4.9 trillion yuan giant: profits doubled, why isn't the stock price rising?

NEXT PICK · Market Insights

4.9 trillion yuan giant: profits doubled, why isn't the stock price rising?

Friday, October 30, 2026

Quarterly net profit grew nearly 126% year-over-year, but in return, its stock price nearly stagnated over the past 12 months—on October 30, NVIDIA closed at $204.64, down another 4.22% in a single day. Behind a $4.94 trillion market cap, the price-to-earnings ratio (P/E) was only 25.87 times.

This company had just delivered $96.2 billion in quarterly revenue and projected about 70% growth for the next fiscal year, yet the market hesitated at this moment.

Is the market misreading, or is it seeing something beyond the financial report titles?

Latest price $204.64 ▼ -4.22%

Price-to-earnings ratio 25.87 times — EPS $7.91

Distance from the 50-day moving average -7.8% ▼ Only +0.2% below the 200-day moving average

What they sell isn't graphics cards, but an entire computing power factory

NVIDIA is no longer the company that sells gaming graphics cards. As of the second quarter of fiscal year 2027, ending July 26, 2026, its revenue reached $96.22 billion, with data centers accounting for $89 billion, a year-over-year increase of 117%, and edge computing accounting for only $7.2 billion.

Its products span AI-accelerated computing, networking, gaming, professional visualization, and automotive, with buyers ranging from OEMs, ODMs, system integrators, distributors, to cloud service providers, independent software vendors, and automakers.

And this transformation doesn't stop at the story. In fiscal year 2026, revenue reached $215.9 billion, a 65% year-on-year increase, and net profit was $120.07 billion, also up 65%.

So why can't customers do without it? The difficulty of replacement is that large-scale clusters must simultaneously solve the coordination of GPU, CPU, network, interconnect, cabinet, cooling, compiler, and software framework. A single chip can compare parameters, but an entire factory is hard to replace just by saying.

The real threshold is the combination of availability, software migration costs, system validation time, and cluster utilization.

At the same time, it is collaborating with Cisco to develop high-density pluggable optical modules, working with ChipAgents to develop Build-A-Chip-Bench, and forming a next-generation AI computing platform alliance with MediaTek.

Platform advantages are ultimately written into the income statement

In the second quarter of fiscal year 2027, revenue grew by about 105.9%, while operating profit increased by about 124.1%, with the operating margin rising to 66.2%. Return on equity (ROE) ranged from 77% to 117% on different scales.

Therefore, the report gave an irreplaceability score of 8/10, not a perfect score. The reason is simple: existing disclosures cannot provide CUDA market share, software revenue splits, migration costs, or customer retention rates. The moat is visible but cannot be measured, especially as substitute chips and customer self-development continue to erode it.

A P/E ratio of 25.87 times—is it cheap or just an illusion?

Looking only at a 25.87x P/E ratio and a forward P/E of 13.52x, this company seems unreasonably cheap.

With a single ruler, the price-to-sales ratio (P/S) exceeds 17 times, the price-to-book ratio (P/B) exceeds 22 times, the enterprise value multiple (EV/EBITDA) exceeds 22 times, and the free cash flow yield (FCF yield) is even less than 3%.

The denominator also needs to be carefully examined

The $7.91 earnings per share over the past four quarters include investment income, with about $7.8 billion in pre-tax equity gains in the second quarter alone. GAAP earnings per share of $2.46 are actually higher than the non-GAAP $2.22. The reason the price-to-earnings ratio looks good is partly because of chip sales.

The P/E to earnings growth ratio (PEG) also depends on the caliber: the research snapshot shows 0.13, Yahoo Finance shows 0.28, and MarketBeat shows 1.65. The gap is so large that readers remind readers that this metric heavily depends on growth assumptions, and once growth slows down, it quickly becomes expensive.

The report focuses on price-to-earnings ratio and cash flow as a supplement, setting three scenarios: 25 times when growth normalizes, 28 times when high growth but profit margins no longer expand, and 35 times when supply, pricing, and profit margins are all realized simultaneously.

Its original estimate was based on a lower profit scale, with downside values of about 11% and upside about 24%, concluding that it was fair rather than cheap.

If you swap the profit threshold to the latest $7.91, the current price is only about 3% away from 25x, and about 35% above 35x. The odds look better, but the premise is that you treat investment returns as sustainable profits. So, are you willing to pay the same multiple for this profit?

Cash flow is the most easily overlooked page in this financial report. Operating cash flow for fiscal year 2026 is $102.72 billion, a year-on-year increase of about 60.3%; Operating cash flow for Q1 fiscal 2027 is $50.34 billion, free cash flow is $48.55 billion, with no indication that it needs external financing to keep operations running.

