NVIDIA's earnings report "captures everyone's attention": Rubin cycle, cyclical financing, and future market share.

NVIDIA will announce its quarterly earnings report for the period ending in July next week. According to the rhythm established over the past few quarters, the market is already quite familiar with the expectations of "better-than-expected performance and upward guidance." What truly determines the elasticity of the stock price are three longer-term issues: whether Rubin can successfully take over from Blackwell, whether AI demand continues to significantly exceed supply, and how much financing and balance sheet responsibility NVIDIA must undertake to secure future demand. From the perspective of brokerage expectations, the fundamentals remain strong. According to reports from the trading desk, Jefferies released a forward-looking report this week, predicting NVIDIA's revenue for the July quarter will reach $95 billion, surpassing the market consensus estimate of around $91.9 billion; for the October quarter, revenue guidance is expected to hit $108 billion, also above the market expectation of approximately $103.7 billion. Morgan Stanley's report is relatively conservative, estimating July quarter revenue at about $91.2 billion and October quarter revenue at around $102.3 billion, but it also believes the recent trend will continue, with quarterly revenue possibly continuing to increase by around $10 billion sequentially. In other words, the core of the current market debate is whether, after delivering strong results, the growth curve for 2027 and 2028 can still be adjusted upward. Blackwell remains strong, but Rubin is the next narrative for valuation. In the short term, demand for Blackwell remains robust. Microsoft and Amazon's cloud services showed more significant improvements during the quarter, and the long-term commitments to computing power from companies like OpenAI and SpaceX continue to bolster market confidence in NVIDIA's data center revenue. Morgan Stanley pointed out that even if Rubin has yet to make a substantial contribution, NVIDIA's current growth is already quite strong; greater acceleration may occur after the January quarter next year. Rubin is one of the most important keywords in this earnings call. Jefferies believes that Rubin's ramp-up speed may significantly surpass that of Blackwell in its early stages. The reason lies in Rubins continuation of the 72-GPU Oberon NVL72 rack form, which allows for the reuse of the liquid cooling data center deployment and manufacturing infrastructure accumulated from GB200 and GB300. At the same time, the next-generation design has significantly reduced manual wiring and piping installation; management previously mentioned that the assembly time for computing trays is expected to decrease from about two hours to five minutes. The implication is that NVIDIA is upgrading its AI chips into a replicable and scalable AI factory system. Jefferies expects that the VR/R200 will account for about 12% of NVIDIA's GPU revenue by the third quarter of the 2027 fiscal year, rising to over 40% in the fourth quarter, and will become the dominant revenue source by the first quarter of the 2028 fiscal year. If this pace holds true, the market's valuation logic for NVIDIA will shift from "how long can the Blackwell cycle continue" to "can Rubin unlock the next round of higher base growth." Vera CPU: Not the main course, but completes the full-stack control. Aside from GPUs, the Vera CPU is also an incremental point of interest in the earnings report. Jefferies anticipates that NVIDIA's CPU revenue in the 2027 fiscal year could reach $20 billion, including Grace and Vera; their calculations show that CPU revenue from GPU sales is about $8.65 billion, with standalone Vera CPU revenue around $11.35 billion, leading to an annualized revenue exit in the fourth quarter of nearly $32 billion. In absolute terms, the CPU cannot compete with NVIDIA's AI GPU business and is more like the icing on the cake. However, its strategic significance is considerable: the Vera CPU strengthens NVIDIA's control in AI server and AI factory architectures through low-power LPDDR5X and consistent connectivity with the Rubin domain's NVLink. This will also create new pressure on AMD and Intel. Especially as large customers increasingly prefer to purchase complete racks and systems, NVIDIA sells a full computing power platform, from GPUs and CPUs to networking and software. OpenAI's large order: locking in demand, but raises questions about circular financing. Of greater concern to the market is NVIDIA's long-term computing power arrangements with OpenAI. According to Jefferies' analysis, NVIDIA has reached an arrangement with SB Energy, in which SB Energy will build, own, and operate the PORTS-Pike Technology Campus in Ohio, leasing it to OpenAI under a 20-year contract, with NVIDIA as the exclusive computing power supplier. NVIDIA will provide credit support for the initial 4.25 GW of IT capacity and has the option for an additional 3.75 GW, totaling up