GB300 Volume Ramp and Vera Rubin Introduction Could Drive Another NVIDIA Beat, with China H200 Shipments Providing Additional Upside

Overview

This report previews NVIDIA's F2Q27 results and F3Q27 guidance. JPMorgan believes the GB300 rack ramp, the on-schedule introduction of Vera Rubin, and potential H200 shipments to China will support another beat, but the share-price reaction may depend more on how management explains medium- to long-term issues such as competition, infrastructure financing, memory supply, and the China business.

Core views

JPMorgan's supply-chain checks indicate that GB300 rack shipments continued to grow in F2Q27, the July quarter, while tail-end GB200 shipments persisted, with total rack volumes expected to rise approximately 15% quarter over quarter to 17,000–18,000 units. Following unexpected strength in the April quarter, networking growth may return this quarter to a low-to-high-teens quarter-over-quarter rate more closely aligned with the compute business. Edge businesses are expected to grow by mid-single digits to high-single digits quarter over quarter, corresponding to approximately 20% year-over-year revenue growth, below the 29% recorded in F1Q27. Combining these factors, the report expects total revenue could reach $94 billion to $95 billion, representing approximately 15% quarter-over-quarter growth and exceeding the market consensus of $92.1 billion. For F3Q27, the October quarter, the report expects rack shipments to maintain a similar pace, increasing 13%–14% quarter over quarter to 19,000–20,000 units, including the first 1,000–2,000 Vera Rubin racks. As Vera Rubin enters the product mix, the blended average selling price is expected to increase by 5%–10%, while compute revenue could grow by approximately the low-to-high teens quarter over quarter; networking is expected to grow at a similar rate. If some H200 shipments to China are also included, JPMorgan believes revenue guidance could reach $107 billion to $108 billion, approximately 3%, or $2.5 billion to $3.5 billion, above the market consensus of about $104.5 billion. Regarding gross margin, the report sees little risk in management reiterating a mid-70% level, although significant memory cost inflation raises questions about whether this level can be sustained over the long term. The report also emphasizes that near-term results and guidance exceeding expectations alone may not serve as a meaningful share-price catalyst. Over the past four quarters, NVIDIA's revenue guidance has exceeded market expectations by an average of 4%, but its share price subsequently declined by an average of 3% over 7 days and 5% over 30 days. Therefore, the key issue for this earnings call is not merely the magnitude of the beat, but management's commentary on the competitive landscape, the China business, infrastructure financing, and changes in memory specifications. Regarding competition in AI computing, the growing share of ASICs, AMD, Cerebras, and other specialized accelerated-computing platforms continues to pressure NVIDIA's market narrative. Moreover, multiple alternative-platform projects will enter volume production over the coming years, making this concern difficult to eliminate completely. The report expects management to emphasize the flexibility of NVIDIA's platform, particularly its adaptability to the rapidly growing enterprise and sovereign AI markets, as well as the reduction in cost per token enabled by generational product upgrades. JPMorgan estimates that Vera Rubin could reduce platform cost per token by approximately 90% relative to Blackwell Ultra, supporting customer adoption and NVIDIA's ability to maintain its leadership in the AI computing market. Nevertheless, the report still expects the shares of GPUs and ASICs/XPUs in the overall addressable market for AI computing to gradually converge over the next several years. Regarding the recently announced large infrastructure financing agreements, the report expects management to characterize them as facilitating infrastructure deployment rather than artificially creating demand. Its view is that underlying AI demand already exists and that NVIDIA is helping customers address power, data center capacity, and financing bottlenecks through limited, asset-backed arrangements that bring in third-party capital. The China business is an important upside variable for October-quarter guidance and subsequent forecasts. H200 shipments to China were approved some time ago, but the ramp has been slow; recent reports indicate initial customer deliveries. JPMorgan estimates that every 100,000 H200 units shipped could generate approximately $3 billion in incremental revenue, meaning actual shipment volumes could materially alter revenue forecasts for the October quarter and beyond. Adjustments to HBM and SOCAMM specifications are another key issue. Research by JPMorgan's global team indicates that Rubin may use two SKUs configured with either 288GB or 192GB of HBM4; Rubin Ultra's configuration has been adjusted from 16-high stacked HBM4E to 8-high or 12-high stacked HBM4E; and the SOCAMM capacity of the Vera CPU has been reduced by 50%, from the previous 1.5TB per unit to 768GB. The report believes these changes are driven almost entirely by tight memory supply: to deliver the volume of compute products required by customers, NVIDIA has had to reduce memory capacity per system. The ultimate mix of the different SKUs remains unclear, while the impact of specification changes on revenue growth and margins is expected to be a major topic during the earnings call. Regarding valuation, the report believes NVIDIA's valuation multiples have fallen well below normal and historical levels. Based on market forecasts for CY27 and CY28 earnings per share, the stock currently trades at approximately 17x and 13x P/E, respectively. Meanwhile, steadily rising earnings-per-share expectations have continued to support the share price; JPMorgan expects this pattern to persist as earnings continue to expand. Its $280 price target is based on NVIDIA trading at 28x P/E and applying an annualized earnings power of approximately $10 as the company exits CY26. On this basis, the report reiterates its Overweight rating and $280 price target.

