Nvidia may be the most impressive business in the market. That does not automatically make every price a bargain. Here is what investors are really betting on at roughly $224 per share.
There are two conversations about Nvidia, and they rarely occur in the same room.
In the first conversation, Nvidia is building the computational infrastructure for the most important technological transition in decades. Its chips power the artificial-intelligence systems that every major technology company is racing to develop. Revenue is growing at a rate normally associated with a small software company that just discovered customers, not a corporation already worth more than $5 trillion. Gross margins resemble those of a luxury tollbooth positioned directly across the only bridge to the future.
In the second conversation, Nvidia is priced as though the future has already arrived, signed a long-term lease, and agreed to annual rent increases. The company may continue producing astonishing numbers and still disappoint shareholders if those numbers are merely astonishing rather than supernatural.
Both conversations are reasonable.
That is what makes Nvidia interesting. The argument is no longer about whether the company is good. Calling Nvidia “good” is like describing the Pacific Ocean as damp. The debate is about expectations: how much future growth is embedded in the stock, how long that growth can persist, and what happens when a company delivering extraordinary results encounters a valuation that has grown accustomed to extraordinary results.
As of August 13, 2026, Nvidia shares traded near $224, giving the company a market capitalization of approximately $5.47 trillion. The stock traded around 34 times trailing earnings, based on reported market data. Nvidia had just reported quarterly revenue of $81.6 billion, up 85% from the prior year, including $75.2 billion from Data Center, up 92%. Management guided to roughly $91 billion in revenue for the following quarter—and did so without assuming any Data Center compute revenue from China.
Those figures are so large that ordinary adjectives become unemployed.
But investors do not earn returns from being impressed. They earn returns when the company’s future results exceed what the current price already assumes. That difference—between business excellence and investment expectations—is where the bull and bear cases collide.
The Numbers Nvidia Has Already Put on the Board
Before debating the future, I want to acknowledge the present, because Nvidia’s recent performance makes lazy bearish arguments difficult to sustain.
Fiscal 2026 revenue reached $215.9 billion, an increase of 65% from the prior year. In the first quarter of fiscal 2027, revenue jumped to $81.6 billion, 20% higher than the immediately preceding quarter and 85% higher than a year earlier. Data Center produced more than 92% of quarterly revenue. GAAP gross margin reached 74.9%, while operating leverage turned much of that gross profit into earnings.
Management’s second-quarter guidance of $91 billion implied another sequential increase of roughly 12%. Annualizing that single-quarter estimate would produce a revenue run rate near $364 billion. I would not mistake a run rate for a forecast—four quarters have an irritating habit of containing events—but it illustrates the scale Nvidia has reached.
At the current market capitalization, that annualized revenue pace implies a price-to-sales multiple of about 15. That would be an extravagant number for a mature hardware company. Nvidia, of course, is not behaving like a mature hardware company. It is behaving like the principal supplier to a global infrastructure buildout while collecting software-like margins.
The balance sheet and cash generation add to the appeal. Nvidia authorized an additional $80 billion in share repurchases and raised its quarterly dividend from one cent to 25 cents per share in May 2026. The dividend is no longer purely decorative, although nobody should mistake Nvidia for a utility unless their utility also grows quarterly revenue by tens of billions of dollars.
This is the foundation of the bull case: Nvidia is not merely enjoying enthusiastic projections. It is producing historic financial results in real time.
The bear does not need to deny any of that. The bear only needs to ask whether the stock price requires the historic performance to continue longer than competition, customer economics, regulation, supply constraints, and mathematics will permit.
The Bull Case: Nvidia Is Becoming the Operating System of AI Infrastructure
The strongest bull argument begins with the idea that Nvidia is not selling an isolated chip. It is selling an ecosystem.
CUDA, Nvidia’s software platform for accelerated computing, has been developed for years. Developers know it. Universities teach it. Companies build around it. Libraries, tools, models, and workflows have accumulated on top of it. That creates switching costs that cannot be captured by comparing benchmark charts alone.