to 8 GW, while investing $1.5 billion into SB Energy. NVIDIA states that each generation of systems deployed at this facility could correspond to about 1.5 million GPUs, yielding between $150 billion and $200 billion in revenue; OpenAI's existing and planned commitments total about 12 GW, expandable to 16 GW, which would correspond to approximately $600 billion in NVIDIA computing power procurement by 2030. This undoubtedly provides a clearer anchor point for NVIDIAs long-term orders. However, the market will also question how much of this demand is driven by voluntary capital expenditure from customers, and how much requires NVIDIA to promote it through credit support, investment, or revenue sharing. Morgan Stanley believes that financing, market share, and gross margin are the three major contentious points for investors at present. NVIDIA's management will likely emphasize that these structures help accelerate AI infrastructure development, but investors will need to see clearer accounting disclosures, including obligations, leases, power payment supports, residual value commitments, and other details. This determines whether the market views it as locking in future revenue or worries it will become growth at the expense of the balance sheet. Gross margin and cost pressures: steady in the short term, but need verification in the long term. NVIDIA's high gross margins have always been a crucial foundation for its valuation premium. Morgan Stanley expects the company to continue to maintain a gross margin in the mid-70% range for the 2027 fiscal year. However, pressures do exist. Costs in HBM, DRAM, advanced packaging, substrates, and front-end wafers are still on the rise, and the ramp-up phase for new products usually accompanies yield issues, testing, and supply chain adjustments. Morgan Stanley cautioned that the consensus expectations for significant recovery in gross margin in the second half of the 2028 fiscal year may be somewhat optimistic. This means that if NVIDIA's management provides a more cautious assessment of long-term gross margins during the earnings call, it could trigger short-term volatility; however, if the management can clarify cost locking, pricing power, and system-wide sales structures, it may instead become a clearing of expectations. The biggest bottleneck may not be in chips but in electricity. As AI capital expenditure progresses, electricity is likely to become the most tangible constraint. Jefferies pointed out that the gap between chip demand expectations and grid supply capability is widening. Large cloud providers are increasingly working with utility companies to build dedicated natural gas plants and transmission facilities, while cutting-edge AI labs and new cloud providers are increasingly adopting behind-the-meter electricity solutions to bridge the time gap between grid access and computing power deployment. This does not mean that the expansion of AI infrastructure will stop, but it will make the rhythm of converting orders into revenue less linear. Solutions such as gas turbines, backup power, and fuel cells are all under advancement, but the equipment delivery cycle, grid approval, local regulation, and community acceptance will all impact the progress of data center construction. Nuclear power and small modular reactors are more likely to be variables after 2030. Therefore, NVIDIA's core challenge in the coming years will be whether customers' data centers, power, and financing can keep pace. Clarity is the catalyst for stock prices. Morgan Stanley maintains its overweight and preferred stock view on NVIDIA, with a target price of $288, noting that NVIDIA's current price-to-earnings ratio is about 17 times FY28, which is not expensive compared to its own history, semiconductor peers, and the broader market. However, low valuation does not necessarily lead to an immediate increase. Over the past four quarters, even as NVIDIA has delivered strong results, the stock price has consistently reacted negatively the following day, indicating that the market has come to see better than expected as a default script. What will truly drive valuation recovery next is whether management can clearly answer four questions: Beyond the October quarter guidance, will the January quarter see a significant acceleration from Rubin; Is Rubin's ramp-up smooth enough to maintain product leadership; What is the financing responsibility behind long-term agreements with OpenAI and others; Will power, regulation, and cost pressures impact the delivery rhythm after 2027. Overall, NVIDIA's earnings report is likely to still reflect strong performance. However, the market has entered the next phase: no longer satisfied with just seeing demand, but wants to see how demand can be fulfilled by supply, power, financing, and margins. NVIDIA's story remains grand, and short-term performance is also supported; but as investment in AI infrastructure reaches levels in the trillions, investors need to assess whether an entire system for AI industrialization can continuously operate.