NVIDIA Earnings Preview Q2 FY27: Blackwell, Rubin and Gross Margin Are Key

-NVIDIA guided US$91.0 billion for the quarter (±2%) with a non-GAAP gross margin of ~75%, but market consensus (~US$92.0 billion, EPS US$2.14) likely understates what investors actually expect; some analysts already project figures of US$93-95 billion.

-The real catalysts are the October guidance (~US$104.0 billion, with bullish reads up to US$107-108 billion), the pace of the Blackwell-to-Vera Rubin transition, financing agreements exceeding US$500 billion, and the China optionality.

-Technically, US$213-215 is the support to defend: losing it opens the way to US$207 and US$195.34, while reclaiming US$220 enables an attempt toward the US$225-235 zone.

The quarter about to be reported starts from an already extraordinary base: last quarter NVIDIA posted record revenue of US$81.6 billion (+85% year-over-year), with Data Center reaching US$75.2 billion (+92%). The company's own guidance of US$91.0 billion (±2%) incorporates a non-GAAP gross margin of approximately 75% and does not include Data Center Compute revenue from China, leaving any recovery in that market as a potential source of upside surprise rather than part of the base case.

Market consensus currently sits around US$92.0 billion in revenue and US$2.14 in EPS, with Data Center representing roughly US$85.4 billion, close to 93% of total expected sales. However, that consensus likely understates the true level of expectations the market will bring to the announcement: some analysts project sales near US$93.0 billion, supported by strength in B300 and sequential Data Center growth above what's embedded in consensus, while others are even more bullish, seeing revenue approaching US$94-95 billion and expecting the momentum to continue over the coming quarters. In other words, a simple "beat" against the US$92.0 billion consensus likely won't constitute a sufficient surprise on its own.

More important than the quarter just ending, therefore, will be the outlook for the period ending in October. Consensus sits around US$104.0 billion, but here too there is a significant gap between official market expectations and the scenarios some analysts are working with, with estimates ranging from US$105.0 billion up to US$107-108 billion in the optimistic case.

Behind these expectations remains Blackwell. Demand for B300 and Blackwell Ultra remains elevated, and supply chain signals suggest that advanced packaging capacity availability is improving; several analysts have accordingly raised their CoWoS capacity estimates and calculate that NVIDIA could have access to enough HBM for approximately 11 million units in 2027. The fundamental takeaway is important: NVIDIA's main challenge continues to look more like how much product it can supply rather than finding buyers for that product. At the same time, rising hyperscaler capex sustains an exceptionally robust demand base, with Reuters noting that Big Tech is targeting combined spending of more than US$730 billion during 2026 — a figure that helps explain why physical infrastructure constraints are becoming as relevant as GPU availability itself.

Attention, however, will begin shifting from Blackwell to Vera Rubin. NVIDIA has indicated that Rubin shipments should begin during the October quarter, and the company has stated it has visibility into at least US$1 trillion of cumulative Blackwell and Rubin demand through 2027; in this way, Blackwell can sustain elevated demand while Rubin begins ramping, with the most significant revenue acceleration expected later on, particularly in the January quarter, once the new product reaches greater scale. As a result, any comments from Jensen Huang on production timelines, yields, HBM memory availability, initial system volumes, and adoption pace will likely matter more for 2027 estimates than a few tenths of a point of difference in this quarter's just-reported EPS.

The transition also puts gross margin back at the center of the discussion: NVIDIA expects approximately 75% for the quarter and has maintained a mid-70% margin expectation for the fiscal year. The question isn't so much whether the company can hit that near-term target, but how much of that profitability it can preserve as system complexity increases and memory, wafer, packaging, and substrate costs rise alongside a new technology transition.

