Financing the AI Boom – what are the risks for credit investors

Written by Matt Cornwell

The level of interconnectivity across the AI ecosystem has ramped up a notch in 2025 and you will be forgiven for struggling to understand how many chips Nvidia have sent Microsoft or how many billions Softbank have committed to spend or how many data centres Meta wants to build. Taking a step back, Microsoft first invested $1bn in Open AI back in 20191, we got the launch of Chat GPT in 2022 and then in 2024 we saw the Hyperscalers start to ramp up AI infrastructure spending. The arrival of the low-cost model DeepSeek in January of this year briefly dampened sentiment but has spurred competition and we close third quarter reporting season with an estimated 1.2trn2 worth of spending commitments announced this year. So, a race to the bottom?

How these deals are funded and the true economics behind each commitment is becoming increasingly opaque too. Vendor financing has been a key feature and can take a couple of forms, but we see both suppliers are funding customer operations through equity investments (e.g. OpenAI’s investments from Nvidia) or manufacturers borrowing themselves to supply lines of credit to supplier operations. This allows the recipient to obtain financing elsewhere beyond what would be available based on current profitability or cash flow levels (Open AI are expected to burn $115bn by 20293). This is nothing new of course, Nortel was the main culprit in the dotcom bubble of similar arrangements, and we know how that turned out for them.

Additionally, we know many of the deals include repurchase and revenue sharing agreements which further muddies the picture for investors on the direction of cash flow. That’s not to mention how the financing of the actual commitments is going to be made, how long these deals run for, where the power is going to come from and most importantly the return expected on these investments. Meta recently showed how not to spend $50bn with their investment into the Metaverse but the level of integration that AI has achieved means we are unlikely to see a repeat here. A more likely outcome could be akin to the dotcom overbuild given that this technology did not become obsolete but returns on the investment proved extraordinarily poor.

The quantum of these deals, along with so many unanswered questions make us bond investors nervous and therefore happy to be on the sidelines. But AI mania is now spreading into the public fixed income realm as hyperscalers look to diversify funding sources we have seen record amounts of demand for AI related bond issuance. Alphabet was the latest issuing $25bn of IG corporate bonds but this was after Broadcom launched an $18bn deal at the end of September and Meta topped both of these, issuing $30bn as they look to finance upwards of $115bn in capex in 2026. Combined orders across these deals were nearly $300bn so there is no lack of appetite amongst bond investors. And this does not include the $27bn raised in the largest single corporate bond issued as part of Meta’s JV with Blue Owl for their Hyperion data centre. So, it is not just the equity markets getting creative from a financing point of view then with Meta guaranteeing the lease payments whilst also creating an off-balance sheet liability, so a win-win for them.

Hyperscalers are driving this surge in issuance but there are knock on effects for other sectors, mainly the utility companies that are trying to grapple with the surge in energy demand. They too have put in funding plans to meet the estimated 90GW4 of additional power required by 2030 due to data centre demand. There is a clear risk of overbuild here, the largest utility companies have thus far taken a cautious approach putting in minimum volume commitments, early exit fees and in some cases collateral on data centres therefore reducing this risk. But counterparty credit risk remains and should be monitored. Noting that the main hyperscalers are low levered with significant cash balances and solid credit ratings (e.g. Alphabet’s AA+ rating is on par with the US government). There is less concern for now in the public fixed income space but given the private equity/credit involvement in

new data centre construction, we worry about recovery rates and the financial implications of a pullback in demand.

Credit markets would not be immune to an equity led sell-off in AI names. The US is more exposed than Europe with ~15% of the IG index closely tied to the AI industry but on the whole fundamentals are more favourable given the level of regulation in the utility sector as well significant cash balances in the tech sector. This is reflected in these names trading inside the index average, although the recent issuance wave has led to some underperformance and has caused reverberations across the broader IG markets. We see improving fundamentals across the IG universe but expect the technology sector to be the outlier given the levels of issuance, noting that this is mainly skewed to the longer end of the curve.

Therefore, we remain cautious and see no need to chase risk in more exposed issuers or drop down the credit spectrum where covenants are weakening or lending standards deteriorating. Nevertheless, we are overweight utilities preferring the core European names generating decent cash flow and backstopped by recent equity raises. Additionally, we are not averse to participating in hyperscaler issuance where pricing is attractive, but do not see any value in extending duration in what is a relatively flat credit curve that does not compensate for the risk of capex overspend in the medium term

Disclaimer:

This article is of a general nature and intended for information purposes only, it is not intended for distribution to any person or entity who is a citizen or resident of any country or other jurisdiction where such distribution, publication or use would be contrary to law or regulation. Whilst all reasonable steps have been taken to ensure that this article is accurate and current at the time of publication, we shall accept no responsibility or liability for any inaccuracies, errors or omissions relating to the information and topics covered in this article. The opinions expressed may be subject to change or amendment at the discretion of the Investment Manager without notice.

Nedgroup Investment (IOM) Limited (reg no 57917C), is licensed by the Isle of Man Financial Services Authority.

Nedgroup Investments (UK) Limited is regulated by the Financial Conduct Authority.