AI Euphoria Meets Reality Check: Markets Soar on Chips Deals as Executives Warn of Disruption�and Fragility

Summary: Investor enthusiasm for AI has lifted markets to records, but a series of signals�earnings wobble, circular financing, and geopolitics�suggest a more complex picture. A massive chips deal between Anthropic and Google Cloud underscores real infrastructure build-out, while Blackstone warns Wall Street may be underestimating how AI will disrupt legacy business models. Only 5% of firms are achieving AI value at scale, pointing to an execution gap. With a Trump�Xi meeting set and tariff threats over rare earths in play, CIOs and CFOs should model supply shocks and demand proofs, not just ride sentiment.

Are markets in an AI bubble�or just repricing the next industrial platform? Investor optimism has propelled indexes to records, even as geopolitical and earnings jitters flicker in the background? The Financial Times is putting the question to readers in a live Q&A on October 30, asking whether artificial intelligence is pushing markets into bubble territory�and what happens if it bursts?

Under the hood: record highs, muted breadth

Even with benchmarks near peaks, short, sharp pauses continue? Wall Street opened muted after a streaming giant missed earnings and was flat when other tech bellwethers underwhelmed, according to Reuters coverage of recent sessions? These wobbles suggest sentiment is lively but fragile, especially beyond the small cohort of mega-cap AI winners?

What�s powering the bull case? Hard investment? Anthropic�s blockbuster deal to secure access to 1 million Google Cloud chips (TPUs) in a pact worth tens of billions underscores how real the build-out is? Google plans to bring more than a gigawatt of dedicated computing capacity online for Anthropic in 2026�roughly the output of a large power plant? TPUs (specialized AI processors) promise better price-performance for training models than general-purpose chips, Google says?

Froth�or fundamentals?

Anthropic�s valuation hit $183 billion after a $13 billion raise in September, part of a wider wave of circular deals in which cloud providers are also investors and customers? That structure worries some analysts who see echoes of 1999�s vendor-financed growth? Yet the capital spending is building real infrastructure�data centers, grid interconnects, and chips�that can support multi-year productivity gains if demand persists?

Blackstone president Jonathan Gray argues markets may be focused on the wrong risk? �People say, �This smells like a bubble,� but they�re not asking: �What about legacy businesses that could be massively disrupted?�� he said, noting the firm moved AI to the first pages of its investment memos? Blackstone has avoided some software and call-center deals it deems vulnerable, while piling into power and data-center adjacencies? The lesson for CFOs: the winners and losers won�t break neatly along �tech� lines; rules-based white-collar work and data-heavy services face the sharpest shocks?

Execution gap inside the enterprise

There�s another reality check: only 5% of companies have achieved �AI value at scale,� according to BCG research cited by the FT? Generative and agentic AI pilots are common, but durable ROI remains rare without process redesign? The biggest near-term value is showing up in client-facing functions and IT, with case studies cutting hours-long tasks to minutes�but only when leaders assign operational owners, set measurable goals, and prevent pilot sprawl? In other words, no strategy, no ROI?

Policy risks you can model today

Geopolitics is a wild card with tangible P&L impact? The White House confirmed President Trump will meet China�s Xi Jinping in South Korea on October 30 amid threats of 100% tariffs if China doesn�t loosen rare earth export curbs? Rare earths�specialty materials used in motors, smartphones, and some chipmaking steps�sit upstream of the AI stack? Tighter controls or tariffs would raise costs for data centers and devices, potentially pinching margins in an industry already racing to secure power, cooling gear, and high-bandwidth networking?

For procurement teams, the takeaway is practical: hedge supply with multi-vendor chips strategies (as Anthropic is doing with different accelerators), lock in power and colocation contracts early, and run sensitivity analyses on tariff scenarios?

Bubble, bull run�or both?

So which is it? Signs of speculative excess are present in rich multiples, venture raises at lofty marks, and interlocking supplier-investor-customer deals? At the same time, the capex is real, workloads are compounding, and leading labs are scheduling gigawatt-scale capacity to train successively larger models? Short-term, earnings hiccups and policy shocks can trigger air pockets; medium-term, productivity proof points and cost curves will decide winners?

Three boardroom questions for the next 12 months:

  1. Can we tie incremental compute spend to measurable revenue or cost reductions within two to four quarters?
  2. Where are we exposed if a single accelerator, cloud region, or supplier faces a pricing shock or export control?
  3. Have we budgeted for workflow redesign�not just tool licenses�to unlock value at scale?

If you want to test your thesis�or your hedges�the FT�s live �Ask an Expert� on October 30 will crowdsource exactly that: what breaks first if optimism turns, and where the durable value really sits?

Glossary: TPUs are specialized processors tuned for AI math; a gigawatt is a measure of power capacity equal to one billion watts; rare earths are a group of elements critical to magnets and electronics?

Found this article insightful? Share it and spark a discussion that matters!

Latest Articles