A competitor may build a capable accelerator. That does not instantly recreate the developer base, software compatibility, networking technology, systems integration, support, and institutional familiarity surrounding Nvidia. Hardware performance matters, but usable infrastructure matters more. Customers are not buying silicon as a collectible. They are buying the ability to train and run models at enormous scale without turning every deployment into an archaeological expedition through incompatible software.
This ecosystem advantage gives Nvidia something more durable than a temporary lead in chip speed. It gives the company a platform position.
The bull case also argues that AI demand remains early. The largest cloud companies continue spending heavily on data centers, networking, power, and accelerators. Governments want domestic AI capacity. Enterprises are moving from experimentation toward deployment. Model developers require more computing power not only for training but also for inference—the process of using trained models to answer questions, generate media, write code, conduct research, and operate software.
Training created the first spectacular wave of demand. Inference could become the larger and more persistent market if AI applications spread throughout business and consumer life.
That distinction matters because skeptics often imagine the AI buildout as a one-time purchase. Companies buy the chips, construct the data centers, and eventually declare the job finished. The bull sees something closer to cloud computing: a recurring expansion of capacity as usage grows, models become more capable, and lower costs create new demand.
The history of computing repeatedly demonstrates that cheaper computation does not necessarily reduce total spending. It can increase consumption by making previously uneconomic applications possible. If AI inference becomes embedded in search, software development, advertising, robotics, healthcare, logistics, entertainment, and industrial systems, the appetite for accelerated computing may remain enormous even as individual tasks become more efficient.
Nvidia is also expanding the amount of each data center it supplies. The company’s opportunity is no longer limited to graphics processors. It includes networking, central processing units, systems, software, interconnects, and complete rack-scale architecture. Selling more of the system can increase revenue per deployment and deepen customer dependence on the platform.
To the bull, Nvidia resembles a combination of chip designer, systems architect, networking vendor, and software platform. Assigning it the valuation framework of an ordinary semiconductor company may therefore underestimate both its economics and its strategic position.
Then there is execution. Technology markets are filled with companies that possess a compelling roadmap and the organizational coordination of a shopping cart with one broken wheel. Nvidia has repeatedly introduced new architectures, managed complex supply relationships, expanded systems capabilities, and translated demand into revenue at a scale few businesses have ever attempted.
Execution deserves a premium because opportunity without delivery is merely a conference presentation.
The Bull’s Valuation Argument
At approximately 34 times trailing earnings, Nvidia is not statistically cheap. It is also not trading at a valuation that automatically requires fantasy if earnings keep expanding rapidly.
Suppose Nvidia’s trailing earnings per share of roughly $6.57 grow at 25% annually for three years. That would produce earnings near $12.83 per share. Apply a 25-times earnings multiple—lower than today’s—and the resulting share price would be about $321. At 30 times earnings, it would be approximately $385.
Those are not forecasts. They are illustrations of what the current price can support.
The bull could reasonably argue that 25% annual earnings growth is conservative for a company whose latest quarterly revenue increased 85% year over year. Even with substantial deceleration, Nvidia could grow into its valuation faster than a static price-to-earnings ratio suggests.
This is one reason simple declarations that “Nvidia is expensive” feel incomplete. Expensive relative to what growth rate? A company growing earnings at 8% and trading at 34 times earnings faces a very different burden from one growing at 30% or more.
If Nvidia can sustain high margins, continue gaining wallet share within AI infrastructure, and compound earnings above 20% for several years, today’s valuation may look demanding but defensible. The market is charging a premium for dominance, but dominance is visible in the financial statements rather than confined to a motivational poster in the lobby.
The bull’s central claim is therefore not that expectations are low. They plainly are not. The claim is that Nvidia’s earnings power is rising fast enough to catch expectations before expectations become fatal.
The Bear Case: Greatness Has Become the Minimum Requirement
The bear case begins with a brutal observation: Nvidia no longer receives much credit for being excellent. Excellence is the admission price.
When a company is valued near $5.5 trillion, growth must generate staggering absolute dollars. A 20% increase in market value would add more than $1 trillion—roughly the value of a major global corporation materializing on top of an already gigantic one.
This does not make further gains impossible. Market capitalization is not a warehouse that runs out of shelf space. But it changes the scale of the required outcome. Nvidia must keep converting AI enthusiasm into earnings large enough to move one of the heaviest objects in financial history.
