Bank of America reiterates a Buy rating on NVIDIA with a $350 target price

Overview

The report focuses on the core investment debate around NVIDIA: whether HBM memory costs will squeeze gross margin, whether custom ASICs will weaken GPU demand, whether positioning is too crowded, and whether supplier and customer ecosystem investments are weighing on cash returns. Bank of America concludes these concerns are already over-reflected in the stock price, while NVIDIA’s combined strengths across AI compute, networking, software, supply-chain commitments, and customer ecosystem support high-quality, sustainable growth.

Core views

Bank of America reiterates a BUY on NVIDIA and believes the current valuation implies an unreasonable 30-35% discount to CY27/28 EPS. The report expects NVIDIA gross margin to remain around 75%, and long-term AI capex share to stay above 65-70%; the performance, efficiency, and rack-level value gains of Vera Rubin versus Blackwell are expected to offset higher HBM costs.

Analysis framework

The report uses issue decomposition, peer valuation comparison, AI data center TAM estimation, a historical review of GPU versus ASIC competition, and free cash flow and ecosystem investment resilience analysis to assess whether NVIDIA’s earnings sustainability and valuation discount are justified.

Methodology notes

• Valuation methodsforward pe valuation

  target price approach based on CY27E P/E

  The 350 USD target price is based on a 26x CY27E P/E excluding cash, within NVIDIA’s historical forward P/E range of 25x to 56x.

• market sizetam analysis

  AI data center systems TAM estimation

  The report expects AI data center systems TAM to grow from 273 billion USD in CY25 to about 1.7 trillion USD by CY30, with AI servers around 77%, networking around 18%, and storage around 5%.

• competitive analysisshare and wallet analysis

  GPU versus ASIC share comparison

  The report reviews market performance after launches of Google TPU, Amazon Trainium, and Meta MTIA, and notes NVIDIA GPU revenue has still grown about 700x, indicating platform breadth and customer wallet share advantage.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• NVIDIA Corporation (NVDA.US)

  Core research target; benefits from continued growth in demand for AI GPUs, AI data center systems, and networking interconnects.

  Strengths

  Full-stack AI hardware and software platform, GPU ecosystem, supplier prepayment commitments, pricing power, scale advantage, and broad customer coverage.

  Weaknesses

  The market remains concerned about HBM costs, ASIC competition, crowded positioning, and the efficiency of ecosystem investments.

  Comparison

  Relative to peers such as Amazon, Meta, Google, Microsoft, and Apple, NVIDIA has an estimated 30-35% discount on CY27/28E PE; versus ASIC solutions, NVIDIA has a more general and broadly supported platform advantage.

  Risks

  Weak gaming demand, intensified AI and accelerator competition, Chinese export restrictions, sales volatility in enterprise and data-center channels, slower capital returns, and potential regulatory scrutiny of NVIDIA’s AI chip leadership.

• META PLATFORMS INC (META.US)

  Used as a peer and customer-ecosystem reference in the context of AI capex and internal ASIC competition.

  Strengths

  Has large-scale AI investment and in-house MTIA capability.

  Weaknesses

  The report does not provide a standalone investment rating analysis for META.

  Comparison

  Meta MTIA is cited to show that custom ASICs have existed for a long time, yet they have not prevented substantial growth in NVIDIA GPU revenue.

  Risks

  If large cloud and internet customers accelerate internal ASIC substitution, NVIDIA’s incremental demand could be affected.

Key data

• Investment ratingBUYMaintains Buy.
• Target price350.00 USDBased on 26x CY27E P/E ex cash.
• Report price195.55 USDPrice disclosed in report.
• Implied upsideabout 79.0%Calculated using target 350.00 USD versus price 195.55 USD.
• CY27E EPS9.09 USDBank of America estimate.
• CY28E EPS13.27 USDBank of America estimate.
• CY30E EPS upside25+ USDBased on AI data center systems TAM around 1.7 trillion USD and an assumed NVIDIA share of roughly 70%+.
• Forward P/E18x forward PEThe report describes it as near a 7-year low.
• CY27/28E valuation discountabout 30-35%Compared with large-cap peers including Amazon, Meta, Google, Microsoft, and Apple.
• HBM cost incrementabout 0.2-0.3 million USD per rackEstimated increase from Blackwell to Rubin.
• Rack price upliftabout 2-3 million USD per rackBlackwell about 3-4 million USD per rack, Vera Rubin about 6-7 million USD per rack.
• Long-term AI capex share65-70%+NVIDIA share Bank of America expects to maintain long term.
• Ecosystem investment scaleabout 65bn USDAbout 35% of CY26E FCF and 17% of CY27E FCF.

Impact & implications

If Bank of America’s view is correct, NVIDIA’s current valuation discount is more likely an excessive market penalty for cost, competition, and crowded-positioning risks than a structural deterioration in profitability. Upcoming earnings, the Vera Rubin product cycle, AI data center TAM expansion, and ability to return capital may become catalysts to close the valuation gap.

Risks

• Weakness in consumer-driven gaming demand.
• Competition in AI and accelerator markets from large listed companies, hyperscaler internal programs, and private firms.
• Underestimation of the impact of export shipment restrictions on China and potential further limitations on regional activity.
• Variable and less predictable sales cadence in enterprise, data center, and autos/adjacent new markets.
• Capital returns may slow.
• Increased government regulatory scrutiny of NVIDIA’s dominant position in AI chips.
• High overweight in institutional ownership and S&P 500 active funds could become near-term supply-demand pressure.
• If investments by suppliers and customer ecosystems do not deliver, free cash flow allocation efficiency could be impaired.

What to watch

• Whether NVIDIA’s next earnings call reinforces product, pricing, and supply-chain moats.
• Performance, efficiency, rack pricing, and HBM cost pass-through from Vera Rubin versus Blackwell.
• Whether HBM cost share of accelerator spending remains in the roughly 20-23% range.
• The gap between hyperscaler capex growth and NVIDIA’s sales growth to hyperscalers.
• Share shifts of alternatives such as Google TPU, Amazon Trainium, Meta MTIA, and AMD.
• Whether CY30 AI data center systems TAM follows the roughly 1.7 trillion USD trajectory.
• NVIDIA’s ecosystem investment scale, returns, and effect on dividend-and-buyback capacity.
• Marginal changes in China export controls and regulatory scrutiny.