As hyperscalers commit vast sums to data centres, the more compelling story for credit investors may be the manufacturers supplying the physical infrastructure.
Much of that capital will be committed by hyperscalers such as Apple, Google, Oracle and Amazon among others. And that capital will then flow through to a much wider ecosystem of companies supplying the physical infrastructure required to build and operate data centres: heating, ventilation and air conditioning (HVAC) systems, electronical components, power generation equipment and other critical industry inputs.
For credit investors, this creates an important distinction. The companies funding the data centre build-out are not the only issuers affected by it. Many of the manufacturers providing these component parts are also represented in global credit markets – and, in some cases, in our portfolios. Understanding whether the data centre boom improves their business mix, cash-flow visibility and credit quality – or actually creates a new source of cyclical and customer concentration risk – is therefore directly relevant to portfolio construction and security selection.
Who bears the downside?
The key question is what happens if the boom proves less robust than expected. If demand disappoints, or a significant unexpected technological change occurs, the hyperscalers could be left with stranded assets, write-downs and a loss of credibility. But would the same be true for the diversified manufacturers that have supplied the build-out? Would HVAC companies such as Carrier and Johnson Controls face meaningful credit stress? Would electrical component providers such as Eaton or Schneider Electric be exposed to a sharp deterioration in fundamentals?
And would Caterpillar, which has attracted attention for its role in providing power generation equipment to data centres, become a credit casualty?
A comparison of investment amounts and free cashflow after investment helps us valuate this. Caterpillar probably has the most aggressive data centre-related investment plans among these manufacturers. Its expected capital expenditure for 2026 is around $3.5 billion, an increase of $1.2 billion from the prior year, much of which is likely related to data centres. Their free cashflow was $9.7 billion in 2025. By contrast, Amazon’s capital expenditure was $151 billion in the 12 months to March 2026, an increase of about $50 billion from 2025, while free cashflow moved from a positive $7.7 billion in 2025 to a negative $2.5 billion over the same period.
That difference is telling. For the hyperscalers, data centre investment is a major strategic and financial commitment, in other words an investment outflow. For the diversified manufacturers the exposure is a smaller part of their overall mix, and a source of revenue. Provided these firms do not become overly dependent on data centres, this should be credit positive.
These manufacturers are traditionally viewed as businesses with lower visibility into future sales and cash conversion than, for example, a military contractor with a long-dated government order. All else equal, this lack of visibility can weigh on credit quality because management teams – and investors – have less certainty around future earnings and cash flows. Data centres are different – they have relatively long planning and construction timelines, need critical components, and create demand that has longer visibility and relatively high certainty.
The result could be a longer-term improvement in business quality for HVAC and electronics providers. A data centre opportunity could provide better revenue visibility and useful diversification, assuming it is not entirely correlated with their existing cyclical exposures. This should be supportive of credit fundamentals.
The concentration question
The main caveat is concentration risk. Even if the various data centre projects carry different names, the ultimate source of demand may still be the same small group of hyperscalers. While data centre revenue in the 10%-30% range may be acceptable for a manufacturer, were that to increase above 30% it would require closer scrutiny. Investors would need to ensure they are not effectively owning data centre bonds dressed up in HVAC clothing.
We will therefore continue to monitor capital expenditures of these non-tech firms with data centre exposures. If rising investment were to put a strain on related credit metrics such as free cash flow relative to net debt, the assumption that these firms are ‘safe’ would need to be revisited. At present, however, the most visible free cash flow strain sits with the firms funding the data centre build-out, not with the diversified manufacturers supporting it.
The bottom line
If the data centre boom ultimately proves overextended, hyperscalers could face stranded assets, accounting charges and reputational damage. However, their core businesses should still support robust cash flow. For HVAC and electronics companies, the fundamental downside looks to be limited to manageable revenue revisions.
On that basis, data centre expansion seems to be a more obvious positive driver for diversified manufacturer credit than for the companies making the largest direct investments in the build-out.
