cannot keep pace with continuously expanding capital expenditure, market confidence in returns on AI investment could be challenged. The resulting pressure could spread from frontier model developers to computing service providers and the hardware supply chain.

“Current annualized global AI revenue is about $230 billion, and we expect it to reach $1.05 trillion by 2030, more than four times the current level,” Hu said. However, he noted that Fitch estimates revenue of $1.05 trillion could support approximately $500 billion in sustainable capital expenditure, while the market expects capital expenditure next year to exceed $1 trillion. “So there is a gap, and revenue needs to grow a little faster to catch up with investment,” he said.

Discussing AI commercialization, Hu said an important factor in realizing commercial value is whether companies can embed AI into existing customer relationships and business use cases. Platforms with established customer bases, proprietary data and distribution channels typically have more avenues for capturing value. However, AI adoption, revenue, profits and cash-flow improvements do not occur in sync, and commercial returns still require further observation.

On infrastructure, Hu noted that as AI applications scale up, particularly as inference demand rises, infrastructure bottlenecks are extending beyond hardware such as GPUs and storage to electricity and data centers. “I think there is one thing we can keep in mind: there will always be bottlenecks, and they will emerge in different places,” he said.

During the on-site question-and-answer session, a media representative asked whether recent changes in market reports about computing-center procurement orders would affect investor confidence. Hu said infrastructure bottlenecks confronting computing-center construction are making investors more cautious about committing funds to such projects. Power constraints, conditions for deploying computing resources and financing arrangements in some regions could all affect project progress.

He noted that even when investors are willing to develop data centers, the projects may not necessarily obtain power supplies according to the original schedule. Issues involving data-center electricity and water consumption and grid-connection approvals in parts of the United States have already sparked discussion. The timing of power connections and the pace of infrastructure construction could also affect when projects begin operations and when computing capacity is delivered.

Potential risk: Commercialization and monetization may lag behind capital investment

Risks typically begin spreading from frontier model companies before extending through the computing supply chain.

“We believe frontier model developers will be the most vulnerable link because they face relatively intense competition from open-source models. Frontier model developers are also the main buyers of computing capacity, so the risks will spread to the computing supply chain as well,” Huang said.

The impact will first spread to cloud computing providers and neocloud service providers. These companies typically need to make substantial upfront investments in GPUs, data-center construction or leased computing resources while bearing the associated financing costs. If customer demand declines or customers are unable to fulfill their contracts, computing service providers could face cash-flow pressure and refinancing risks.

The pressure could spread further through the hardware supply chain, including server, memory and certain semiconductor companies. Huang said hardware capacity planning often needs to be carried out one or even several years in advance. If AI demand suddenly slows, companies may need to adjust orders and inventories, while capacity utilization could also be affected. Because production and capacity expansion involve fixed expenditures, the companies concerned may not be able to adjust costs quickly.

“By comparison, large cloud service providers have relatively strong resilience to risk,” Huang said. These companies have diversified businesses, broad customer bases, robust cash flow and financing capabilities. AI is only one part of their overall operations, so even if their AI businesses underperform expectations, the impact on their overall credit profiles would be relatively limited.

Regarding data-center projects, Huang said the projects covered by Fitch Ratings have performed relatively steadily overall, primarily because some have long-term contracts, tenants with strong credit quality and a degree of structural protection. The credit performance of such projects depends more on whether contractual counterparties can fulfill their obligations than entirely on short-term changes in AI demand.

(Article source: The Paper) Embedded advertisement in the middle of the article body

Source

Eastmoney