A second theme that could end up being as important as the operating results is the financing of AI infrastructure. NVIDIA recently announced agreements with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR aimed at mobilizing more than US$500 billion in capital for compute infrastructure. Several sources have indicated that NVIDIA could back up to US$125 billion, equivalent to roughly 25% of potential projects. Shortly after, the company committed up to US$105 billion in guarantees tied to OpenAI's massive Ohio campus, along with an investment in SB Energy. These initiatives can be viewed positively as a natural extension of NVIDIA's ecosystem, since once chip availability improves, the main bottlenecks shift to land, power, physical infrastructure, and financing.

But this strategy also raises one of the most relevant questions for the stock's valuation. The risk is that a growing share of demand ends up depending on the supplier itself directly or indirectly financing its customers — particularly AI labs with balance sheets far weaker than the major hyperscalers. Notably, these structures are fueling concerns about "circular financing" and about who ultimately bears the risk if the economic returns on AI infrastructure fall short of current projections. As such, this earnings release will also be an opportunity for NVIDIA to clarify more precisely how much capital will genuinely remain with third parties, what residual exposure its own balance sheet might carry, and to what extent these structures simply remove financial friction from otherwise genuine economic demand. A convincing explanation could turn this debate into a positive catalyst; an unclear expansion of guarantees could produce the opposite effect.

China represents another important source of optionality. The guidance provided for the quarter explicitly assumed zero Data Center revenue from that market, but the regulatory situation has begun to show some movement: several outlets reported that small shipments of H200 chips have started entering China, with ByteDance and Tencent each receiving approximately 10,000 processors, though restrictions remain and considerable uncertainty persists around Beijing and Washington's policies. As such, any incremental China revenue could be a positive versus prior guidance, but it would be premature to treat it as a structural growth driver. More important will be how NVIDIA assesses the recoverable size of that market and the stability of export rules over coming quarters.

Finally, the market share debate isn't going away either. AMD is making progress on rack-scale solutions and recently reported that its Data Center revenue more than doubled year-over-year, while hyperscalers like Google continue investing aggressively in proprietary ASICs and TPU-related technology. The competitive threat doesn't necessarily imply an immediate contraction for NVIDIA — the market's absolute growth remains fast enough to allow multiple vendors to advance — but it does shape the multiple investors will be willing to pay for 2027 and 2028 earnings.

All of this sets up a release where the relationship between numbers and expectations will be particularly delicate. Notably, NVIDIA's stock has reacted negatively the day after each of its last four earnings reports despite posting very strong numbers — evidence that the market has learned to price in results ahead of time. At the same time, NVIDIA's multiple has compressed significantly relative to the rest of the sector, reducing some of the valuation risk that characterized earlier stages of the rally. The most favorable scenario, then, would be a combination of revenue near the high end of expectations, forward guidance clearly above consensus, confirmation that Rubin will ramp without meaningful delays, margin stability around 75%, and greater clarity that the new financing structures expand the market without shifting disproportionate credit risk onto NVIDIA's balance sheet. That said, the August 26 results will likely confirm that immediate demand for AI compute remains exceptionally strong. The real test for the stock will be whether NVIDIA can convince the market that this strength can extend through 2027 without meaningfully eroding margins, market share, or financial quality.

From a technical standpoint, NVIDIA maintains a bullish underlying structure, though in the short term it is going through a consolidation phase with fading momentum. The price at US$216.48 sits below the 10-day EMA (US$218.83), but still above the 21-day EMA (US$215.21) and the 50-day SMA (US$207.28), preserving the medium-term positive bias. The US$213-215 zone is the immediate support and the key level to defend: if it holds, there's a moderately favorable probability of stabilization and a fresh attempt toward US$220.15, followed by US$225-235. A technical entry could be considered on a confirmed bounce at US$213-215 or after reclaiming US$220 on a daily close, with profit-taking zones at US$225-230 and US$233-235.

RSI at 53 reflects neutral momentum, while ADX at 18 confirms the absence of a strong directional trend, favoring sideways moves and false breakouts. The main risk is a break below US$213: in that scenario, the probability of a decline toward US$207 and subsequently US$195.34 would increase. A daily close below US$213 would be a reasonable signal for a defensive exit or reduced exposure.