If the ambitious AI investment plans are to be believed, an estimated additional 80–120GW of capacity needs to be constructed over 2025–2030.
However, at such scale over a relatively short period there are risks supply will not be able to match demand. The IEA forecasts in their base case scenario ~110GW of data centres capacity to be added between 2025–30, of this ~20GW may be at risk of being delayed due to grid constraints. Data centres operators are racing to bring as much renewable, nuclear and fossil fuel energy sources online as quickly as possible to meet future demand.
How will all this investment be funded?
To date the majority has been funded by the free cash flow of the world’s most profitable companies. However, as these companies’ free cash flows begin to approach zero and investments continue to rise, three alternative sources of financing are becoming more prevalent:
- Debt financing – e.g. Oracle must raise debt to invest at a similar scale to the largest players. Hyperscalers have yet to meaningfully use debt financing.
- Equity financing – e.g. OpenAI funding rounds being used to potentially self-build capacity.
- Vendor financing – notably Nvidia has agreed to invest up to USD100bn into OpenAI to support the companies’ ambitions.
The shift to debt and vendor financing indicates we have entered a new stage of this cycle, although how long it is before the eventual correction occurs depends ultimately on whether:
- the collective belief in continued technological advancement and monetary returns from investing in AI is maintained
- there is enough remaining capital to be deployed. For the time being, there is but that may not last.
US vs. China
Both the US and China view AI as a strategic technology with implications for economic growth as well as national security. Whilst Chinese tech giants have increased their AI investments they are yet to match the scale of their US counterparts. Alibaba has been the biggest spender and most vocal to date, in September 2025 announcing their intention to invest “over RMB 380bn” (USD53bn) into AI over a 3-year period. China may well have the energy advantage as it mobilises its prowess in sales and nuclear energy. Outside of the US, China is likely to be the second largest spender when it comes to building AI infrastructure.
The US has targeted slowing down Chinese AI progress through semiconductor export restrictions and to a lesser extent tariffs. China in response has leveraged US industrial dependencies on rare earths.
As US Secretary of Commerce Howard Lutnick so bluntly put it: “We don’t sell them our best stuff, not our second best stuff, not even our third best. The fourth one down, we want to keep China using it. We want to keep having the Chinese use the American technology stack, because they still rely upon it”.12
In parallel to trade negotiations each country is following a path of trying to localise strategic resources: for China chip fabrication and design, for the US reshoring manufacturing and accessing critical materials.
Humanoid robots – from sci-fi to the factory floor
An emerging application of AI comes in the form of autonomous robotics. Humanoid robots, having once lived purely in the realm of science fiction, are increasingly showing promising signs of reaching technological maturity. From factory floors to hospital hallways, humanoid robots are stepping into trials as a new age labour force.
The promised benefits are ambitious, a tireless 24/7 worker that can slot into human-designed spaces opens an almost limitless set of applications: from factories, logistics warehouses, healthcare services, retail and beyond. For economies battling ageing demographics and frequent staffing shortages robotics are a compelling proposition to achieve productivity gains.
Today these autonomous robots are predominantly in pilot trials across sectors rather than seeing imminent mass adoption. Technological capability and cost must continue to make progress, but the trends so far are promising. Costs have come down from USD100k+ to USD35–60k in some leading models, with some Chinese firms suggesting they’ll be able to reach <USD15k per unit.
Robotics may be another sector the US and China battle over, with plausibly massive economic rewards from being able to automate many more manual labour tasks there are strong incentives to localise this industry.
