Guest opinion: The fight for capital – AI’s next bottleneck | By Sean Mills, Senior Portfolio Manager
The AI investment cycle is entering a phase in which access to capital is becoming as important as access to computing power itself. Building data centres, securing electricity, and purchasing equipment require enormous upfront spending. As that spending outpaces internally generated cash flows, financing the continuation of the expansion becomes an investment issue in its own right.
The inflation backdrop compounds this challenge. Energy supply disruption associated with the Iran conflict has raised costs, while AI investment adds new demand pressures for electricity, equipment, and construction capacity. These supply and demand price pressures feed straight through into input costs and therefore higher inflation. The Federal Reserve’s September rate increase illustrates this renewed pressure on monetary policy.
AI businesses are therefore confronting two rising prices: the cost of building compute capacity and the cost of funding it. S&P Global Ratings expects negative free operating cash flow across six major Hyperscalers in 2026 and 2027, with debt, leases and other financing arrangements supporting expansion. Profitable businesses can still need outside funding when investment opportunity expands faster than cash generation can keep up.
The battle for capital
This borrowing however also changes the bond market. In August, J.P. Morgan Asset Management found Hyperscaler debt trading at wider spreads than similarly credit rated bonds after adjusting for duration. Credit spreads measure the extra yield over comparable government bonds. Hyperscaler debt issuance in 2026 is already almost 70% of Treasury bond issuance – that puts additional pressure on yields through oversupply of debt itself. Investors want compensation for the heavy issuance and uncertain future cash generation. Increased borrowing therefore competes for investors’ capital, potentially forcing other issuers to offer even higher yields in the fight for capital.
Higher funding costs need not stop the strongest projects. SpaceX management recently reported payback periods below one year for new compute deployments at current pricing, although the call provided no detailed cost reconciliation. Projects offering sufficiently high returns can justify expensive financing as long as they produce an economic profit at the end of it – simply; the return of the investment must compensate sufficiently above the cost of capital used to make the investment. However, even short payback periods can leave funding gaps when successive rounds of investment grow faster than the resulting cash receipts – it is a difficult balance to successfully navigate.
How AI is reshaping capital allocation
This creates an important possibility: the AI boom could tighten financing conditions elsewhere before it restrains its own expansion. A developer earning modest returns on housing or commercial property may abandon a project at a borrowing cost an AI business can now willingly absorb. Higher expected productivity could redirect capital across the economy, raising the hurdle rate for competing investment. Together this may result in capital being allocated to its best & most efficient use which would directly impact global economic growth.
For investors, that makes financial strength and financing discipline increasingly valuable. Selective lenders may receive better compensation, while existing bonds with longer maturities remain exposed to rising yields and higher refinancing costs. Real assets face a double effect: stronger demand for their services, but potentially lower valuations as discount rates rise. Attractive project economics still require dependable cash flows to service debt.
Conclusion: the price of AI’s success
Over time, AI could ease inflation through productivity gains. But as the opportunities to invest multiply, competition for the capital to fund them could intensify. Cheaper goods would not necessarily mean cheaper capital. The consequences would reach far beyond technology, shaping which businesses can expand, which innovations reach the market, and which projects never get built. The price of AI’s success may be a higher cost of capital for the rest of the global economy. The next phase therefore may reward those who allocate, strategically execute, and provide capital most effectively.
Warnings
The views, thoughts and opinions expressed within this article are those of the author, and not those of Capital International Group Limited (Group) and/or any of its subsidiary companies and as such are neither given nor endorsed by the Group or any company within the Group.
Information in this article does not constitute investment advice or an offer or an invitation by or on behalf of any company within the Group to buy or sell any product or security or to make a bank deposit.
Any reference to past performance is not necessarily a guide to the future. The value of investments may go down as well as up and may be adversely affected by currency fluctuations.
The Group, its subsidiary companies, clients, and officers may have a position in, or engage in transactions in any of the investments mentioned. Opinions constitute views as at the date of issue thereof and are subject to change.
