NVIDIA's results and FY28 growth framework are strong; memory costs lower the gross-margin trajectory, but the long-term demand foundation remains intact

Overview

The report focuses on NVIDIA's F2Q27 results, F3Q27 guidance, its first FY28 revenue framework, the Vera Rubin product ramp, memory-cost pressures, and AI infrastructure financing arrangements. JPMorgan believes demand and the product cycle are stronger than previously assumed. Although gross margins face near-term pressure, purchase commitments provide support, leading the firm to raise its earnings forecasts and price target while maintaining its Overweight rating.

Core views

NVIDIA's F2Q27 revenue was USD 962 × 100 million, up 18% quarter over quarter and 106% year over year, marking the fourth consecutive quarter of accelerating year-over-year growth and exceeding the market expectation of USD 921 × 100 million. GAAP EPS was USD 2.22, above the market expectation of USD 2.09; GAAP gross margin was 75.0%, also slightly above the market expectation of 74.8%. Data Center revenue reached USD 890 × 100 million, up 18% quarter over quarter and 117% year over year, exceeding the market expectation of USD 873 × 100 million and representing the primary source of the earnings beat. Demand within Data Center was broad-based. Hyperscale revenue was USD 487 × 100 million, up 13% quarter over quarter and 102% year over year, primarily driven by continued demand for Blackwell Ultra; ACIE revenue was USD 403 × 100 million, up 25% quarter over quarter and 138% year over year, driven by AI-native clouds, enterprises, sovereign customers, and hyperscale customers supplementing their internally built capacity. Sovereign AI revenue increased 35% quarter over quarter and more than tripled year over year, raising international revenue to 38% of total revenue; China Data Center compute revenue represented less than 1% of Data Center revenue. Networking revenue increased 18% quarter over quarter, including a 2.6-fold year-over-year increase in Spectrum X Ethernet revenue; Edge Computing revenue was USD 72 × 100 million, up 13% quarter over quarter and 27% year over year, driven by Blackwell workstation sales. The midpoint of F3Q27 revenue guidance was USD 1,080 × 100 million, significantly above the market expectation of USD 1,045 × 100 million, with sequential growth primarily driven by AI-native customers. Management expects the Hyperscale business to reaccelerate in F4Q27 and FY28 as Vera Rubin supply expands. Vera Rubin began volume production shipments during the report month and has received purchase orders from all major hyperscale clouds, AI clouds, and system OEMs; it is expected to contribute approximately 20% of F3Q27 Data Center revenue. Management called it "the fastest product ramp in NVIDIA's history," while stating that tail-end demand for Blackwell is narrowing. JPMorgan believes this progress alleviates previous concerns about the architecture transition and is key to reducing execution risk for the FY28 framework. Management provided its first formal forward-looking FY28 revenue framework, calling for approximately 70% year-over-year growth; JPMorgan estimates FY28 revenue at approximately USD 6,950 × 100 million and believes the framework may still be conservative because management explicitly stated that revenue is supply constrained and that unconstrained demand growth is significantly higher. The growth thesis is primarily based on agentic inference expanding the total addressable market, global sovereign AI momentum, and continued release of supply-chain capacity. The hyperscale cloud industry's backlog has exceeded USD 2 trillion; revenue per gigawatt is expected to increase from USD 180 × 100 million for Hopper and USD 250 × 100 million for Grace Blackwell to more than USD 400 × 100 million for Vera Rubin; as Vera Rubin scales, FY28 CPU revenue is expected to more than double. Gross margin is the report's primary near-term pressure point. F3Q27 gross-margin guidance is 74.0%, above the market expectation of approximately 73.5%, but down roughly 100 basis points sequentially because memory-cost inflation exceeded previous assumptions. Management lowered the broader gross-margin trajectory, expecting a trough of 71%—72% in F4Q27, followed by stabilization at 72%—73% in FY28 as implemented price increases begin to take effect in