AI enters its political phase: implications for US credit

This is an article from our Q4 Active Fixed Income Outlook. 

Over the past few years, artificial intelligence (AI) has primarily been viewed as a technology and investment story. However, it is now becoming a political one in the US and globally. The November US midterms may matter less because of the precise composition of Congress than because they reveal which AI concerns resonate with voters. AI still trails the economy, immigration and foreign affairs as a priority, but public anxiety about its consequences is increasing. For US credit investors, the key question is where that anxiety begins to constrain the investment cycle, alter the supply outlook and change the compensation required for AI-linked risk.

The emerging fault line is not simply Democratic versus Republican. President Trump’s administration has emphasised American AI leadership, competition with China, faster infrastructure development and relatively light federal regulation. Democrats generally favour stronger safeguards, but the coalitions are more complicated. Populist Republicans are also questioning the costs and concentration of AI, while governors and state legislators from both parties have resisted attempts by Washington to pre-empt state rules or limit their ability to regulate data-centre construction. 

That makes AI unusual: Both parties want the US to lead China, but neither can ignore the domestic costs of that race. We believe three political paths could emerge after the midterms, each with important implications for US credit.

First, China could win the political argument

If national security and technological competition dominate, policymakers may continue to prioritise rapid infrastructure development, permitting reform and access to energy. Bottlenecks in power, chips, powered buildings and skilled labour would likely extend the investment cycle rather than stop it. For US credit investors, this would support continued issuance from hyperscalers, utilities, data-centre developers and the broader supply chain – potentially including large, well-telegraphed deals from mega-cap technology issuers as well as securitised and project-related financings tied to data centre capacity. The opportunity would remain substantial, but so would supply pressure and the need for issuer and structure selection.

Second, an infrastructure backlash could become the primary constraint

The first meaningful restrictions may fall not on frontier models but on the infrastructure surrounding AI: data centres, electricity, water and tax incentives. These costs are visible to households and can generate bipartisan pressure in ways that abstract model-safety arguments cannot. Local opposition is already raising project costs and timelines, and could change where data centres are built. 

For US bondholders, this scenario could increase permitting, completion and refinancing risk. Attempts to secure capacity could pull capital expenditure and issuance forward, worsening market indigestion. Longer term, fewer projects could limit overbuilding risk and benefit completed assets. The result would probably be greater dispersion between operating projects and speculative developments, and among utilities able to allocate infrastructure costs without provoking a political response.

Third, labour could become the enduring political fault line

In the near term, the buildout is creating demand for specialised construction, engineering, electrical and technical workers. Those shortages can raise wages, construction costs and inflation before AI produces economy-wide disinflation. Over 
time, broader enterprise adoption could displace workers, particularly in entry-level and routine cognitive roles.

The absence of widespread layoffs this year does not settle that question.Enterprise deployment is harder than many expected: Companies must connect models to proprietary data, redesign workflows and build the governance, security and evaluation processes required for reliable use. Limited displacement today may therefore say more about implementation frictions than AI’s eventual labour-market effects.

That uncertainty has market implications. A gradual transition could support productivity, profits and company fundamentals. A more abrupt displacement cycle could impair consumption and create weakness in consumer, banking and lower-quality sectors of the market, even if GDP remains supported by investment. It could also produce a complicated rates outcome: weaker employment and disinflation at the front end, but stronger productivity, fiscal responses and a higher neutral rate 
at the long end.

From politics to financing

Across all three paths, politics ultimately becomes a financing question. A year on from the first mega AI debt deals, the buildout is clearly dependent on public bonds, securitised debt, private capital and hybrid structures. Recent research from Goldman Sachs tracks $578 billion of global AI-related debt issuance through mid-September this year.[1] 

This year has also shown that with risk-free yields elevated and AI deals receiving a choppier reception, capital is not unlimited. Equity and IPO funding can strengthen the cushion beneath creditors, but safety, political and return-on-investment concerns could undermine valuations. The most consequential question is therefore whether the industry can finance the buildout, manage its physical footprint and demonstrate that its benefits are broadly shared.

For US credit investors, the implication is a more selective – not simply more bearish –view of AI-linked risk. Strong demand, elevated all-in yields and disciplined issuance windows can still support spreads, particularly for well-capitalised hyperscalers and data-centre assets with clear contracted cash flows. But valuations leave less room for broad market tightening, and the next phase of the buildout is likely to bring more differentiation by issuer quality, funding structure and exposure to political or infrastructure constraints. 

The better opportunities may come where technical pressure has created adequate compensation in higher-quality AI-linked issuers, balanced with defensive duration and select wider-trading credits that can benefit if conditions merely stop worsening.The main risks are supply indigestion, weaker earnings quality, refinancing pressure and the possibility that politics turns AI from a growth tailwind into a constraint on where, how and at what cost the buildout can be financed. 

This backdrop reinforces our constructive but selective approach to AI-related US credit, which we view as a relative-value, issuer-specific opportunity rather than a single-sector allocation. Within technology, we favour hyperscalers with stronger balance-sheet capacity, greater funding visibility and more attractive valuations, while remaining cautious toward issuers facing repeated issuance or less predictable financing needs.

This is an article from our Q4 Active Fixed Income Outlook. 

Assumptions, opinions, and estimates are provided for illustrative purposes only. There is no guarantee that any forecasts made will come to pass.

[1]. Goldman Sachs,  as of 18 September 2026