Market scepticism grows as AI infrastructure spending reaches new heights

Skepticism is growing over the aggressive capital expenditure (capex) plans of major technology companies heavily investing in artificial intelligence (AI) infrastructure. According to BCA Research, the market may be entering a “verse Moment,” a phase where investor enthusiasm gives way to scrutiny and potential punishment of AI firms for their rising capex commitments.

Leading players such as Meta, Amazon, Microsoft, Google (Alphabet), and Oracle are projecting combined AI-related assets worth roughly $2 trillion by 2030. This scale of investment implies annual depreciation costs of around $400 billion if these assets maintain a five-year lifespan, an amount exceeding their current combined trailing earnings. This massive financial ask has raised questions about its sustainability and whether the anticipated productivity gains from large language models (LLMs) and AI innovation will materialise as hoped.

Meta Platforms, in particular, exemplifies this trend. The company has sharply increased its capex forecast for 2025, narrowing it to a range between $66 billion and $72 billion. These investments include building extensive data centres and expanding use of AI chips such as Nvidia’s, and are part of CEO Mark Zuckerberg’s vision for AI to serve more than a billion people within the year. Meta’s formidable bond offerings totalling up to $30 billion and multi-year financing deals, including a $27 billion agreement for its ‘Hyperion’ data centre, underscore the scale and ambition behind this push.

However, voices within the AI research community are becoming increasingly cautious. Yann LeCun, Meta’s former AI chief, has publicly doubted whether current LLMs can generate truly new knowledge, an essential factor for transformative productivity improvements justifying such large capital outlays. Supporting this view, some studies suggest AI tools may actually reduce efficiency for experienced programmers, challenging the prevailing narrative of an AI-driven productivity bonanza.

BCA Research also pushes back on the notion that early AI leadership creates long-term strategic advantages. Unlike some tech innovations that benefit from strong network effects or scalable intellectual property, AI infrastructure appears more akin to the airline industry: capital-intensive, commoditised, and hostile to durable first-mover benefits. The technology, and much of the infrastructure supporting AI workloads, can be replicated by competitors with sufficient capital, diluting any sustainable competitive edge.

Market indicators appear to reflect this growing caution. Meta shares have dropped nearly 20% from their August highs, Oracle’s stock has fallen about 40% from peak levels, and credit spreads for both Oracle and CoreWeave have widened markedly. These movements are viewed by BCA as early signs that investors may be starting to question aggressive AI spending.

Adding to these concerns is the broader macroeconomic context. While some market participants anticipate a Federal Reserve rate cut in December that could alleviate pressure on equities, BCA’s fixed-income analysts argue that the likelihood of the Fed holding rates steady is underestimated by the market. A hawkish surprise, coupled with weakening global economic data across manufacturing, trade, and European activity, could prompt a more pronounced risk-off sentiment.

Reflecting these risks, BCA recommends a modest underweight to equities within balanced portfolios, maintaining a 57% allocation against a 60% benchmark in its GIS FullView Portfolio, with a bias toward further reductions pending an early December reassessment.

From a wider perspective, bond markets have seen a marked shift as these tech giants increasingly seek debt financing to support their AI and cloud infrastructure ambitions. Collectively, Amazon, Microsoft, Google, Oracle, and Meta have raised close to $100 billion through bond sales this year alone, nearly doubling the prior year’s data centre investment outlays to around $400 billion.

Jefferies forecasts that the world’s largest cloud and AI infrastructure providers will collectively spend an estimated $417 billion on capex in 2025, a 64% rise from 2024 and a 168% increase since 2023. This explosive growth, driven by unrelenting demand for AI capacity, continues to put pressure on infrastructure supply chains and raises questions about whether such surges in spending will eventually yield commensurate returns.

In summary, while the technology sector’s fevered investments in AI infrastructure signal confidence in transformative digital progress, scepticism is mounting over the scale, pace, and efficacy of the associated capital spending. Market corrections, cautious analyst forecasts, and emerging doubts from AI researchers collectively spotlight the potential for a “verse Moment” where the market recalibrates its enthusiasm, imposing a more cautious and selective view of the AI investment boom.