But in the second quarter, net profit was $59.69 billion, but free cash flow was only $21.34 billion, down about 56% quarter-on-quarter. Third-party analysis shows that increases in accounts receivable and inventory swallowed up a large amount of cash, and the quarterly shareholder return of about $26 billion already surpassed the quarter's free cash flow.

The conclusion remains self-funding, but this judgment will need confirmation in the next financial report. Supply and capacity commitments jumped from $119 billion to $279 billion, mainly for memory purchases, with future cash consumption expected to be significant.

The company is breaking through, but its stock price is stuck above the long-term moving average

The report positions NVIDIA as a breakout rather than a mature phase, with a very complete logical chain.

Revenue rose from $46.7 billion in the second quarter of fiscal year 2026 to $57 billion, $68.1 billion, $81.6 billion, and $96.2 billion, with operating profit margin climbing from 60.8% all the way up to 66.2%. Clearly, this growth is not a one-quarter phenomenon.

The industry has shifted from a surge in volume to full-rack deployment, and for the first time, the company has forecast about 70% growth for the next fiscal year.

But the stock price is not the same description

The current price is about 7.8% below the 50-day moving average and only about 0.2% above the 200-day moving average. Over the past 30 days, the return is about -10%, and the RSI (14) is near 34. The mid-term trend has been broken, and the long-term trend is being tested.

This does not mean the market has rejected this stage; rather, it seems unwilling to pay a premium for now. A breakout period does not mean an unconditional main rise; what matters next is the fulfillment of third-quarter guidance, the direction of profit margins, and the progress of memory and advanced packaging capacity expansion.

There is more than one opponent; the real pressure comes from the customer

Competition should be viewed in three layers

Comparing single-chip competitors, system-level competitors, and customer-developed chips to all semiconductor companies is itself a misunderstanding.

AMD is the most direct competitor to general-purpose accelerated computing, competing for cloud vendors' training, inference, and high-performance computing budgets, and also offering multi-vendor bargaining value to major clients. But the focus of competition is not just peak computing power; it also includes interconnects, compilers, operator libraries, and migration costs. Currently, the gap between CUDA and ROCm lacks hard data quantization.

Intel's role is more complex; it can leverage its CPU, x86 ecosystem, and enterprise channels to form an alternative platform, but its AI product deployments and financial data are insufficient to prove how much market share it has taken away, and the weak latest guidance leaves it overwhelmed.

Broadcom approaches from both sides of custom AI accelerators, network chips, and high-speed interconnects, limiting NVIDIA's bargaining power.

Perhaps the most wary is AWS Trainium. Amazon offers it alongside NVIDIA GPUs, indicating customers can take a heterogeneous approach. It may not immediately replace all GPUs, but it proves that the largest customer also holds bargaining chips, while unit cost and deployment share still lack public data.

The moat can be broken down into five layers

The technology accumulation has been commercially validated, but the number of patents has not been disclosed; Conversion costs are high, but it cannot be proven to be permanent; The developer ecosystem is self-reinforced, but Amazon Bedrock has added more than 20 hosting models, indicating that entry value may be being reallocated.

Its scale advantage gives it the ability to invest sustainably, but cloud providers also have the scale to do in-house R&D; Regulation and supply chain are double-edged swords. CoWoS-L's tightness protects the delivery priorities of locked customers and ensures revenue realization depends on partners. No layer is insurmountable, but the cost of bypassing them is too high which is why the report gave it an 8-star rating.

Pricing is a new variable in this competition. Analysts say NVIDIA has raised prices to customers by about 15% to pass on memory costs, but raising prices when AI returns are still unclear may also stimulate cloud vendors to optimize models, improve self-developed chip utilization, and delay some purchases. Ultimately, it's a comparison of performance per watt, training cost per dollar, and delivery capability.

Can the $710 billion rack be considered its ceiling?

TrendForce estimates that the combined output value of GB300, VR200, and VR300 NVL72 cabinets will exceed $710 billion by 2027, a year-on-year increase of 214%, with shipments increasing by more than 50%.

This is a proxy measure of prosperity, but it cannot be directly equated with NVIDIA's available revenue**, nor can it be added up to data center capital expenditure.

Total Reachable Markets (TAMs) should include training, inference, general-purpose acceleration, high-performance computing, network interconnects, and software. Currently, there are no authoritative global figures, so reliable penetration rates cannot be discussed. Servable Markets (SAMs) are limited to data center AI computing and networking, with automotive, gaming, and professional visualization being increments that are difficult to quantify.

The most reliable agent available to the market (SOM) is actual revenue: the second quarter simple annualized figure reached $384.88 billion, but this is not company guidance.

The report's three-year scenario is $320 billion to $380 billion in 2026, $420 billion to $560 billion in 2027, and $480 billion to $650 billion in 2028.