The bear also worries about customer concentration and customer motivation. Major cloud providers are among Nvidia’s largest buyers, but they are simultaneously developing custom accelerators. Alphabet has TPUs. Amazon has Trainium and Inferentia. Microsoft and Meta have internal silicon efforts. These companies may continue buying enormous quantities of Nvidia systems while shifting selected workloads toward their own chips.
They do not need to replace Nvidia entirely to affect the economics. They need only reduce Nvidia’s share of incremental spending, pressure pricing, or create credible alternatives for certain workloads.
Customers have an obvious incentive to do this. Nvidia’s gross margins are Nvidia’s achievement and its customers’ expense. A 75% gross margin is a beautiful sight from inside the company and a recurring invitation for everyone outside it to design an escape route.
Competition also extends beyond custom silicon. AMD and other vendors are improving accelerators, software, and systems. Open standards may gain traction. Model architectures could evolve toward greater efficiency. Workloads may become more specialized. No single development must dethrone Nvidia; several modest pressures could gradually reduce pricing power or slow share gains.
Then comes the capital-spending question.
The current AI investment cycle is being financed by some of the richest companies ever created. That is comforting until it is not. Their balance sheets can support enormous expenditures, but shareholders will eventually demand evidence that AI revenue and productivity gains justify the spending.
If monetization lags infrastructure costs, customers may slow deployment. They do not need to cancel AI. They only need to move from “build everything immediately” to “please show me the return on this next data center.” For Nvidia, a moderation in customer urgency could matter even if the long-term AI thesis remains intact.
Infrastructure cycles are rarely smooth. Shortages become gluts. Double ordering disappears. Supply catches demand. Customers digest previous purchases. Products transition from one architecture to another. A business can remain strategically essential while experiencing quarters that remind investors semiconductors were not invented as a substitute for volatility.
China and the Cost of Geopolitics
Nvidia’s second-quarter fiscal 2027 guidance excluded Data Center compute revenue from China. That detail is both bullish and bearish, which is exactly the kind of thing investors enjoy because it permits everyone to win an argument.
The bullish interpretation is that Nvidia could guide to $91 billion without relying on Chinese Data Center compute sales. Demand elsewhere is strong enough to support extraordinary growth even while a significant market is constrained.
The bearish interpretation is that geopolitics can close or reshape markets regardless of customer demand. Export controls may limit which products Nvidia can sell, create inventory charges, encourage domestic Chinese alternatives, and introduce uncertainty into product planning. Nvidia must design around rules that can change faster than semiconductor roadmaps.
The risk is larger than one quarter of lost sales. Restrictions can accelerate competing ecosystems. If Chinese developers and companies are forced to use alternatives, those alternatives receive investment, users, feedback, and time to improve. A protected market can become an incubator for future competition.
Nvidia’s global dominance is real, but it operates inside national-security policy, trade restrictions, and an increasingly fragmented technological world. Investors assigning a premium for near-monopoly economics should also assign a discount for the political attention near-monopoly economics attract.
Margins May Be the Most Important Number
Revenue growth gets the headlines because revenue numbers are large and behave well in graphics. I pay equal attention to gross margin.
Nvidia reported a GAAP gross margin of 74.9% in the first quarter and guided to approximately the same level for the second. That margin reflects remarkable pricing power, product value, and supply-chain economics. It also creates a demanding baseline.
If gross margin remains near 75% while revenue grows, earnings can expand rapidly because operating expenses do not need to rise at the same rate. That is the bull’s operating-leverage engine.
If gross margin declines because of competition, product mix, system costs, pricing pressure, export restrictions, or customer bargaining power, the earnings trajectory can disappoint even while revenue continues growing.
Consider the psychology of expectations. Investors may tolerate slowing revenue growth if margins remain resilient. They may tolerate modest margin pressure if revenue repeatedly exceeds guidance. They become less tolerant when both weaken together.
The bear does not require Nvidia’s margins to collapse. A decline from extraordinary to merely excellent could be enough to reduce the valuation multiple, especially if growth is slowing at the same time. Stocks are priced at the intersection of earnings and the multiple applied to those earnings. When both move in the wrong direction, the mathematics becomes impolite.
