Interview, Fireside Chat
Chip stocks retreat but AI spend advances; upcycle has room to run
- Hyperscaler and neocloud spending is projected at approximately $1 trillion for the current year, rising to a range of $1.35 to $1.4 trillion next year, representing 35 to 40 percent growth and a 20 percent increase over forecasts made three months prior.
- The largest hyperscalers anticipate acceleration in core search, e-commerce, and customer acquisition businesses, while new clouds are expected to grow rapidly by servicing overflow demand from public hyperscalers constrained by immediate land, power, and shell capacity.
- High-bandwidth memory demand is forecast to exceed supply by a factor of three to 4x, suggesting the current memory cycle will persist longer than previous iterations, with the broader semiconductor cycle anticipated to last a minimum of another two years.
- AI investment is predicted to contribute 30 to 40 basis points of GDP growth next year and approximately 70 basis points in 2026, potentially constituting the third year of a five-year AI cycle.
- Frontier model pricing is expected to remain strong while consumer offerings may shift to more productive infrastructure, leading to value accrual at the underlying computing infrastructure layer as model-level margins erode.
- Intelligence is expected to increasingly move toward the edge via co-pilot-equipped PCs and consumer devices with potential robotics applications, though bulk reasoning and processing will remain in the network core to avoid unfavorable battery trade-offs on phones.
- A smarter phone ecosystem is expected to generate a virtuous cycle where users rely more heavily on cloud infrastructure, with hyperscalers providing spending trajectory guidance at the start of the new year to determine 2027 and 2028 outlooks.
- Strong cash flow generation is expected to drive increased reliance on buybacks for shareholder returns, provided government regulations do not become more onerous than current levels regarding data center construction or model adoption.
- AI policy frameworks are expected to require support for reducing inflation and the cost of capital to facilitate data center construction, while China continues significant infrastructure investment and robotics adoption to maintain global competition.
- U.S. leadership in AI is considered critical to ensuring technology deployment in Asia and the Middle East is not dictated by the Chinese technology ecosystem.
- New clouds face a financial tension between monetizing existing capacity and funding forward development, a challenge that intensifies if the cost of capital increases.