英伟达高质量增长与估值吸引力，维持买入评级

Durable high-quality growth at compellingvalue multiple, reit Buy Maintain Rating: BUY | PO: 350.00 USD | Price: 195.55 USD Addressing key debates for unique growth franchiseDespite its AIdominance, NVDA stock is up just 3% YTD vs. SOX +82%. We address 4 07 July 2026 key investor concerns: 1) Gross margin pressure from higher memory costs, 2) CustomASIC competition, 3) Crowded investor ownership, and 4) Unproductive use of cash invendor financing vs. stronger buybacks/dividend. Our analysis suggests that at NVDA’scurrent valuation, investors might already be implicitly discounting an unjustified ~30-35% headwind to CY27/28 EPS ests (effectively delta between NVDA and growth peerforward PE).We strongly disagree with the EPS discount and see as an enhanced Buyopp’ty for a unique, durable growth franchise now trading at a 7-yr low 18x forward PE. Equity Vivek AryaResearch AnalystBofASvivek.arya@bofa.com Duksan JangResearch AnalystBofASduksan.jang@bofa.com Michael ManiResearch AnalystBofASmichael.mani@bofa.com Memory inflation predictable, pricing power unappreciatedWe believe investors overstate HBM cost pressure while underestimating NVDA’s pricing power, scale, and $119bn of supply-chain commitments. HBM content per rack may riseby ~$0.2-0.3mn from Blackwell to Rubin, but rack pricing could increase by $2-3mn(from ~$3-4mn to ~$6-7mn), driven by upgrades across compute, networking, andsoftware. We therefore expect gross margins to remain around the mid-70% range. Liam PharrResearch AnalystBofASliam.pharr@bofa.com NVDA GPUsales up ~700x since ASICs launched in 2015 GoogleTPU (2015), Amazon Trainium (2020), and Meta MTIA (2023) have all beenaround, yet NVDA GPU revenue has grown ~700x since 2015. NVDA sales to hyperscalerrose 115% YoY, nearly 2x cloud capex growth, suggesting continued wallet-share gains.Over the long-term, we expect NVDA to sustain a 65-70%+ share of AI capex. Price195.55 USDPrice Objective350.00 USDDate Established20-May-2026Investment OpinionC-1-752-Week Range157.34 USD - 236.54 USDMrkt Val (mn) / Shares Out4,865,284 USD / 24,880.0 Valuation alreadybakes in 30-35% EPS headwind Despite similar AI opportunities and memory-cost pressures as Amazon, Meta, Google,Microsoft, and Apple, NVDA trades ~30-35% below their 22x/19x CY27/28E PE. Weexpect upcoming NVDA earnings to reinforce its moats in products, pricing, and supplychain. While ownership concentration (1.15x weighting, 78% ownership within activeS&P 500 funds) and strategic investments (~$65bn) remain overhangs, we estimate thelatter consumes <35% of FCF, leaving substantial capacity for dividends and buybacks. See glossary on page 8 Timestamp: 07 July 2026 06:44PM EDT iQprofileSMNVIDIA Corporation Company SectorSemiconductors Company Description Nvidia designs and sell graphics and video processing chipsfor desktop and notebookgaming PCs, workstations, gameconsoles, and accelerated computing servers andsupercomputers. Investment Rationale Our positive view on Nvidia is based on its unique full-stackand leadership in artificial intelligence silicon, hardware andsoftware. The company's strong balance sheet and freecash flow returns further enables ecosystem investmentsand enhanced shareholder returns. Stock Data Average Daily Volume166,123,520 Memory inflation predictable, Price power un-appreciatedWe believe investors simultaneously overestimate the GM impact of rising memory input cost, while underestimating NVDA’s pricing power, long-range planning, product co-design, scale and supply chain alignment via over $119bn in prepurchase commitmentsacross memory, wafers, packaging and power. To first order, NVDA high bandwidth memory content per rack goes up by ~$0.2-0.3mnincrementally from Blackwell to Rubin, while product pricing is likely to go up by over $2-3mn (from $3-4mn/rack in Blackwell, to $6-7mn/rack in Vera Rubin) because of theadvancements in multiple other parts of the rack such as an upgraded Vera CPU,upgraded NVLink and Quantum Ethernet networking (none of which need HBM memory),and a host of other software features that speed-up (and lower cost) time to first token. Moreover, we flag NVDA’s comments that Vera Rubin can deliver 10x more performanceper watt (~10x lower inference token cost), 3.3x more powerful on inference, and up to5x more powerful on training than Blackwell. Overall we expect NVDA to maintain its roughly mid-70s GM, blended across its breadthof cloud, enterprise, neocloud, sovereign customers, and across its breadth of productsin the rack-scale architecture. NVDA GPU revs up 700x since ASICs launched in 2015 NVDA has faced competition from custom ASIC since Google first launched its TPU in2015, Amazon Trainium launched in 2020 and Meta MTIA in 2023. Despite thesecompetitions, NVDA GPU accelerator sales have grown by a factor of 700x since 2015.In fact, per its new segment disclosure, NVDA sales to hyperscalers surged 115% YoY,nearly 2x the growth in cloud capex, suggesting growi