F1Q28. NVIDIA has locked in most of its FY28 HBM requirements at known prices; most of the company's approximately USD 3,500 × 100 million in total supply commitments relates to memory, primarily HBM, and covers the next three years. JPMorgan believes these long-term commitments provide a floor for next year's gross-margin trajectory, but do not eliminate pressure from memory costs and supply on specifications, product mix, and profitability. NVIDIA is also expanding the financing architecture supporting future AI compute demand. Its USD 1,050 × 100 million PORTS-Pike guarantee supports the construction of approximately 4.25GW of IT load for OpenAI in Ohio and includes a right of first negotiation for an additional approximately 3.8GW; the report estimates that each product generation could correspond to approximately 1.5 million GPUs and USD 1,500 × 100 million—USD 2,000 × 100 million in NVIDIA revenue across multiple 20-year upgrade cycles. For certain emerging cloud service providers, NVIDIA supports base-price cloud service agreements and receives 50% of the difference between the base price and the actual market lease price, thereby creating usage-linked recurring revenue beyond hardware sales; approximately 5—20 such transactions are currently in the pipeline and could contribute billions of dollars in incremental revenue over the long term. In addition, the company has established a USD 5,000 × 100 million private-capital financing platform with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to support infrastructure construction for frontier AI laboratories. NVIDIA's residual credit support is capped at approximately 25% of each financing, and no capital has yet been committed under the framework. The company has invested nearly USD 500 × 100 million in frontier AI laboratories to date; OpenAI has committed to large-scale deployment of NVIDIA AI infrastructure through 2030, totaling approximately 12GW; AI laboratories supported by NVIDIA's balance sheet are expected to contribute approximately one-quarter of the business in FY28. Management believes the related support is commensurate with demand intensity, ecosystem returns, and the investment-grade credit quality of the underlying compute buyers. JPMorgan acknowledges mitigating factors, including the replaceability and redeployability of compute infrastructure, phased and capped guarantees, and NVIDIA's limited residual role in the financing hierarchy, but expects "circular financing" to remain an ongoing area of debate. The report also emphasizes that NVIDIA's vertically integrated platform spanning chips, rack systems, and the software stack is difficult to replicate and provides a competitive advantage over other AI compute vendors; the migration of traditional enterprise workloads to accelerated computing also provides structural demand independent of AI training and inference cycles. The company returned a record USD 260 × 100 million to shareholders this quarter, including USD 200 × 100 million in repurchases and USD 60 × 100 million in dividends, bringing the year-to-date free-cash-flow return ratio to 60%, above its "more than 50%" plan. Quarterly operating cash flow was USD 241 × 100 million, down from USD 503 × 100 million in the prior quarter; period-end cash, cash equivalents, and marketable securities were USD 994 × 100 million, up from USD 806 × 100 million in the prior quarter. JPMorgan raised its FY27 adjusted EPS forecast by 5.8%, from USD 8.73 to USD 9.24. Its FY26A, FY27E, and FY28E revenue forecasts are USD 2,159.38 × 100 million, USD 4,087.00 × 100 million, and USD 6,950.41 × 100 million, respectively, while adjusted EPS forecasts are USD 4.77, USD 9.24, and USD 15.87, with corresponding year-over-year revenue growth of 65.5%, 89.3%, and 70.1%. The corresponding adjusted P/E multiples are 44.0x, 22.7x, and 13.2x, while EV/EBITDA multiples are 37.8x, 18.4x, and 10.5x. The price target applies an approximately 20x valuation multiple to what the report describes as CY26 EPS of USD 15.87, raising the Dec-27 price target to USD 320 from the previous Dec-26 target of USD 280 and maintaining the Overweight rating. The report also cautions that AI ASICs/XPUs, AMD, and specialized accelerators may continue gaining share in the AI compute market, and that the market shares of GPUs and ASICs/XPUs may eventually converge.