Analysis framework

JPMorgan first uses supply-chain checks to estimate rack shipments for GB300, GB200, and Vera Rubin, then combines growth rates in compute, networking, and edge businesses with the average selling price uplift from Vera Rubin to derive quarterly revenue and next-quarter guidance. The report then uses guidance beats and share-price performance over the past four quarters to demonstrate that near-term results may not be the primary catalyst, before further analyzing the effects of competition, infrastructure financing, China H200 shipments, memory specification changes, and valuation on the share price.

Methodology notes

• Industry/Sector Analysis FrameworkVolume-price decomposition

  Decomposition of rack shipment volumes and blended average selling prices

  The report separately estimates rack shipment volumes and the 5%–10% increase in average selling price after Vera Rubin enters the product mix, then uses these estimates to assess growth in the compute business and total revenue.

• Industry/Sector Analysis FrameworkSupply-demand framework

  Supply-chain checks and memory supply constraint analysis

  The report uses supply-chain checks to assess the shipment pace of GB300 and Vera Rubin and attributes the reductions in HBM and SOCAMM specifications to tight memory supply, thereby analyzing the effects of supply constraints on shipments, revenue, and margins.

• Competition and Strategy FrameworkMoat / competitive advantage

  Platform flexibility and cost-per-token advantage

  The report explains NVIDIA's ability to sustain customer adoption and its leadership in AI computing through the breadth of its platform applicability and generational reductions in product costs, while also assessing competitive pressure from ASICs, XPUs, and other specialized acceleration platforms.

• Valuation MethodPE/PEG valuation

  P/E-based price target using annualized earnings power upon exiting CY26

  The $280 price target applies a 28x P/E multiple to approximately $10 of annualized earnings power upon exiting CY26; the report also compares market-implied CY27 and CY28 P/E multiples of approximately 17x and 13x with historical levels.

• Event Trading and Behavioral FinanceEvent-driven analysis

  Share-price reactions following historical guidance beats

  The report compares the magnitude of revenue guidance beats over the past four quarters with the subsequent average 7-day and 30-day share-price performance to determine whether an earnings beat alone would be sufficient to serve as a catalyst this time.

Asset mapping & comparison

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

• NVIDIA Corporation (NVDA.US)

  The report's core research subject, expected to benefit from the GB300 volume ramp, Vera Rubin introduction, AI computing demand, and potential H200 shipments to China.

  Strengths

  Leadership in the AI computing market, platform flexibility, applicability to enterprise and sovereign AI, and an approximately 90% reduction in Vera Rubin's cost per token relative to Blackwell Ultra.

  Weaknesses

  Faces competition from ASICs, XPUs, AMD, Cerebras, and others, while tight memory supply has forced specification reductions for certain Rubin, Rubin Ultra, and Vera CPU products.

  Comparison

  The report expects NVIDIA to maintain its leadership in AI computing, but the shares of GPUs and ASICs/XPUs in the overall addressable market for AI computing may gradually converge over the next several years.

