The first full week of October 2026 showed one thing clearly: building AI now costs so much that even the biggest companies are borrowing to keep up.
Record-sized money
- SoftBank is reportedly talking to Gulf investors about a fund of up to $100 billion that would buy established businesses and use AI to run them better. The talks are said to be at an early stage.
- SpaceX is reportedly lining up around $40 billion in bonds and bank loans to buy Nvidia AI chips. Broadcom and Oracle are also in large financing deals.
- Morgan Stanley estimates AI infrastructure could need about $1.5 trillion of outside financing by 2028.
Not every bet is working
Australian AI data-centre company Firmus withdrew a planned IPO, with reports that only a small part of its planned capacity is running so far. It is a reminder that big promises and real capacity are not the same thing.
Investing in science
Nvidia pledged $1 billion over five years for research and computing in the US, across AI, quantum computing, healthcare and energy.
People want a slower pace
An AP-NORC poll of over 2,000 US adults found that 64% feel AI is developing too quickly. Around eight in ten said keeping AI under human control and protecting workers should be government priorities.
Why it matters
The AI race is no longer only about who has the smartest model — it is about who can pay for the chips, power and data centres. At the same time, public pressure for safety and rules is growing, from Washington to Brussels to New Delhi.