Asia-Pacific technology hardware firms are benefiting from the artificial intelligence (AI) investment boom, but material downside risks could threaten strong demand across the AI-related value chain over the next two years, S&P Global Ratings said on Thursday.

According to its report, Asia-Pacific tech hardware firms are getting rich off the AI investment boom, and its base case assumption remains that demand will stay strong along the AI-related value chain over the next two years, but downside risks are material.

“We stress-tested rated AI supply chain companies in Asia-Pacific against two slowdown scenarios–the first largely driven by bottlenecks, the second by a change in investment appetite,” said S&P Global Ratings credit analyst Cathy Lai. “The results show some variance in resilience.”

According to the report, companies with significant AI exposure stand to benefit the most from AI-related spending, which it forecast will well-exceed $1 trillion annually in 2027-2028.

This is driving growth for Asia’s tech hardware sectors, leading to a number of high-profile rating upgrades from us in recent months.

Key downside risks include grid limitations, land scarcity, and regulatory hurdles that could delay or cancel plans for data centers or other AI-related projects.

A change in investment appetite is another potential overhang that could moderate capital expenditure (capex) by the big spenders, including Alphabet Inc., Amazon Inc., Microsoft Corp., Meta Platforms Inc., Oracle Corp., and SpaceX.

S&P stress tests looked at four key areas of the AI-related supply chain:

Firstly, foundries which manufacture chips designed by companies such as Nvidia Corp. Major examples include Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC).

Secondly, memory makers including SK Hynix Inc. and Samsung Electronics Co. Ltd. that produce chips that store data.

Thirdly, other components, such as cooling systems used by data-center servers.

Fourthly, original design manufacturers (ODMs) that design and assemble servers, sourcing many parts from the above three subsectors.

Revenue, earnings before interest, taxes, depreciation, and amortization (EBITDA) and cash flows would be resilient in scenario 1, which tested delays due to bottlenecks or infrastructure limitations, said S&P.

Under its harsher scenario, in which investment appetite wanes for AI and capex peaks early, the rating agency expects more pronounced EBITDA declines, particularly for memory makers.

Interestingly, it believes free cash flows for many rated issuers would be less affected than EBITDA, and in some cases, even show a positive, albeit temporary, impact relative to our base case.

This is because of a reduction or delay in the tech firms’ capex, as well as a temporary reversal of the substantial working capital outflows seen in recent years–particularly ODMs.

It is noted that over the past two years, ODMs have experienced significant working capital pressure due to increased inventory holdings and longer receivable periods; these trends are likely to reverse during the initial period of a downturn.

In S&P view, foundries would hold up the best under both scenarios; especially TSMC, due to its market leadership in advanced-technology products.

Other factors for resilience across the sectors include substantial capacity undersupply and revenue diversification.

Moreover, its financial forecasts for issuers in highly cyclical industries—such as the memory sector–incorporate downside risks.

“Rated Asia-Pacific tech-hardware issuers maintain sufficient buffers against delays in AI spending plans,” said S&P Global Ratings credit analyst Clifford Kurz. “A fundamental shift in long-term AI demand could pose credit risk.”

BMI sees AI supercycle to boost growth in East Asia