This range now appears conservative

Some media estimates based on market expectations that about 70% growth means revenue for the fiscal year ending January 2028 will reach $690 billion to $700 billion.

The most sensitive factors are not software TAMs, but shipment volume and average selling price. Upstream capacity determines how much you can deliver, price determines how much you can earn.

Confidence from the bullish side: demand moves faster than supply

The core of the bull market is not a single quarter's performance beyond expectations, but the simultaneous existence of profitability, platform expansion, and demand suppressed by supply, all supported by solid data:

▲ Bull Case

Bearish reason: Even good companies face bad prices

Bears do not deny the company's operations; they worry that under high expectations, any slowdown could cause a double blow to profits and multiples:

▼ Bear Case

On November 17, three questions await one answer

Over the past month, the market's focus has shifted from how much AI demand will be to when investments will pay off. Intel's stock price plunged nearly 29% after disappointing guidance, Amazon's free cash flow over the past 12 months was -$42.4 billion, the Philadelphia Semiconductor Index fell more than 7% in a single day, and NVIDIA was not spared.

On the capital side, bullish orders reached $11.3 million, 1.8 times the bearish side. This is more like investors betting on earnings through adjustments, not as institutions confirming demand improvement with real money. If guidance still fails to clarify order visibility and investment recovery paths, the rebound will be difficult to sustain.

•	Fiscal Year 2027 Q3 Earnings Report: After the U.S. market closed on November 17, guidance was $108 billion, depending on whether both revenue and gross margin meet targets
•	Will the roughly 70% growth outlook be maintained? In the same call, if the upward revision is positive, the downward revision puts valuation pressure
•	Price increase transmission: In the next two quarters, quantitative gains will deliver positive gross margins, but if obstacles occur, cost pressures will be amplified
•	GB300 will be the main shipment driver in 2026: quarter by quarter this year, with a positive impact
•	VR200 a surge in volume and cabinet output value growth: 2027, high impact, and the pace will determine whether market expectations can be realized
•	CoWoS-L Expansion and HBM Supply Improvement: Moderate bullish for the next 12 to 24 months, but management says memory bottlenecks may persist until the end of 2028
•	Regulation and public opinion: The Department of Justice inquired about the Hugging Face deal, and The New York Times reported on contacts of Jensen Huang. The short-term sentiment is negative, and NVIDIA has denied the latter's hints

Performance has doubled steadily, and the stock price has remained unchanged for a year

Over the past 12 months, NVIDIA has delivered a rare, steep curve: revenue for fiscal year 2026 of $215.9 billion, Q2 net profit rising from $26.42 billion to $59.69 billion. The demand for AI infrastructure has indeed moved from pilot to large-scale deployment, and the company's business expectations have proven correct.

But the stock price followed a different path. According to the report snapshot, the return over the past 12 months was only about 0.9%, the current price has fallen about 16% from the 52-week high, and in October it fell about 10%. This does not mean the fundamentals have deteriorated; rather, it suggests that strong performance may have already been priced into earlier valuation expansion.

The August earnings were the latest emotional turning point, with revenue exceeding expectations and an even more astonishing outlook. However, guidance of a peak and decline in gross margin and a drop in second-quarter cash flow also emerged simultaneously; By the end of October, Intel and Amazon's earnings reports once again raised questions about AI investment returns, and this stock began to pay for expectations.

History leaves three layers of lessons

The earnings cycle must be verified by revenue, profit margin, and cash flow together; The price-to-earnings ratio cannot be interpreted without the growth base; The most common mistake is annualizing quarterly earnings per share, or treating price increases under tight supply and demand as permanent. To return to the upward trend, subsequent revenue and profit margins must provide answers.

⚠️ Risk Notice • Relying on third-party manufacturing, packaging, and testing, CoWoS-L may be tight and may limit shipments and revenue recognition • Memory costs rising more than expected, gross margin may drop to 71% to 72% in Q4 • High absolute valuations combined with growth or downward revisions in interest rate expectations may trigger valuation compression • Free cash flow fell about 56% quarter-on-quarter, and the rise in receivables and inventory requires continued observation • Cloud customers are concentrated and self-developed is accelerating, weakening pricing power; There are also regulatory and geopolitical review risks

🟡 Neutral Solid performance but not cheap; cash flow and gross margin determine the next step

💬 Discussion Good company but not cheap, not worth chasing highs. Live buy and sell positions posted on NVDA official accounts.

Data source

•	Source: NextPick real-time snapshot + NVIDIA earnings and investor announcements + SEC filings + TrendForce + mainstream financial media; The fiscal year labels, cash flow criteria, and price increase expressions in the research report have been corrected according to company disclosures.