Nvidia Q426: The secret everyone knows

Nvidia continued its explosive growth trajectory, beating analyst expectations,  largely driven by insatiable demand for its Data Centre infrastructure.

Since our latest update, we have highlighted our concern about the unrealistic investors expectations and the solid fundamentals of the company, resulting in our Neutral call and a TP of $220. Since then, Nvidia’s share price has largely remained flat, as investors rotated into adjacent AI beneficiaries such as foundry and memory players.

Ironically, despite the (once again) earnings double beat, the share price of Nvidia has tumbled from the post-earnings peak of $203 to $184 (as of 27/2). The recent US-Iran conflict has deteriorated the sentiment further lately.

We acknowledge the divergence between the share price and underlying fundamentals, yet it also reflects that investors are demanding higher levels of upside surprise at current valuation levels.

Figure 1: Our previous call for Nvidia in 2025

Nvidia Q426 result

Table 1: Nvidia’s Q4 26 results

In USD	Q4 FY26	Y/Y Change	Q4 FY25	Notes
Total Revenue	$68.13B	73%	$39.33B	Exponential computing demand
Data Centre	$62.31B	75%	$35.6B	Exceptional Blackwell demand and +263% y/y growth in networking (NVLink/InfiniBand). Hyperscalers remained ~50% of segment revenue.
Gaming	$3.73B	47%	$2.5B	Sales channel inventory naturally moderated following a strong Q3 holiday demand season.Driver (Y/Y): Strong underlying RTX and Blackwell architecture demand.
Prof. Visualization	$1.32B	159%	$508M	Skyrocketing enterprise adoption of agents and RTX PRO GPUs for agentic workflows.
Automotive & Robotics	$604M	6%	$569M	Continued steady adoption of self-driving platforms (NVIDIA DRIVE AV software).
Gross Margin	75.00%	-1.0 pts	76.00%	Slight moderation due to the introduction and scaling costs of the new Blackwell architecture.
Non-GAAP EPS	$1.62	82%	$0.89	Crushed analyst consensus ($1.54) on the back of strong profitability and massive $41.1B in FY26 share repurchases/dividends.

Source: Bloomberg Finance L.P., iFAST Compilation, Data as of 26 February 2026.

In the latest earnings, Nvidia continues to defy gravity, but the magnitude of upside surprises is naturally compressing as the revenue base expands.

Nvidia delivered a record-breaking quarter with $68.1 billion in revenue, a 73% y/y increase driven almost entirely by its Data Center segment, which now accounts for 91% of total sales and is seeing insatiable demand for the new Blackwell architecture.

However, this growth comes with distinct trade-offs: Nvidia is effectively relocating manufacturing capacity away from its Gaming segment, which fell -13% sequentially, by diverting scarce manufacturing capacity to high-margin AI chips, effectively capping consumer growth to feed enterprise demand.

On top of that, the networking revenue delivered a remarkable 263% y/y growth, reaching a record $11.0 billion. This is driven by the NVLink switch systems required to connect thousands of GPUs into massive supercomputers. Having said that, Nvidia is increasingly selling full-stack architecture rather than discrete chips. By bundling the GPUs, high-speed networking, and proprietary software into a single, integrated data centre solution, Nvidia is fundamentally shifting its business model from a component supplier to a full-stack infrastructure provider.

Nvidia’s Q1 27 Guidance

Table 2: Nvidia’s Q1 27 Guidance

	Q1 27 Guidance	Notes
Revenue	$78b	Expect a strong demand in Blackwell. Not assuming any data centre compute revenue from China
Gross Margin	75%	inclusive 0.1% impact from stock-based compensation expense
Operating expenses	$7.5b	Inclusive of $1.9b of stock-based compensation expense
Tax rate	17%-19%

Source: Nvidia, iFAST Compilation, Data as of 27 December 2026.

The company provided aggressive guidance of $78 billion for the current quarter, signalling that the AI infrastructure boom is accelerating into the agentic AI era, where models require significantly more compute power for reasoning and inference.

Management highlighted a critical evolution in workload mix. Historically, demand was driven by training huge models (GPT, Gemini, Claude). Now, the demand is shifting to inference (running them). Generally, Agentic AI models require significantly more computing per query because they perform multi-step reasoning before producing outputs. This expands the Total Addressable Market (TAM) for inference chips by 10x, debunking the bear case that demand would collapse once training was finished.