What Growth Does Today’s Price Actually Require?
I prefer turning the valuation question backward.
Instead of asking what Nvidia should be worth, I ask what Nvidia must earn for today’s price to produce a reasonable future return.
At $224.09, a 10% annualized return for three years would require a share price near $298, ignoring dividends. The earnings required at that future price depend on the valuation multiple the market is willing to pay:
| Three-year exit P/E | EPS required in year three | Approximate EPS growth from $6.57 |
|---|---|---|
| 25x | $11.93 | 22% annually |
| 30x | $9.94 | 15% annually |
| 35x | $8.52 | 9% annually |
| 40x | $7.46 | 4% annually |
This table captures the entire debate.
If Nvidia retains a premium multiple near 35 or 40 times earnings, the company does not need heroic EPS growth to generate a 10% annual return from today’s price. But maintaining that multiple would require investors to remain convinced that growth beyond year three is durable and high quality.
If the market eventually values Nvidia at 25 times earnings—a respectable multiple for an excellent large company—EPS must compound around 22% annually merely to deliver that same return.
That growth is possible. It is not trivial.
The danger is multiple compression. Nvidia could grow EPS 15% annually, reach roughly $10 per share, and still produce a mediocre stock return if the market assigns a 22-times multiple. The business would have succeeded. The stock would have encountered arithmetic.
This is why the phrase “priced in” cannot be answered with a single percentage. The current price reflects assumptions about revenue growth, margins, capital intensity, competition, geopolitical access, and the future earnings multiple. Change one variable and the conclusion shifts.
A Simple Bull, Base, and Bear Framework
I would frame the next three years through scenarios rather than one theatrical price target pretending uncertainty has been eliminated.
Bull scenario
AI infrastructure spending remains aggressive, inference demand accelerates, Nvidia preserves its ecosystem advantage, and gross margins remain in the low-to-mid 70% range. EPS compounds near 30% annually, reaching about $14.45 after three years. At a 35-times multiple, the stock could approach $506.
This scenario assumes Nvidia continues to earn platform economics, not ordinary hardware economics. It also assumes customers’ custom chips supplement rather than materially displace Nvidia systems.
Base scenario
Growth decelerates as the business becomes larger, but AI demand remains durable. Competition and customer silicon take portions of specific workloads without breaking Nvidia’s broader ecosystem. EPS grows around 20% annually to approximately $11.35. At a 28-times multiple, the stock would be worth roughly $318.
This would represent a strong business outcome and a respectable stock return, though far below the spectacular gains investors have come to regard as a recurring subscription benefit.
Bear scenario
Cloud capital spending slows, custom accelerators gain share, export restrictions persist, and margins normalize. EPS grows only 8% annually to about $8.28, while the market applies a 22-times multiple. The resulting value would be approximately $182.
Notice that the bear scenario does not require Nvidia’s earnings to fall. It requires growth to become ordinary and the valuation to recognize that ordinariness. A stock can decline while the company continues making more money. This confuses people only because markets are expectations machines disguised as price tickers.
These scenarios are illustrative, not predictions. Their purpose is to identify the variables that matter and prevent a single point estimate from impersonating knowledge.
What the Bulls May Be Underestimating
Bulls may underestimate how quickly expectations can move ahead of execution.
Each earnings beat raises the baseline for the next report. A company can train investors to regard extraordinary performance as routine. Eventually, “record revenue” becomes less important than whether the record exceeded the unofficial record imagined by traders.
Bulls may also underestimate the customers. Nvidia’s largest buyers are not passive consumers. They are technically capable, financially motivated, and large enough to fund alternatives for years. Nvidia’s moat is formidable, but the opposing armies possess several trillion dollars and do not enjoy paying tolls.
Finally, bulls may underestimate cyclicality. AI may be a secular transformation while AI infrastructure spending remains cyclical. The internet changed the world; telecommunications investors still learned that a correct technological thesis can coexist with terrible timing.
What the Bears May Be Underestimating
Bears may underestimate how far Nvidia’s advantage extends beyond the GPU.
