Stock Markets June 30, 2026 08:59 AM

Nomura: AI infrastructure build-out likely has further to run despite recent chip pullback

Firm points to data center supply constraints and hyperscaler spending patterns that could extend the investment cycle into 2027

By Ajmal Hussain
Share
Twitter Reddit Facebook LinkedIn

Nomura's latest note argues the AI infrastructure investment cycle has not yet peaked. While semiconductor shares have eased after sharp gains, the bank highlights supply bottlenecks, rising memory costs and continued hyperscaler capital intensity as reasons to expect more upside through 2027. Nomura's data center build tracking and supply-side timelines support a continued demand-supply imbalance, and the firm says it remains a buyer on weakness.

Nomura: AI infrastructure build-out likely has further to run despite recent chip pullback
Summarize with
ChatGPT Perplexity Claude Grok Gemini

Key Points

  • SOX has risen sharply in recent months - about 85% since Nomura's March update and around 211% since the firm's May 2025 reassessment - before a recent pullback; this rapid appreciation explains why the firm views the latest weakness as a healthy correction.
  • Nomura's global data center build tracking and a two-year greenfield construction timeline starting in late 2025 suggest capacity may be insufficient heading into 2027, supporting the case for continued investment in AI infrastructure.
  • Price inflation in memory and potential shifts in supply bottlenecks from large foundries to smaller component makers could force hyperscalers to maintain or increase spending even amid free cash flow pressures, affecting semiconductor, memory, and data center sectors.

Nomura told clients that the current AI infrastructure investment cycle has probably not reached its high point, even after a pullback in semiconductor equities. The bank points to ongoing supply constraints and hyperscaler spending trends as reasons the cycle could extend further into 2027.

In a note distributed on Tuesday, analyst Aaron Jeng highlighted the magnitude of recent gains in chip stocks. The Philadelphia Semiconductor Index (SOX) advanced roughly 85% since Nomura's previous cycle update in March and about 211% since the firm revisited the AI theme in May 2025, before the market reversed some of those gains.

Jeng described the recent sell-off as a healthy correction after a rapid run-up, and he listed several risk factors that could weigh on the group. Those include component supply mismatches, pressure on hyperscalers' free cash flow in 2027, and macroeconomic headwinds tied to a rising yield environment.

Despite these headwinds, Nomura does not view the cycle as having peaked. The bank argues that hyperscaler capital expenditure may need to increase further into 2027, even if free cash flow is strained, a dynamic that is being amplified in part by sharply higher memory prices.

Nomura's internal tracking of global data center builds also points to upside beyond the firm's March estimates. On the supply side, the bank notes that greenfield data center projects that begin construction in late 2025 would generally require around two years to complete. That timeline implies capacity could be insufficient as 2027 approaches.

According to Nomura, the nature of the bottleneck may also shift. Rather than constraints originating primarily with major foundries such as TSMC, limitations may migrate toward smaller component suppliers that feed the broader ecosystem.

Jeng emphasized that pricing moves and ongoing upward earnings revisions would remain the primary catalysts for additional upside in semiconductor and infrastructure-related stocks. He added that Nomura would "still be buyers into weakness," signaling the firm sees continued opportunity despite noted risks.


Context and implications

The note frames the current market action as a pause rather than an end to an investment cycle driven by AI compute demand. Nomura's supply-side timelines and proprietary build tracking underpin its view that demand will outstrip near-term capacity, and that cost pressures - notably in memory - could compel sustained hyperscaler investment despite tighter free cash flow metrics.

Risks

  • Component supply mismatches that could disrupt the timing and scale of infrastructure deployments - this directly affects semiconductor suppliers and component makers.
  • Hyperscalers' potential free cash flow pressures in 2027, which could limit or delay capital spending plans and impact data center construction timelines and equipment purchases.
  • Macroeconomic risk tied to a rising yield trend, which could increase financing costs and weigh on valuation-sensitive technology and infrastructure investments.

More from Stock Markets

Circle Internet Shares Slide After Major Consortium Announces Competing Dollar Stablecoin Jun 30, 2026 Chinese Smartphone Makers Deeply Cut 2026 Volume Targets as Memory Shortage Bites Jun 30, 2026 Oppenheimer Cuts Ratings on Major U.S. Banks, Urges Shift Toward Alternative Asset Managers Jun 30, 2026 Digital Realty Shares Drop After Closing Blackstone Purchase and Following Large Secondary Sale Jun 30, 2026 Morgan Stanley Sees More Room for Post-Conflict Rotation in EEMEA Stocks Jun 30, 2026