The narrative has successfully shifted from just training models to running them (inference). The $78B revenue guidance for Q1 FY27 implies a massive $10B sequential jump, confirming that the Blackwell ramp is not just on track but accelerating. On top of that, it explicitly assumes $0 revenue from China Data Centre compute due to export controls.

Blackwell is dominating, all eyes on Rubin

The transition to the Blackwell architecture provides a renewed average selling price (ASP) uplift. The demand backlog is so deep that Nvidia has revenue visibility extending into 2027. While Blackwell is currently driving top-line beats, Rubin is the margin and revenue expansion story for late FY27 and FY28. Current industry estimates place the GB300 NVL72 rack system at approximately $3.5 million, and the R100 chips will command a price premium due to its advanced 3nm process and HBM4 density. Assuming a similar unit ramp to Blackwell, this ASP uplift could potentially drive an incremental $12.6-$20.5 billion in top-line revenue.

Beyond the tech giants, Nvidia noted multi-billion dollar demand from Sovereign AI, nations building domestic compute clouds. This demand is stickier and less price-sensitive than corporate IT spend, acting as a defensive buffer against a potential hyperscaler capex pause.

Figure 2: Nvidia’s target product roadmap

Source: Bloomberg Finance, iFAST Compilation, Data as of 25 Nov 2025.

China contribution as a wildcard

Forward guidance currently models zero Data Centre compute revenue from China due to US export controls. While this structural void is fully priced into the stock, it creates an asymmetric upside risk. Should geopolitical tensions thaw, or if Nvidia successfully secures export licenses for a new tier of compliant, downgraded chips, China coming back online could provide an immediate, unmodeled boost to financials. Investors must note, however, that any realised revenue from this region would likely be subject to heavy inbound tariffs, potentially diluting its net margin contribution.

Figure 3: Nvidia’s China contribution since FY19

(The Question Again) Will ASICs erode Nvidia's dominance?

Nvidia's biggest customers are becoming its biggest competitors. Google (TPU), Amazon (Trainium/Inferentia), and Microsoft (Maia) are aggressively ramping up their own internal silicon. While they still need Nvidia for training, they are incentivised to move inference workloads to their own cheaper, custom chips, potentially eating into Nvidia's volume growth over the medium term.

To tackle the issue, Nvidia acquired Groq, known for its ultra-low latency token generation lately. By integrating Groq's compiler technologies with TensorRT and CUDA, Nvidia is effectively offering ASIC-level inference speeds within its general-purpose GPU ecosystem. This completely deepens the structural moat, ensuring developers do not need to leave the CUDA environment to achieve peak inference performance, thereby locking in hyperscaler compute spend for the Agentic AI era. We are still observing the impact of integrating Groq into Nvidia’s ecosystem in the next few quarters.

75% gross margin is an indicator of a strong moat; H2 FY27 is the key

Nvidia reported a robust 75.2% non-GAAP gross margin for Q4, but sustainability of margin remains a focal point for investors. Upstream supply chain pressures are mounting, specifically regarding the rising costs of High Bandwidth Memory (HBM3e/HBM4) from SK Hynix and Micron, alongside TSMC’s continuous CoWoS packaging price hikes.

While these input costs are a meaningful headwind, we believe Nvidia can defend the 74%–75% margin range through the aforementioned Rubin ASP uplift and the increasing software revenue mix from Nvidia AI Enterprise. However, any yield issues in the early Rubin wafers could temporarily compress margins, requiring close monitoring in H2 FY27.

Figure 4: Nvidia’s gross margin

Sustainability of CAPEX and supply chain bottlenecks as headwinds

Hyperscalers (Microsoft, Meta, Google, Amazon, ~>50% of Nvidia’s total revenue) have been spending historic amounts of capital (Capex) on Nvidia chips. Yet, this level of capital spending is unlikely to expand indefinitely.

Investors must question how long these companies can sustain $30B–$50B annual run rates on hardware before their own shareholders demand massive ROI from AI software revenue which has yet to fully materialise.

Besides, even if demand is infinite, supply is not. Nvidia is physically constrained by TSMC's capacity for CoWoS (Chip-on-Wafer-on-Substrate) packaging. This bottleneck forces Nvidia to make hard allocation decisions (sacrificing Gaming for Data Centre), which caps their surprise potential. Shipment volumes remain constrained by TSMC’s CoWoS packaging capacity, meaning revenue beats will become smaller and harder to come by.