The bear case often sounds simple: competitors will make faster or cheaper chips. But customers buy working systems and developer productivity, not isolated benchmark victories. Software ecosystems can preserve pricing power longer than hardware comparisons suggest.
Bears may also underestimate the size of inference. If AI usage expands across billions of people and devices, efficient inference could generate more total demand rather than less. The declining cost of computation may unlock applications that do not yet exist.
Most importantly, bears may underestimate Nvidia’s ability to widen its opportunity. Networking, CPUs, rack-scale systems, software, robotics, automotive platforms, and sovereign AI initiatives can create additional revenue streams. The company is trying to turn leadership in accelerators into control of a larger computing architecture.
Betting against an expensive stock is not the same as betting against an adaptive company. Nvidia has repeatedly made skeptics look as though they brought a pocket calculator to a supercomputing contest.
What I Would Watch From Here
I would focus on five indicators rather than reacting to every rumor that crosses a trading screen.
First, Data Center growth. The absolute revenue matters, but the sequential growth rate will reveal whether demand is still accelerating, stabilizing, or entering a digestion period.
Second, gross margin. A sustained level near 75% supports the platform thesis. Meaningful deterioration would suggest changing mix, higher costs, weaker pricing, or stronger competition.
Third, cloud capital expenditure and monetization. Nvidia’s customers can fund the buildout, but investors should watch whether those customers produce growing AI revenue and productivity gains. Their return on investment eventually becomes Nvidia’s demand environment.
Fourth, inference economics. Training headlines created the boom, but inference usage could determine its duration. I want evidence that deployed AI applications are generating recurring compute demand at scale.
Fifth, competitive adoption. Announcements about custom chips matter less than actual workload migration, utilization, software support, and customer behavior. Every large customer will discuss alternatives. The financial question is how much spending those alternatives capture.
My Verdict: The Growth Is Expensive, but Not Fully Preposterous
At roughly 34 times trailing earnings, Nvidia’s valuation prices in continued excellence. It does not necessarily price in perfection, because earnings are still growing rapidly enough to reduce the multiple if the share price merely stands still.
I view the stock as neither an obvious bargain nor an obvious bubble based on the current figures. It is a high-expectation security attached to an exceptional business. The bull case can work if Nvidia compounds EPS above 20% for several years and retains a premium valuation. The bear case can work if growth slows faster than expected, margins compress, or the market decides that even a dominant semiconductor platform deserves a more ordinary multiple.
The current price appears to embed substantial growth, but not the full continuation of the latest 85% revenue increase. Nobody rationally expects a $5.5 trillion company to compound at 85% for long; after three years, the result would begin making national economies look like neighborhood lemonade stands.
What the market appears to expect is a controlled deceleration: rapid AI demand, durable platform leadership, resilient margins, and earnings growth strong enough to prevent valuation compression from overwhelming the operating gains.
That is achievable. It is also a narrow bridge.
For investors, the central question is not, “Will AI be important?” That question has become too easy. The better question is, “Will Nvidia capture enough of AI’s economic value, for long enough, to exceed the expectations already represented by $5.47 trillion?”
I can believe Nvidia will remain one of the world’s most important companies and still demand a margin of safety before buying its stock. Admiration is not valuation. A compelling product roadmap does not suspend the laws of expected return. And a stock does not reward investors for correctly identifying greatness after everyone else has identified it too.
The bull sees an expanding platform at the center of a new computing era.
The bear sees a magnificent business carrying the heaviest expectations in the market.
I see both—and that is precisely why the price matters.
Valuation snapshot
Market data as of August 13, 2026.
Share price: approximately $224.09
Market capitalization: approximately $5.47 trillion
Trailing P/E: approximately 34.1
Trailing EPS: approximately $6.57
Q1 fiscal 2027 revenue: $81.6 billion
Q1 fiscal 2027 Data Center revenue: $75.2 billion
Q2 fiscal 2027 revenue guidance: $91.0 billion, plus or minus 2%
Q2 guidance assumption: no Data Center compute revenue from China
Sources
This article is for informational and educational purposes only. It is not personalized investment advice or a recommendation to buy, sell, or hold any security. Valuation scenarios depend on assumptions that may prove incorrect.
Comments
Post a Comment