Nvidia earnings test AI demand and financing

Nvidia’s fiscal 2027 second-quarter report on Wednesday will test whether the company can keep converting unprecedented demand for AI systems into durable revenue growth as data-center construction becomes constrained by financing, power availability and rising component costs. Wall Street expects quarterly revenue of about $92.0 billion, adjusted earnings per share of $2.09, and data-center revenue of roughly $85.4 billion.

The results arrive after Nvidia shares gained more than 12% in August, according to Goldman Sachs, leaving the market focused less on whether the chipmaker can beat near-term forecasts and more on the quality of its outlook. Goldman Sachs said a revenue beat and higher guidance may not be enough to lift the stock without clearer evidence that cloud customers can earn attractive returns from their AI spending, that Nvidia’s financing commitments are manageable, and that the company has room for substantial capital returns.

Jefferies has set a higher bar than the broader market, forecasting $95.0 billion in quarterly revenue and an October-quarter revenue outlook of about $108.0 billion. Broader estimates cited by Jefferies stand near $91.9 billion for the reported quarter and $103.7 billion for the following quarter.

Options market expects a narrower reaction

Options traders are pricing an implied move of about 5.4% after the results, according to ORATS founder Matt Amberson. That would be Nvidia’s smallest implied earnings move since August 2021 and below the 7.4% average implied move over the past 12 quarters.

Amberson said Nvidia’s actual share-price moves following earnings have been smaller than implied volatility in each of the past eight quarters, prompting a reduction in the event premium embedded in options prices. The pattern reflects a market that has repeatedly expected dramatic reactions from Nvidia’s reports but has often seen a more contained response after the numbers are released.

Forecast markets show a similar divide between confidence in Nvidia’s financial performance and uncertainty about the stock’s immediate reaction. They place the probability of the company beating estimates at 96%, while assigning only a 43% probability that shares close above $220 following the report.

Morgan Stanley forecasts July-quarter revenue of about $91.2 billion and October-quarter guidance near $102.3 billion. The bank expects revenue to keep rising by about $10 billion sequentially in coming quarters, retained an overweight rating and $288 price target, and said Nvidia was trading at roughly 17 times its estimated fiscal 2028 earnings.

Financing commitments move into focus

Questions around customer funding may be among the most closely watched parts of Nvidia’s earnings call. Large AI projects increasingly require financing structures that extend beyond a straightforward purchase of chips and servers, placing suppliers, cloud platforms and project developers closer to the risks of the infrastructure buildout.

Jefferies described one proposed framework involving SB Energy’s PORTS-Pike Technology Campus in Ohio, which would be built and operated by SB Energy and offered to OpenAI through a 20-year lease. Nvidia would serve as the exclusive compute supplier under the arrangement.

The structure described by Jefferies would have Nvidia provide credit support for the first 4.25 gigawatts of IT capacity, hold an option over the remaining 3.75 gigawatts, and invest $1.5 billion in SB Energy. Jefferies estimated that each system generation at the site could require about 1.5 million GPUs, representing $150 billion to $200 billion in Nvidia revenue under its assumptions.

Jefferies also cited OpenAI commitments of about 12 gigawatts, with capacity potentially expanding to 16 gigawatts. The firm estimated that volume could equate to roughly $600 billion in Nvidia compute purchases before 2030. Those figures remain dependent on construction schedules, available electricity, financing and customers’ ability to turn AI capacity into commercial services.

Morgan Stanley estimated that five hyperscale cloud providers, along with Nvidia and Broadcom, have committed more than $3.1 trillion through guarantees, leases and other forms of support intended to finance AI infrastructure spending by third parties. The bank identified financing exposure, market share and gross margin as the three central areas of debate heading into Nvidia’s report.

Power shortages could affect delivery timing

AI demand is colliding with a more basic constraint: electricity. Jefferies said the gap between projected chip demand and available grid capacity is widening, leading hyperscalers to explore dedicated gas-power facilities while some AI labs use behind-the-meter generation to avoid delays in connecting to the grid.

Data centers need more than accelerated computing hardware before they can produce revenue. Turbine and backup-power delivery cycles, grid-interconnection approvals, local regulations and site construction can all delay deployment, Jefferies said. That means Nvidia can continue to report strong orders while revenue conversion becomes less predictable across individual quarters.

The power bottleneck also affects the market’s interpretation of customer demand. A delayed data-center opening does not necessarily signal weaker appetite for Nvidia systems, but it can push hardware delivery schedules or cause customers to phase purchases around power availability rather than chip availability.

Rubin ramp and margin pressure

Nvidia’s next major product transition, centered on the Rubin platform, will be another focal point. Jefferies said Rubin retains the 72-GPU NVL72 rack design, allowing customers and manufacturers to reuse liquid-cooling systems and construction work developed for GB200 and GB300 deployments.

That degree of reuse could reduce installation friction during the changeover. Jefferies said Nvidia management had previously cited a reduction in compute-tray assembly time from about two hours to roughly five minutes. The firm forecasts Rubin and R200 products will account for about 12% of GPU revenue in fiscal 2027’s third quarter, exceed 40% in the fourth quarter and become Nvidia’s leading revenue source in fiscal 2028’s first quarter.

Citi, which maintained a $300 price target, said Nvidia has secured high-bandwidth memory supply for 2026 and 2027. HBM is a critical component in AI accelerators, and supply stability would support Nvidia’s ability to deliver systems into large infrastructure projects.

Supply security does not eliminate cost pressure. Nvidia has told large customers that server prices using Vera Rubin and Grace Blackwell chips will rise by more than 15% early next year, with higher HBM costs cited as the main reason. DRAM, advanced packaging, substrates and wafers are also becoming more expensive across the hardware supply chain.

Morgan Stanley expects Nvidia to maintain its fiscal 2027 gross-margin outlook in the mid-70% range. The bank cautioned that consensus estimates for a sharp margin rebound in the latter half of fiscal 2028 could prove too optimistic if higher manufacturing and memory costs persist through the Rubin transition.

Competition adds another layer to expectations

Nvidia remains the dominant supplier of AI accelerators, but major technology companies are expanding internal chip programs. Google and Meta are among the platforms developing custom silicon, while Broadcom has become a key partner for companies seeking alternatives to general-purpose GPU purchases.

The article cited a reported OpenAI-Broadcom project known as “Jalapeño,” which was said to have exceeded Blackwell in some benchmark tests. Such claims depend heavily on workload design and testing conditions, but custom chips can shift portions of demand away from Nvidia where customers operate sufficiently large, specialized workloads.

Bespoke Investment Group said Nvidia’s share performance has diverged from the broader semiconductor group in recent months. The PHLX Semiconductor Index has weakened since a late-June high, while Nvidia’s seven-session slide marked its longest consecutive losing streak since the stock’s bull run began in October 2022.

Wednesday’s report will therefore be judged against a demanding standard: evidence that Nvidia can sustain its revenue trajectory while its customers confront the practical limits of financing multi-gigawatt campuses, sourcing power and absorbing higher system costs.