#我的七夕交易分享 Morgan Stanley: NVIDIA has fallen after earnings for four consecutive quarters—will this time be different?

What is the market really pricing in on the eve of NVIDIA's earnings?
Morgan Stanley's view of NVIDIA's upcoming second-quarter results is clear: guidance will continue to be raised above expectations, but the stock may not necessarily react positively in the short term.
The real catalyst will not be released until Rubin enters mass production next year. At the current 17x FY28 P/E, NVIDIA is valued relatively low both within the semiconductor sector and against the broader market. Frankly, the fact that the stock fell the day after beating earnings expectations for four consecutive quarters is telling in itself. The market has long priced in near-term earnings upgrades, and everyone is now focused on the longer-term logic—whether market share will erode, whether the circular financing model can work, whether gross margins will decline later, and how much revenue Rubin can contribute in the second half of the year.
Morgan Stanley expects management to speak optimistically about all these points on the call, but without substantial incremental information, routine earnings beats will have difficulty driving the stock higher. For quarterly earnings, Morgan Stanley estimates revenue of approximately $91.2 billion for the July quarter and $102.3 billion for the October quarter, representing sequential growth of 11.7% and 12.2%, respectively. These figures are slightly below consensus estimates, mainly because Morgan Stanley is leaving more upside for the January quarter next year, after Rubin's large-scale ramp-up.
Regardless of how the growth is distributed, the core view is that current shipments remain below actual end-market demand, and the growth momentum will continue into next year. This is the industry's most important fundamental support. Demand for Blackwell is even stronger than previously expected.
The June survey of the Asian supply chain had already shown increased 4nm wafer starts. Looking at the situation now, previous forecasts that the October quarter would be driven primarily by Blackwell revenue were probably too conservative. Accelerating inference demand is the main driver. Beyond frontier labs, demand from enterprise and sovereign customers is also rising, and this trend is spreading faster than many expected. This table compares the key data from Morgan Stanley and market consensus, clearly showing where the two sides diverge.
Revenue, gross margin, and EPS forecasts for the next two quarters are nearly identical, with the divergence concentrated in FY2028. Morgan Stanley's FY2028 revenue forecast is approximately $36 billion above consensus, while its gross margin forecast is more than 2 percentage points below consensus. Put simply, the market expects gross margins to recover later, while Morgan Stanley believes cost pressures will not be absorbed so easily, but that revenue upside will be greater than the market expects.
How smoothly is Rubin progressing?
Rubin's overall shipment schedule remains on track, with formal shipments beginning in the third quarter. Morgan Stanley's third-quarter revenue forecast is $9 billion, corresponding to 150k wafers at an average price of $60k per card. A new-product ramp can never be completely smooth; small issues every week are normal in the industry and do not affect the overall delivery schedule. Recently, core testing partner KYEC lowered its fiscal-year guidance, and many people immediately interpreted this as meaning Rubin shipments were below expectations. But anyone who has worked in testing and packaging understands that new-product testing taking longer than initially planned is entirely normal. It does not mean shipment volumes were cut; it only means revenue recognition is pushed back by one quarter.
Morgan Stanley's analysts also maintained their full-year forecast for total Rubin testing volume, merely shifting more volume from the third quarter to the fourth quarter. The impact on NVIDIA's overall revenue is small. Moreover, the 150k-wafer forecast is itself far below the supply chain's revised expectation of 500k wafers, leaving a substantial margin of safety. Supply remains the core issue for the next two years. Rubin's supply constraints will be more pronounced than Blackwell's, while strong demand for Blackwell can also continue into next year rather than collapsing quickly when the new product arrives. The transition between the two generations will be much smoother than in previous cycles, with no obvious gap period.