The market’s ‘want more’ narrative still here to stay

Table 3: Price movement of Nvidia’s during earnings season

Fiscal Quarter	Revenue (Actual)	Revenue Beat (vs Consensus)	Earnings Day Move	5-Day Post Trend
FY26 Q4	$68.1B	+$2.5B (+3.9%)	-3%	Pending
FY26 Q3	$57.0B	+$2.4B (+4.3%)	-3%	-2%
FY26 Q2	$46.7B	+$1.1B (+2.4%)	-1%	-3%
FY26 Q1	$44.1B	+$1.0B (+2.3%)	3%	2%
FY25 Q4	$39.3B	+$1.2B (+3.1%)	1%	4%
FY25 Q3	$35.1B	+$1.9B (+5.8%)	-1%	-3%
FY25 Q2	$30.0B	+$1.3B (+4.5%)	-6%	-8%
FY25 Q1	$26.0B	+$1.5B (+5.6%)	9%	20%
FY24 Q4	$22.1B	+$1.5B (+7.2%)	16%	18%
FY24 Q3	$18.1B	+$2.0B (+12.0%)	-2%	-6%
FY24 Q2	$13.5B	+$2.3B (+20.4%)	0%	1%
FY24 Q1	$7.2B	+$0.7B (+10.3%)	24%	30%

Source: Bloomberg Finance L.P., iFAST Compilation, Data as of 26 Dec 2026.

Our review of recent earnings reveals a structural compression in Nvidia’s ability to surprise the market. In FY24, the company consistently delivered double-digit revenue beats (10–20%), fueling massive rallies. By contrast, FY26 has seen these beats compress to the low single-digits (2–4%). Consequently, the stock has reacted negatively or neutrally to four of the last five earnings reports (e.g., -3% in Q4 FY26), despite fundamental beats.

This decoupling signals that the market has fully priced in perfection, creating an asymmetric risk profile where the upside from a beat and raise is capped, but the downside from any execution stumble is magnified.

Maintain TP of $220, Neutral

We are rolling forward our valuation framework to FY28 to fully capture the maturity of the Blackwell cycle and the initial ramp of the subsequent Rubin architecture. On the other hand, we expect semiconductor AI demand to remain healthy but become less explosive. As such, Nvidia faces rising costs for HBM and slightly more price competition from AMD/Intel, causing margins to dip from their current record highs. Additionally, the expectation from investors as proven by the past few post-earnings movement is that investors are already pricing in a significant valuation premium.

Applying the Gartner Hype Cycle framework, we opine the current divergence between dipping/flattish share price and solid earnings is suggesting the market may be transitioning between peak of inflated expectation and the trough of disillusionment, therefore we adjust our fair PE multiple lower to 28x (from previous 30x).

Figure 5: Gartner Hype Cycle

Source: Gartner, iFAST Compilation

Combining above, our $220.00 target is based on a downwardly adjusted 28x PE multiple applied to our upwardly adjusted FY28 EPS estimate of $8.00.

Table 4: Nvidia’s valuation

	FY25	FY26	FY27e	FY28e
EPS	3.0	4.8	6.7	8
EPS Growth		59.53%	40.46%	19.40%
PE (x)	61.5	38.6	27.5	23.0
Current Price	180
Fair PE (x)				28
Upside Potential				24.4%
Target Price				220

Source: Bloomberg Finance L.P., iFAST Compilation, Data as of 4 March 2026.

Given this backdrop, from a risk-to-reward perspective, we advocate for a Neutral stance on Nvidia. Investors should maintain their core long-term holdings to benefit from the AI secular trend but avoid overexposure or chasing momentum at current valuations.

With limited details shared on the next-generation Rubin architecture, the next major catalyst is Nvidia’s flagship GTC event on March 16, 2026. This event coincides with major customer LLM refresh cycles, a period that historically drives strong positive seasonality for the stock.

Summary

The AI revolution is in the early innings, but the easy money phase of pure multiple expansion is likely over. The stock will now trade on earnings execution and the ability to navigate supply bottlenecks.

While the secular growth story for AI remains intact and stronger than ever, both the physical and financial constraints to scaling are beginning to emerge. Investors may consider accumulate on weakness on periods of weakness, as the next phase of growth depends on execution amidst tightening constraints.

Disclaimers:

This research report was prepared with the assistance of artificial intelligence (AI) tools. iFAST Financial Pte Ltd does not rely exclusively on AI for content generation; the content of this report – including all investment theses, ratings, price targets and conclusions – has been independently reviewed and verified by the research analyst(s) to ensure accuracy and professional integrity.

For specific disclosure, at the time of publication of this report, IFPL (via its connected and associated entities) and the analyst who produced this report hold a NIL position in the abovementioned securities.