高级节奏：人人都正在采用人工智能

The High-Grade Beat Everyone’s an AI Adopter CITI'S TAKE Daniel SoridAC+1-212-723-1992daniel.sorid@citi.com Mathew JacobAC+1-212-816-1547mathew.jacob@citi.com With thanks toVictor Attah Tech is no longer one trade —Idiosyncratic dispersion within TMT has surged to its54th percentile, the highest of any sector and roughly 18 percentage points abovethe next. With Semis and Hardware sitting near their tightest levels while Softwarehas widened almost 24bps since January, AI is shaping the sector in vastly differentways. The days of treating Tech as a monolithic allocation are over. The AI footprint keeps growing —A back-of-the-envelope exercise in trying toestimate how much of the index has revenue exposed to the AI buildout suggeststhat at least 40% of the index has some exposure. But the reality is that the lines arebecoming blurrier between companies with and without AI exposure. From carmanufacturers repurposing EV batteries to insurance brokers underwriting datacenter construction, non-traditional players are finding creative ways to capture the What goes up, can also come down —While non-Tech issuers are improving theirfundamentalsby tapping into AI-related revenue,they are also becomingincreasingly dependent on the buildout's continuation. As more companies acrossthe index tie their credit stories to AI spending, index holders are accumulatingexposure to the technology that extends well beyond their Tech allocation. Rich and cheap look different through an AI lens —We introduce 7 groupings thathighlight divergence in valuations stemming from AI: Semiconductors, Hardware,Software,Technology Services,AI&Power Generation,Physical AI,andHyperscalers. These 7 groups cut through traditional sector lines to show thedifferent ways AI is impacting pricing. While most of these groups trade close to 3-year tights, Hyperscalers, Software, and Tech Services sit notably wider. As the AI IG Credit Dashboard —We also include our compendium of unique credit marketvisualizationsand insights across supply,fundamentals,relative value,andregression-implied fair values – a full suite of tools to identify value in the U.S. IGcredit market. RELATED: Artificial Intelligence:Citi’s Inference Ahead - Open v. Closed See Appendix A-1 for Analyst Certification, Important Disclosures and Research Analyst Affiliations. Worsening sentiment on the impactof AI © 2026 Citigroup Inc. No redistribution without Citigroup’s written permission.Our August poll asked readers if they thought GenAI would improve, weaken, or maintain US IG issuer credit quality over the next 36 months. October, November, and June polls asked readers theirthoughts on whether the risks of credit quality degradation from GenAI developments outweighedthe opportunities, were in balance with them, or if the opportunities outweighed the risks. Forcomparative purposes, similar positive, negative, and neutral responses are grouped together. Source: Citi Research As the AI debt binge marches forward, reader sentiment on Gen AI’s impact on IGcredit has soured dramatically. When we first polled readers in August 2025,almost half (47%) expected a net positive credit impact from generative AI. Today,that view has completely flipped, with 47% now expecting credit degradation. Thissentiment reversal is also reflected in spreads. US IG Tech spreads widened from We believe the shift is driven by two key factors: the unprecedented AI debt bingeand the increasing concern of AI disruption. After US IG tech companies issued$262bn over the 32-month period spanning from January 2023 to August 2025,they have since issued a staggering $362bn, driven by large-scale issuance fromHyperscalers. As a result, the Hyperscaler share of the Tech IG universe has surgedfrom roughly 28% to 40% by face value. This balance sheet risk is compounded bythe threat of technological disruption. The pace of AI advancement is accelerating;as our colleagues noted in Artificial Intelligence - Citi’s Inference Ahead - Rationing How exposed is US IG to the AI boom The proliferation of spending on AI has rapidly taken over and reshaped the pricingdynamics of a large portion of the index. This is partially structural, with the surgein issuance from Tech increasing the sector’s relative share of the index over thepast year to roughly 11.5% of the index (Figure 2). The focus on AI has touchednearly every corner of pricing for TMT in drastically different ways depending on how the issuer’s business model relates to AI. Figure 3 shows that since late 2025,idiosyncratic dispersion within TMT bond spreads has increased notably from the20thpercentile (looking back since 1998) to the 54thpercentile. This level ofdispersion is notably higher than every other sector, with the next highest sector, The divergence reflects the market's evolving assessment of AI-related creditrisk.Issuers with direct exposure to AI infrastructure buildout (chips, hardware)have benefited from strong demand and tightening spread