Since the launch of OpenAI’s ChatGPT more than three years ago, most of the focus has been on the potential of artificial intelligence to disrupt, with not that much thought or action devoted to those that would be disrupted. That changed last week.
The convulsions that gripped the US sharemarket last week were triggered by the release of new tools for Anthropic’s Claude AI’s chatbot that are designed to automate a range of tasks in a number of service industries, notably legal and data services.
They were immediately seen as an existential threat to software businesses, particularly “software as a service” businesses, with stocks seen as exposed to that AI-generated disruption tumbling.
That wasn’t, however, the only contributor to a tumultuous week for AI-exposed or related stocks.
The companies that are the engine room for AI – the handful of mega tech companies building the large language models and data centres – lost nearly $US800 billion ($1.14 trillion) of market capitalisation last week after Amazon disclosed that its capital expenditures this year were expected to total about $US200 billion, with a heavy emphasis on AI investments. Last year Amazon spent $US135 billion.
With Google’s parent, Alphabet, saying its capex would be $US185 billion, Facebook’s Meta targeting $US135 billion, Microsoft more than $US100 billion and Oracle $US50 billion, the so-called “hyperscalers” will spend more than $US670 billion this year, or more than 2 per cent of the US GDP.
That number doesn’t include spending by the bigger unlisted AI vehicles like OpenAI, Anthropic or xAI, or the host of others providing services and capital to support the rollout of AI and its infrastructure.
There were, therefore, two quite disparate influences on the market last week.
One was the sudden realisation that AI as a disruptor would actually disrupt, potentially wiping out large swaths of existing industries and their value – or at least taking a very large bite out of their revenues and value – and the other was a more emphatic response to the awareness that developed last year that them scramble to dominate AI is getting so expensive that it is raising a question mark over its sustainability.
Anthropic’s release of its plug-ins was a demonstration of AI’s emerging substance. The reaction to Amazon’s capital spending plans was an amplification of investors’ fears that the ever-escalating spending required to realise the hyperscalers’ AI ambitions would result in ever-diminishing returns to existing investors.
Those realisations make this year a critical one for AI and the companies capitalising a potentially transformative technology with spending that is now nearing, or indeed-outstripping, even the massive legacy cash flows of the mega-techs.
Anthropic has demonstrated the potential of AI to abruptly undermine the economics of long-established sectors, with implications for not just for equity investors, but credit providers. Debt held against the perceived security of strong and stable cash flows from software as a service provider, for instance, suddenly doesn’t look as secure.
For investors in the hyperscalers, their dominance of asset and capital-light activities and the massive cash flows they generated might be enabling them to fund their AI investments, but it is also transforming them into capital-intensive businesses, with no certainty as to what the potential returns might be, if they materialise, or when they might eventuate on the scale required to justify the investments.
The more questioning approach to AI that began developing last year has now morphed into something of a wake-up call for investors.
Instead of seeing AI as either a transformative technology, with its pioneers all prospective winners, the response to the Anthropic release and Amazon’s spending budget shows investors are becoming more discriminating and recognising that there will be both winners and losers from the adoption of AI and that the risk-reward equation for the hyperscalers is evolving rapidly as the capital expenditures balloon.
The risks are growing rapidly in line with the spending, but the nature and scale of the rewards, despite the potential impact and implications of Anthropic’s new tools, remain opaque, and it remains uncertain whether any or all of the hyperscalers would actually emerge as winners for their investors even if AI were to live up to its potential.
What is obvious from the way the capex budgets for AI investments have exploded in the past three years is that all the major participants in the frenzy are becoming more vulnerable because they are increasingly dependent on access to external capital.
Even the hyperscalers are now tapping debt markets to fund an AI spending spree that is shaping as one of, if not the biggest, industry investment programs in history.
The more questioning approach to AI that began developing last year has now morphed into something of a wake-up call for investors.
Should credit providers become leery, or equity providers decide the potential rewards either don’t compensate for the risks, or that those rewards are too distant to invest today, or the demand for capital is greater than the ability or willingness of the markets to supply it to a single sector, the financial and economic consequences would be destructive and traumatic.
There’s also a more prosaic challenge for those now committed too deeply to pull back on their AI commitments.
The scramble to build data centres is already starting to inflate construction costs in the US. It is already having an impact on electricity prices. The centres’ energy and water requirements are outstripping existing supply, while new supply could take years to build.
Some of the physical infrastructure and inputs required to validate the enormous sums being poured into AI may therefore provide their own constraints on the financial returns.
The scale of capital and infrastructure requirements suggests there is a compelling case for a level of consolidation of the sector that goes well beyond the current circular, incestuous and, in some instances, illusionary nature of some of the funding.
“AI euphoria”, as some have characterised much of the period since the ChatGPT launch, is giving way to a more discerning and pragmatic attitude by capital providers. Not all those companies pursuing their AI ambitions can be winners, so some differentiation is required.
Equally, even if only some of those ambitions are realised, there will be losers, so existing business models – even those of the most technology-driven companies – perhaps especially those – will need to be evaluated for their vulnerability to AI.
The disruptive impact of AI and its coincidence with the investor backlash against the scale of capital commitments being made to develop is pushing the sector rapidly towards a critical moment, one where the question marks over its impacts and the sustainability of the spending become clearer. Last week’s events suggest that moment may not be too far away.
The Market Recap newsletter is a wrap of the day’s trading. Get it each weekday afternoon.
Stephen Bartholomeusz is one of Australia’s most respected business journalists. He was most recently co-founder and associate editor of the Business Spectator website and an associate editor and senior columnist at The Australian.Connect via email.From our partners
date: 2026-02-09 01:57:00