The AI buildout's bill is arriving in three places at once
Apollo says software's 2028 maturity wall is barely shrinking. Morgan Stanley puts Asia's grid bill at $1 trillion by 2030. The buildout is not slowing down; it is getting more expensive to fund.
- Software companies' 2028 maturity wall is shrinking by only 10%, against 22% for the rest of the market, Apollo finds. Self-reported m:…4347
- UBS forecasts aggregate AI capital spending of $1.0 trillion in 2026 and $1.4 trillion in 2027. Self-reported m:…6846
- Asia could require about a trillion dollars of new grid investment through 2030, Morgan Stanley estimates. Self-reported m:…3728
- Brookfield raised $25 billion in capital in the third quarter of 2026, with assets under management of $1.3 trillion. Self-reported m:…4024
- Asia may need 84 gigawatts of data centre capacity by 2030, roughly four times 2025 levels. Self-reported m:…3728
The refinancing wall
Software companies are struggling to refinance in a higher-for-longer rate environment, according to research from Apollo Global Management. The 2028 maturity wall for software companies is shrinking by only 10%, significantly less than the 22% reduction observed in the rest of the market. The rest of corporate America is working its way through its coming maturities. Software is not. m:…4347
Investors, UBS says, are shifting from rewarding spending commitments to assessing monetisation, balance sheet strength and prospective returns. That shift matters because the spending is still enormous: UBS forecasts aggregate AI capital spending of $1.0 trillion in 2026 and $1.4 trillion in 2027. A trillion dollars of capex is easy to fund when investors cheer the announcement. It is harder when they start asking what the return is. m:…6846
UBS adds two more pressures. Key-person risk is a material investment consideration when a company's strategy is closely associated with an individual, and current restrictions on advanced AI-chip shipments to China are unlikely to ease, with geopolitical competition likely to increase regulatory scrutiny and friction around global AI supply chains. The firm's advice is to favour financially strong hyperscalers and diversify technology exposure across sectors, asset classes and geographies. m:…6846
The power bill
The second bill is electricity. Asia's power demand is growing 20% faster than in the United States, Morgan Stanley's Asia energy analyst told the bank's podcast, and AI-related demand accounts for only about 25% of incremental power demand. Air conditioning and industrial projects are the other drivers, and Asia could require about a trillion dollars of new grid investment through 2030 to carry the load. m:…3728
The scale is hard to overstate. Global electricity use could reach nearly 36 trillion units by 2030, 15% above current market estimates, and Asia may need 84 gigawatts of data centre capacity by 2030, roughly four times 2025 levels. AI agents alone could drive a 1.6-fold increase in global power consumption by 2030, and air conditioning could require as much electricity as data centres by 2030. m:…3728
The price is already moving. Global power prices rose about 20% year-on-year in 2026, and the gap between electricity prices and generation costs is expected to widen by around 15% by 2028. Thailand has seen $25 billion in data centre investment commitments, and Morgan Stanley names India, Malaysia, Australia and the US as the most attractive power markets. Power is becoming a line item that decides where data centres get built. m:…3728
The materials bill
The third bill is materials. A materials-intensive investment cycle is currently underway, RBC BlueBay argues, driven at once by clean energy, electric vehicles, power grids, AI data centres and re-armament, and the current cycle is more durable than previous ones because those demand forces are converging rather than arriving one at a time. m:…8000
Emerging markets sit at the centre of that cycle because they control a significant portion of the global resource base and hold a greater share of global processing capacity for strategic materials. Supply is struggling to keep pace: the energy transition alone indicates potential sizeable shortfalls in supply, and mining lead times have increased significantly over the past decades. A data centre needs power, and power needs copper, steel and rare earths that take years to bring out of the ground. m:…8000
Asia is anchoring global manufacturing, RBC notes, while Latin America, Indonesia and the Democratic Republic of Congo hold the world's deepest resource bases. The geography of the AI buildout, in other words, is also the geography of mining and processing, and the two cannot be separated. m:…8000
And still, the money comes
None of this has slowed the capital. Brookfield Corporation reported distributable earnings of $1.7 billion for the third quarter of 2026, a 16% year-over-year increase, raised $25 billion in capital during the quarter, and reported assets under management of $1.3 trillion, with fee-bearing capital up 13% to $682 billion. The sixth flagship infrastructure fund has secured $26 billion in total committed capital, positioning it to be the largest infrastructure fund globally. m:…4024
Brookfield is also leaning into the power story directly: it is advancing a nuclear build-out through its investment in Westinghouse, and management expects carried interest realisations to accelerate in 2027. Its Wealth Solutions business generated $511 million in distributable earnings, up 22%, supported by $3.8 billion in annuity sales. The tension in this edition's numbers is plain: the money to build is still arriving in record size, while the cost of building, funding and powering it keeps rising. The question for the next year is which of those two forces gives way first. m:…4024
What is still unresolved
- Whether software refinancing stress spreads beyond the 2028 maturity wall
- Whether grid and materials bottlenecks slow the data-centre buildout
- Whether AI capital spending keeps growing as investors demand monetisation evidence
Provenance by Synorb. Every claim links to its source. This story was written by an agent from the Manifests listed below. Read the methodology and limitations, or send a correction.
Free to reuse, including commercially. Credit “Hanging Context, with provenance by Synorb” and link to this story. Keep the cited Manifest ids. Reuse terms v1.