  Risks

  Weak PC gaming demand, reduced deep-learning adoption or GPU deployments by hyperscale customers, and intensifying competition could all cause revenue and earnings to fall below expectations.

Key data

• F2Q27 rack shipments17,000–18,000 unitsExpected to increase approximately 15% quarter over quarter, including the GB300 ramp and tail-end GB200 shipments
• Expected F2Q27 revenue$94 billion–$95 billionExpected to increase approximately 15% quarter over quarter, above the market consensus of $92.1 billion
• Edge business revenue growthMid-single digits to high-single digits quarter over quarter, approximately 20% year over yearYear-over-year growth is below the 29% recorded in F1Q27
• F3Q27 rack shipments19,000–20,000 unitsExpected to increase 13%–14% quarter over quarter
• Initial Vera Rubin shipments1,000–2,000 racksExpected to begin shipping in F3Q27, the October quarter
• Vera Rubin blended average selling price impactIncrease of 5%–10%Estimated contribution to blended ASP after entering the product mix
• Expected F3Q27 revenue guidance$107 billion–$108 billionApproximately 3%, or $2.5 billion–$3.5 billion, above the consensus of about $104.5 billion
• Historical magnitude of guidance beatsAverage of 4%Revenue guidance relative to market consensus over the past four quarters
• Average share-price performance after guidance7 days: -3%; 30 days: -5%Average performance following revenue guidance announcements over the past four quarters
• Vera Rubin cost-per-token improvementApproximately 90% reductionEstimated platform cost relative to Blackwell Ultra
• China H200 revenue sensitivityApproximately $3 billion in incremental revenue per 100,000 unitsJPMorgan's estimate for H200 shipments to China
• Rubin HBM4 configuration288GB or 192GBTwo SKU specifications estimated by the report
• Rubin Ultra HBM4E configurationAdjusted from 16-high to 8-high or 12-highThe report believes the adjustment is primarily due to tight memory supply
• Vera CPU SOCAMM capacity768GB50% lower than the previous 1.5TB specification
• Market-implied P/E multiplesApproximately 17x for CY27; approximately 13x for CY28The report believes these are well below normal and historical valuation levels
• Price-target valuation assumptions28x × approximately $10A 28x P/E multiple applied to approximately $10 of annualized earnings power upon exiting CY26

Impact & implications

The report believes the GB300 volume ramp and Vera Rubin introduction provide high visibility into another near-term revenue beat, while China H200 shipments could further raise guidance. However, because the market has become accustomed to NVIDIA exceeding expectations, the direction of the share price is more likely to depend on management's ability to explain its competitive advantages, memory constraints, financing arrangements, and the sustainability of its China business. Over the medium term, rising earnings expectations and valuation multiples below historical levels are the primary basis for JPMorgan's continued Overweight view.

Risks

• Macroeconomic uncertainty could weaken PC gaming demand; the report notes that approximately 53% of NVIDIA's business exposure comes from PC gaming, so consumer weakness would create downside risk to forecasts.
• If hyperscale customers' adoption of deep learning or deployment of GPUs in data centers falls below expectations, or if competition intensifies significantly, NVIDIA's revenue and earnings could fall below the report's forecasts.

What to watch

• Monitor whether F2Q27 revenue reaches $94 billion to $95 billion and the F3Q27 revenue guidance provided by management.
• Monitor the actual progress of the GB300 rack ramp, tail-end GB200 shipments, and the initial 1,000–2,000 Vera Rubin racks.
• Monitor actual China H200 shipment volumes and their contribution to October-quarter and subsequent revenue forecasts.
• Monitor how management responds to share gains by AI computing competitors such as ASICs, XPUs, AMD, and Cerebras.
• Monitor the impact of memory specification adjustments for Rubin, Rubin Ultra, and Vera CPU on revenue, product mix, and gross margin.
• Monitor management's commentary on mid-70% gross margins and memory cost inflation.
• Monitor how large infrastructure financing agreements use asset-backed structures and third-party capital to alleviate power, data center, and financing bottlenecks.