Investors are starting to doubt whether the global artificial intelligence boom can keep paying for itself. AI runs on data centres that cost billions to build, and companies are finding it harder to raise that money. The Philadelphia Semiconductor Index, which tracks the chipmakers that supply those data centres, has fallen 27% from its June peak. Even US tech giant Meta is feeling the squeeze. Its shares dropped 7% after it revealed plans for a new class of shares to help fund AI spending of at least $US135 billion this year.
Borrowing has become more expensive across the board. Oil above $US100 a barrel has kept inflation high, and in Australia four rate hikes by the Reserve Bank this year have lifted the cash rate to 4.6%. AI companies are being hit hardest because they are borrowing huge sums at once to build data centres whose revenue is still years away, leaving lenders with more risk and too much AI debt to absorb. That has pushed yields on bonds issued by US data centre operator CoreWeave to 15%, a level fund managers say is approaching distressed territory.
Australia’s AI players have hit the same wall. Firmus failed to raise money from public investors, pulling its $44 billion ASX float, which would have been the biggest since Telstra in 1997, after investors baulked at the valuation. Sydney-founded data centre operator AirTrunk then postponed its own $US1.5 billion listing in Singapore. NextDC shares fell 40% after power connection delays at its Melbourne site forced it to cut earnings guidance by more than a third. Its borrowing costs have since jumped by a third in a week. That matters well beyond NextDC: a $155 billion pipeline of Australian data centre projects depends on that kind of funding. The pain has also spread to suppliers whose revenue relies on these projects going ahead. Maas Group, which builds equipment for Firmus, and SKS Technologies, which supplies NextDC, have each lost about half their value in a fortnight.
With AI stocks now seen as high risk, fund managers are shifting into safer sectors such as insurance and defence, or holding cash until conditions settle. Demand for AI may still be strong, but the funding wall means capital will now flow only to companies with the strongest balance sheets.