Analysis framework

The report first compares quarterly revenue, EPS, gross margin, and the performance of each Data Center subsegment with market consensus expectations, then assesses the product ramp and demand durability based on F3Q27 guidance, Vera Rubin shipments, and customer orders. It subsequently builds its FY28 revenue forecast using supply-demand constraints, revenue per gigawatt, cloud-industry backlog, and CPU growth, while projecting the gross-margin trajectory by incorporating HBM purchase commitments and the timing of price increases. Finally, the report analyzes how infrastructure financing may unlock demand and generate incremental revenue, applies a P/E valuation to the updated earnings forecasts to derive the price target, and discusses risks including alternative computing platforms, memory costs, and financing controversies.

Methodology notes

• Valuation MethodPE/PEG valuation

  Forward P/E Price-Target Method

  The report applies an approximately 20x valuation multiple to what it describes as CY26 EPS of USD 15.87 to support a Dec-27 price target of USD 320; the multiple is described as reflecting long-term growth expectations.

• Industry/Sector Analysis FrameworkSupply-demand framework

  Supply-Constrained Revenue Framework

  The report treats approximately 70% FY28 growth as the revenue floor achievable under available supply and evaluates actual growth potential and the gross-margin trajectory using unmet demand, supply-chain expansion, HBM purchase commitments, and the pace of product shipments.

• (Out-of-Vocabulary Method)

  Results Versus Consensus Analysis

  The report compares actual revenue, EPS, gross margin, and next-quarter guidance with market consensus expectations item by item, using the magnitude of the beats to identify demand strength and the basis for forecast increases.

• Quantitative/Factor/Portfolio TheoryStyle factor analysis

  Quantitative Style Exposures and Correlations

  The report includes rankings for value, growth, momentum, quality, low-volatility, and ESG quality factors, as well as the stock's historical correlations with variables such as the US market, technology sector, semiconductor industry, and US 10-year yield, as supplementary descriptions of style and risk characteristics.

Asset mapping & comparison

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

• NVIDIA Corporation (NVDA)

  The report believes the company benefits from continued Blackwell Ultra demand, the rapid Vera Rubin ramp, expansion in agentic inference, sovereign AI construction, and demand unlocked by AI infrastructure financing.

  Strengths

  Its vertically integrated platform spanning chips, rack systems, and the software stack is difficult to replicate; it has broad customer orders, multi-year demand visibility, strong cash resources, and most of its FY28 HBM requirements already secured.

  Weaknesses

  Near-term gross margins are affected by memory-cost inflation, while some future demand requires guarantees, capital support, or revenue-sharing arrangements to be unlocked.

  Comparison

  The report believes NVIDIA's platform-integration capabilities provide an advantage over AI compute peers, but expects the shares of GPUs and ASICs/XPUs in the overall AI compute market to potentially converge over time.

  Risks

  Alternative ASICs/XPUs, AMD, and specialized accelerators gaining market share; memory costs and supply constraints affecting specifications and profitability; and circular-financing controversy arising from balance-sheet support.

