For nearly four years, Wall Street has been happy to underwrite the boom in artificial intelligence spending, providing blank cheques without asking too many hard questions.
asserted
Street → underwrite → questions
Now, it seems, investors have become uneasy about the massive spending spree they have helped ignite.
asserted
they → seem → spree
Late last week, as a collective, the biggest tech companies lost almost $US1 trillion ($1.4 trillion) of market value, headed by a 13 per cent, or $US500 billion, plunge in the value of Google’s parent, Alphabet.
asserted
companies → lose → parent
Shares in the four big hyperscalers – Alphabet, Microsoft, Meta and Amazon – all fell heavily after issuing third-quarter results that showed capital expenditures soaring and, for all but Microsoft, revenue growth from their core business weakening.
asserted
growth → fall → business
Concerns about the rate of spending by the hyperscalers – as a group, probably the most profitable and valuable companies in history – have risen even as their investments in AI have accelerated.
asserted
investments → spend → AI
In the latest quarter they invested more than $US200 billion, compared with $US113 billion in the same quarter of last year.
asserted
they → invest → year
They expect to spend at least $US230 billion in the December quarter.
asserted
They → expect → quarter
Increasingly, that spending is outstripping even the vast cash flows they generate from their legacy businesses.
asserted
they → outstrip → businesses
Of the four biggest hyperscalers, only Microsoft is generating positive free cash flows.
asserted
Microsoft → generate → flows
It had $US18.6 billion of free cash flow in the quarter, whereas Alphabet (minus $US9.1 billion), Amazon (minus $U33.5 billion) and Meta (minus $US2.1 billion) racked up nearly $US45 billion of net cash outflows between them.
asserted
Alphabet → have → them
In the same quarter last year they had positive net cashflows of more than $US36 billion.
asserted
they → have → billion
They’re all reporting record levels of capex and have foreshadowed higher levels of spending in the fourth quarter and into next year.
asserted
They → report → year
They’ll spend around $US750 billion between them this year and, from their guidance, more than $US1 trillion in 2027.
asserted
They → spend → 2027
With their spending outstripping cash generation they’ve had to raise new equity and, increasingly, borrow.
asserted
they → outstrip → equity
Alphabet and Meta have, between them, raised almost $US90 billion of new equity since June and have programs that could raise a further $US100 billion.
uncertain
that → raise → billion
All the hyperscalers are issuing debt to fund their AI spending.
asserted
hyperscalers → issue → spending
Last week S&P Global Ratings lowered the outlook for Meta’s AA- credit rating to negative, citing the increased spending on AI, its increasing leverage and its negative free cashflows.
asserted
Ratings → lower → AI
Also last week, Blackstone, which is the key investor in a consortium financing Anthropic’s $US63 billion purchases of semiconductors from Google, told its investors that the leading frontier lab may not gain an investment grade rating.
uncertain
lab → finance → rating
If Anthropic – about to try to raise $US100 billion of equity in an initial public offering valuing it at up to $US2 trillion – were to issue bonds with a junk bond rating, it would diminish the pool of potential investors and increase the cost of the funding.
asserted
it → try → funding
It would also sound a warning bell for other, lesser, participants in the AI ecosystem.
asserted
It → sound → ecosystem
It’s not just access to debt and the cost of gaining that access that is becoming an issue for even the biggest and best of the AI aspirants.
asserted
that → ’ → aspirants
Access to equity – and its cost – is becoming tighter and more expensive.
asserted
Access → become → equity
The company that started the AI frenzy, OpenAI, deferred its planned $US1.5 trillion-plus float this year into 2027, but is still trying to raise $US30 billion of new equity at a valuation of $US1.4 trillion – or was trying to raise that amount on those terms.
asserted
that → start → terms
The Financial Times is reporting the amount it is seeking has been cut to $US20 billion and the valuation to $US1.25 trillion as the market turmoil caused demand for its shares to weaken.
asserted
demand → report → shares
Like most of the standalone AI companies, OpenAI has been able to raise vast amounts of capital at valuations that have increased at a mind-boggling rate.
asserted
that → raise → rate
It was valued at “only” $US852 billion in March, a month after Anthropic raised $US30 billion at a valuation of $US380 billion.
asserted
Anthropic → value → billion
If those valuations were ever to reverse, those AI companies without legacy revenues would lose their access to capital and die a near-instant death.
asserted
companies → reverse → death
Their lifeblood is continuing access to capital at ever-increasing valuations.
asserted
lifeblood → continue → valuations
Even for the hyperscalers, their ability to raise capital easily and cheaply is being threatened by the more rigorous scrutiny from investors who previously signed blank cheques.
asserted
who → raise → cheques
Those risks, unique to AI, were highlighted again last week when Nvidia and fellow chip-makers AMD and Broadcom committed to taking up $US5 billion of Meta’s $US25 billion equity raising.
asserted
Nvidia → highlight → raising
Meta is, of course, a major customer of the companies and therefore they are helping to fund the purchases of their own products.
asserted
they → help → products
That’s a typical AI transaction – a circular finance transaction – where the chipmakers with cash, Nvidia in particular – have provided vendor financing to the frontier labs, hyperscalers and data centre operators to support demand for their chips.
asserted
chipmakers → ’ → chips
Shifts in access to external funding, or its cost, could generate implosions across the sector.
uncertain
Shifts → generate → sector
The crash in Alphabet shares showed even hyperscalers aren’t immune.
asserted
hyperscalers → show → shares
However, the issue that drove its shares to their biggest loss in its history wasn’t the risk of a loss of access to funding markets, but the impact of AI on what had appeared to be a financial fortress.
asserted
what → drive → AI
Google, Meta and Amazon have been so dominant, and the “moats” that have protected their massively profitable businesses so deep and wide, that it was almost unthinkable their defences could be breached and the massive margins and returns on capital threatened.
uncertain
margins → protect → capital
Shifts in access to external funding, or its cost, could generate implosions across the sector.
uncertain
Shifts → generate → sector
Google’s dominant search business generates more than half its revenues, with the advertising revenue it generates very high margin.
asserted
it → generate → margin
It’s third-quarter report, however, showed that its search revenues were less than analysts predicted and their growth rate of 12 per cent significantly lower than the 17 per cent generated in the second quarter and 19 per cent in the first.
asserted
rate → show → first
Google is getting strong growth from “AI Mode” queries using its Gemini AI platform – but that growth is in revenues that are much lower-yielding than those high-margin revenues from advertising.
asserted
that → get → advertising
…and 5 more, not listed.
The Aporia
