Investors shifted away from the hottest AI names last week, creating a clear market test as funding costs and central bank action altered the risk-reward for growth exposure.

AI has demonstrated important productivity potential but the compute and infrastructure buildout are expensive, and those costs rose after The Federal Reserve moved to higher rates following the U.S.-Iran conflict.

Large-cap tech companies are expected to fund a substantial AI infrastructure program, with an estimated total spend cited at $1.5 trillion, and higher interest rates make that capital more costly.

Financing flows have tightened the link between hardware vendors and their customers. Nvidia has not only sold GPUs to major buyers but has at times offered direct capital support or equity to firms building GPU-powered infrastructure.

Recent financing examples include Firmus in Australia, which abandoned a planned IPO and entered talks for further funding from Nvidia; ZanKore in Indonesia, which secured financing with Nvidia credit support; Nscale in the UK, which raised capital with Nvidia as an investor; and GMI Cloud in the US, which received equity investment from Nvidia.

That pattern means demand for GPUs is intertwined with vendor-provided financing and investments, creating a highly connected system that depends on continued AI spending to justify the scale of current commitments.

Long-term Treasury yields cooled last week yet growth stocks continued to decline, pointing to a persistent tension: investors chose long-duration bonds yielding over 5% rather than holding into the uncertainty of AI-focused growth names.

The broader market showed rotation toward traditional sectors with more predictable revenue profiles and lower AI infrastructure costs, with healthcare, staples and discretionary seeing renewed investor interest.

The October FOMC meeting approaches and the prevailing view in markets is that rate cuts are unlikely, with expectations tilted toward another rate increase of 25 basis points, a dynamic that places additional pressure on rate-sensitive growth companies.

Separately the authors predict a multi-year horizon before AI delivers broad productivity gains, reflecting the scale and cost of the current infrastructure investment cycle.