Key data

• F2Q27 RevenueUSD 962 × 100 millionUp 18% quarter over quarter and 106% year over year, above the market expectation of USD 921 × 100 million
• F2Q27 GAAP EPSUSD 2.22Above the market expectation of USD 2.09
• F2Q27 GAAP Gross Margin75.0%Above the market expectation of 74.8%
• Data Center RevenueUSD 890 × 100 millionUp 18% quarter over quarter and 117% year over year, above the market expectation of USD 873 × 100 million
• Hyperscale RevenueUSD 487 × 100 millionUp 13% quarter over quarter and 102% year over year
• ACIE RevenueUSD 403 × 100 millionUp 25% quarter over quarter and 138% year over year
• Sovereign AI and International RevenueSovereign AI increased 35% quarter over quarter and more than 3-fold year over year; international revenue accounted for 38%Demand from emerging regional cloud service providers was an important driver
• China Data Center Compute RevenueLess than 1% of Data Center revenueF3Q27 guidance continues to assume zero revenue from this item
• F3Q27 Revenue GuidanceMidpoint of USD 1,080 × 100 millionAbove the market expectation of USD 1,045 × 100 million
• Vera Rubin ContributionApproximately 20% of F3Q27 Data Center revenueVolume production shipments have begun, and management described it as the fastest product ramp in the company's history
• FY28 Revenue FrameworkApproximately 70% year-over-year growthSupply constrained; JPMorgan estimates approximately USD 6,950 × 100 million
• Revenue per GigawattMore than USD 400 × 100 million for Vera RubinUSD 250 × 100 million for Grace Blackwell and USD 180 × 100 million for Hopper
• Gross-Margin Trajectory74.0% in F3Q27; trough of 71%—72% in F4Q27; 72%—73% in FY28Memory-cost inflation creates pressure, with price increases gradually taking effect from F1Q28
• Supply CommitmentsApproximately USD 3,500 × 100 millionMost relates to memory, primarily HBM, and covers the next three years
• PORTS-Pike GuaranteeUSD 1,050 × 100 millionSupports approximately 4.25GW of construction and includes a right of first negotiation for an additional approximately 3.8GW; each generation could correspond to approximately 1.5 million GPUs and USD 1,500 × 100 million—USD 2,000 × 100 million in revenue
• Private-Capital Financing PlatformUSD 5,000 × 100 millionNVIDIA's residual credit support is capped at approximately 25% of each financing, and no capital has yet been committed
• Investment in Frontier AI LaboratoriesNearly USD 500 × 100 millionThe related laboratories are expected to contribute approximately one-quarter of FY28 business
• OpenAI Deployment CommitmentApproximately 12GW in total through 2030Committed to large-scale deployment of NVIDIA AI infrastructure
• Shareholder ReturnsUSD 260 × 100 millionIncluding USD 200 × 100 million in repurchases and USD 60 × 100 million in dividends; the year-to-date free-cash-flow return ratio was 60%
• Cash and Operating Cash FlowPeriod-end cash and related assets of USD 994 × 100 million; quarterly operating cash flow of USD 241 × 100 millionUSD 806 × 100 million and USD 503 × 100 million, respectively, in the prior quarter
• FY27 Adjusted EPS RevisionUSD 9.24Previously USD 8.73, increased by 5.8%
• FY26A/FY27E/FY28E RevenueUSD 2,159.38 × 100 million / USD 4,087.00 × 100 million / USD 6,950.41 × 100 millionCorresponding year-over-year growth of 65.5% / 89.3% / 70.1%
• FY26A/FY27E/FY28E Adjusted EPSUSD 4.77 / USD 9.24 / USD 15.87Corresponding growth of 59.3% / 93.8% / 71.7%
• FY26A/FY27E/FY28E ValuationAdjusted P/E of 44.0x / 22.7x / 13.2x; EV/EBITDA of 37.8x / 18.4x / 10.5xBased on the financial forecasts presented in the report

Impact & implications

JPMorgan believes the quarterly beat, Vera Rubin's early entry into a rapid ramp, and the supply-constrained FY28 framework collectively improve visibility into future revenue growth. Locked-in HBM pricing commitments and implemented price increases support a gross-margin recovery after a trough in F4Q27, while financing platforms, guarantees, and revenue sharing may help AI customers with weaker balance sheets expand construction and generate recurring revenue beyond hardware sales. However, growth is becoming increasingly dependent on large AI infrastructure projects, financing arrangements, and continued supply expansion, while memory costs and rising share for alternative computing platforms could still affect profitability and the competitive landscape.

Risks

• AI ASICs/XPUs, AMD, and specialized accelerators may continue gaining share in the AI compute market, and the report expects GPU and ASIC/XPU shares to potentially converge over time.
• Memory-cost inflation and supply constraints may continue to compress platform specifications and create uncertainty regarding long-term gross margins, revenue mix, and profitability.
• NVIDIA's investments, guarantees, and credit support for AI laboratories and infrastructure projects may continue to generate "circular financing" controversy, although the report notes that the support is phased and capped and that the underlying assets can be redeployed.

What to watch

• Monitor Vera Rubin volume production shipments and conversion of customer orders, whether its share of F3Q27 Data Center revenue reaches approximately 20%, and whether the Hyperscale business reaccelerates in F4Q27 and FY28.
• Monitor whether gross margin troughs within the 71%—72% range in F4Q27 and recovers to 72%—73% in FY28 after price increases take effect in F1Q28.
• Monitor the completion of approximately 5—20 revenue-sharing transactions, actual capital deployment by the USD 5,000 × 100 million financing platform, and the scale of NVIDIA's residual credit support.
• Monitor whether agentic inference, sovereign AI, and hyperscale cloud demand can support the FY28 revenue framework of approximately 70% growth, and whether incremental supply-chain capacity can unlock higher unconstrained demand.
• Monitor the assumption of zero China Data Center compute revenue in F3Q27 and the market's impact on subsequent Data Center revenue.