英伟达愿景实现：2万亿美元及更多

NVDA’s Vision Coming to Life: $2Tand Counting When tracking AI capacity additions over the LTM, AI TAMsdon't seem so outlandish anymore and NVDA looks like themost interesting name in our group. NVDAOVERWEIGHTUnchanged U.S. Semiconductors &Semiconductor CapitalEquipmentNEUTRALUnchanged When Jensen first forecast a $1T industry by the end of the decade, we admittedly balked. Withthe wave of announcements that have come over the last 6-9mo, we now estimate over $2T ofplanned spend at ~40GW of power in total. Within that, we attribute ~65-70% to compute& networking with more deals likely in the pipeline, which starts to make the updated guidanceof $3-4T look much more real. Given the variation of data available for each project, we usedboth the conversion of 1GW=$50-60B of total spend and used the more recent 1M GPU per 2GWassociated with the OpenAI deal announced last week. When summed, this equates to $1.5T ofcompute & networking spend and 19M GPUs, which we acknowledge isn't perfectly pro-forma.We also acknowledge that some of these dollars will go towards custom silicon but in thetracked announcements thus far there is little reference to any specific programs. We introducean AI capacity tracker that aggregates announced compute deployments, power, and chips thatwe will update real time andofferto clients. We see this as a positive for all accelerator names(AVGO + AMD) but we see this largely flowing into the NVDA P&L over the next 5+ years, movingnumbers materially higher and making this the most attractive name in our space. We move ourprice target to $240. Price TargetUSD 240.00raised 20% from USD 200.00 Price (24-Sep-25)USD 176.97Potential Upside/Downside+35.6%Source: Bloomberg, Barclays Research Market Cap (USD mn)4300371Shares Outstanding (mn)24300.00Free Float (%)96.1052 Wk Avg Daily Volume (mn)232.4Dividend Yield (%)0.02Return on Equity TTM (%)109.42Current BVPS (USD)4.11Source: Bloomberg Price PerformanceExchange-Nasdaq52 Week rangeUSD 184.55-86.62 See our tracker below our colleagues' commentary. Please reach out if you would like to receivethis in excel. Source: IDCLink to Barclays Live for interactive charting U.S. Semiconductors & SemiconductorCapital Equipment Tom O'Malley+1 212 526 0692thomas.omalley@barclays.comBCI, US In Figure 1 below, we show how estimated compute and networking revenue alreadydwarfs hyperscaler RPO and the total announced programs are already trending towards the2030 NVDA TAM. U.S. Internet When we compare these numbers to cloud/AI capex, it's very clear that numbers need to movehigher. Consensus shows 22% growth in 2026E and 12% in 2027E for the 4 major hyperscalers(AMZN, MSFT, GOOGL, ORCL) plus META. These estimates have been creeping higher with eachquarter for the past three years, and we expect more of the same on the back of OpenAI’s recentannouncements around Stargate expansion and the Nvidia funding. Perhaps a better proxy forthe coming AI wave is the growth rate for the 4 major hyperscalers' backlog. This figure wastrending in the 30% range for the past year and popped to 86% comingoffof this last quarter(crossing to north of $1T for the first time ever), mostly a result of ORCL’s backlog spike. In termsof AI adoption, ChatGPT message volumes are up 482% Y/Y in June 2025 and users (WAU) are up250%+ per recent disclosures, a decent proxy for overall usage of AI (see: “How We Are Using AIThree Years In”). IT Hardware and Communications Equipment The surge in compute demand is cascading across the hardware ecosystem, with switching andEMS players poised to benefit the most. Cloud capex spend has moved up significantly with ourmodel estimating over 20% Y/Y growth for 2026 following over 50% growth in 2024 and 2025. Asdemand for advanced backend networks solution has skyrocketed, switching vendor ANET hasemerged as a winner here and we expect that to continue. In addition, CLS is a dominantwhitebox player and counts GOOG, AWS and META as major customers, with OpenAI entering onthe compute side. The AI theme creates a tailwind for the EMS space given the need foroutsourcing liquid cooling, optical transceivers manufacturing, and services such as rackassembly in the data center. Although this is a positive, we note EMS companies have highcustomer concentration risk. As for server, overall revenues move higher for companies likeDELL and SMCI, but earnings growth is limited given the low-margin nature of the business. U.S.Software From asoftwareperspective, the hyperscalers serve as Nvidia's largest customers in this AIbuildout, and the feasibility of this production is of the utmost importance for the hundreds ofbillions of dollars of compute expected to be spent in the coming years.Microsoft,Oracle, andCoreWeave, among others, underpin the vast levels of investment being seen today with thelarge AI training clusters built on top of the data center infrastructure they provide. Incredibledemand here is represented by the hundreds of billions of dollar