The three core issues in the market debate
The three issues the market cares most about in this earnings cycle are the direction of gross margins, the sustainability of the financing model, and whether market share will be eroded. Morgan Stanley's overall view is constructive, but it does not expect the call to contain unexpectedly new information. On gross margins, management will likely reiterate that FY2027 gross margins will remain around the mid-70% range, with capacity locked in early across the supply chain to offset cost pressures as much as possible. However, Morgan Stanley remains cautious about the market consensus that gross margins will recover in the second half of FY2028. Rising costs for DRAM, front-end wafers, packaging, and substrates are industry-wide, with no signs of relief in the short term.
If the market subsequently lowers its gross margin expectations, sentiment could be hit in the short term. But once the stock falls far enough, it could instead mark the point at which the negative news is fully priced in, since this issue has already been discussed by the market for a long time. On the financing model, Huang Renxun already explained the rationale in a blog post when the $500 billion partnership was announced, and the comments on the call should be consistent with that post. This revenue-sharing model is a new exploration for the industry, and there is still substantial market disagreement; more implementation data will be needed for validation. The debate over market share is similar. Rubin offers a significant improvement in AI factory economics compared with the already-leading Blackwell, and supply chain feedback has been very positive so far. But mass production is still in its early stages, and it will take several more quarters to determine how much impact it will have on the shares of ASICs and AMD. It is too early to draw conclusions now.
The contrast between long-term demand and valuation
At GTC, NVIDIA management said that the demand opportunity for Blackwell plus Rubin from 2025 to 2027 was $1 trillion. This figure includes networking but excludes businesses such as Groq, standalone CPUs, RTX, and software. The market currently expects combined data-center revenue of $1.054 trillion from 2025 to 2027, while Morgan Stanley's model projects $1.09 trillion. Considering that 2025 still includes approximately $30 billion in Hopper and related networking revenue, the $1 trillion target is not particularly difficult to reach. What many people have overlooked is that, in addition to hardware revenue, new-product contributions and perpetual cash flows from revenue sharing will become additional growth drivers in 2027. Current market expectations for 2027 are basically still at the level of five months ago and do not fully reflect subsequent upward revisions to demand. Over the past year, NVIDIA's valuation has gone from significantly above the semiconductor index to significantly below it, and is even below the overall valuation of the S&P 500. A 17x FY28 P/E is relatively low even among global technology giants. Morgan Stanley believes that at this valuation, there are many potential upside drivers and the risk-reward ratio is highly attractive.

Core assumptions under three scenarios
Morgan Stanley listed assumptions for three scenarios in its report, corresponding to different earnings outcomes and valuations.
Under the base case, 2027 EPS is approximately $13.08, with a 22x multiple implying a price of $288. The valuation is in line with the broader market and below that of computing semiconductor peers. The rationale is that with current high market share and high gross margins, there are few near-term drivers for further multiple expansion; a premium valuation will have to wait until new businesses gain traction.
Under the bull case, 2027 EPS is approximately $14, with a 23x multiple implying a price of $330. Drivers include data-center revenue continuing to exceed expectations, networking, Rubin systems, and software lifting the value of the full stack, AI PC adoption opening up the client market, and automotive business deployments generating recurring licensing revenue.
Under the bear case, 2027 EPS is approximately $10, with a 16x multiple implying a price of $160. Risks include a significant slowdown in data-center growth, supply catching up with demand faster than expected, aggressive competitors entering and taking share, customers beginning to develop custom chips in-house, and greater-than-expected effects from tariffs and export controls.

Industry inventory and funding conditions
In addition to content on NVIDIA itself, this weekly report also updated the inventory situation across the semiconductor industry. Semiconductor companies currently have 114 days of inventory overall, up 2 days sequentially, compared with a normal seasonal increase of 5 days; the actual increase is below seasonal levels. Current inventory is 23 days above the ten-year historical median, so overall levels remain elevated. Client-side inventory stands at 60 days, up 9 days sequentially, slightly above the seasonal increase of 8 days. Distributor inventory stands at 61 days, down 2 days sequentially, whereas a normal seasonal pattern would be a 4-day increase, indicating that distributor inventory is being worked down better than expected.
In terms of short interest, as of August 14, NVIDIA's short interest was only 1.2%, placing it in the lowest tier across the semiconductor sector. The industry average was 5.7%, with a median of 4.7%. Companies such as Skyworks have short interest above 20%, while small-cap names in quantum computing and analog chips also generally have higher short interest. There is little disagreement in the market over NVIDIA, and short positioning is very low.

Do you think the competitive landscape in the AI chip industry will change after Rubin enters mass production? Share your thoughts in the comments. $NVDA