All financial figures and market data are current as of July 20, 2026.
I have a complicated relationship with Nvidia stock.
Every time I look at the valuation and decide the market has finally become too enthusiastic, Nvidia reports another quarter that makes my concerns look like they were prepared using a calculator from 1997.
Revenue rises. Data-center demand breaks another record. Management raises the bar. Analysts revise their estimates. The stock climbs. Then someone announces that Nvidia is obviously overvalued because it has already gone up, which remains one of the financial world’s favorite ways to confuse price movement with business analysis.
Still, I cannot ignore the question.
Nvidia shares recently traded around $203, giving the company a market capitalization of approximately $4.96 trillion. That is not merely a large valuation. It is the kind of number that causes me to check whether I accidentally leaned on the keyboard.
At nearly $5 trillion, Nvidia is no longer an overlooked semiconductor company waiting for Wall Street to discover artificial intelligence. It is one of the most closely followed businesses on Earth. Everyone knows the story. Everyone has seen the charts. Every institutional investor, portfolio manager, technology analyst, message-board philosopher, and person who bought one fractional share last Tuesday has an opinion.
The easy money based on the market simply recognizing Nvidia’s importance has probably been made.
The question now is whether the company can continue growing into an enormous valuation—or whether investors have priced the stock as if revenue will rise forever, competitors will remain permanently behind, customers will never question their spending, governments will avoid disruptive regulation, and every new chip will arrive exactly on time carried into the data center on a velvet cushion.
After looking at the business, the valuation, and the risks, my answer is this:
I still consider Nvidia a buy for patient long-term investors, but I would not treat it as a stock that must be purchased immediately at any price. At roughly $203, I see it as a cautious or gradual buy rather than a table-pounding bargain. Existing shareholders have strong reasons to continue holding. New investors should consider building a position in stages and leave room for volatility, because a nearly perfect company can still produce a deeply imperfect stock return when expectations become too ambitious.
That distinction is the entire investment case.
Nvidia Is Still Producing Numbers That Look Fictional
Before I worry about the valuation, I need to acknowledge what Nvidia is accomplishing as a business.
In its first quarter of fiscal 2027, which ended April 26, 2026, Nvidia reported revenue of $81.6 billion. That was an increase of 20% from the previous quarter and 85% from the same quarter a year earlier.
An 85% annual growth rate would be remarkable for a young company operating from a small revenue base. Nvidia produced it while already generating tens of billions of dollars every quarter.
Data-center revenue reached $75.2 billion, rising 92% year over year and 21% sequentially. Data-center compute revenue was $60.4 billion, while networking revenue reached $14.8 billion, up 199% from a year earlier. The company also maintained a GAAP gross margin of 74.9%. Nvidia fiscal 2027 first-quarter results
I read those numbers and have to remind myself that Nvidia sells computing infrastructure, not enchanted beans.
The company guided for second-quarter revenue of approximately $91 billion, plus or minus 2%. That guidance assumes no data-center compute revenue from China.
If Nvidia hits the midpoint, quarterly revenue will have increased by more than $46 billion in only one year. For perspective, many successful public companies will not generate $46 billion in total sales during their entire corporate lives. Nvidia appears capable of adding that much annualized business while commentators debate whether the growth phase has already ended.
During fiscal 2026, Nvidia generated $215.9 billion in revenue, up 65%. Operating income reached $130.4 billion, net income rose to $120.1 billion, and operating cash flow totaled $102.7 billion. At the end of that fiscal year, Nvidia held approximately $62.6 billion in cash, cash equivalents, and marketable securities. Nvidia fiscal 2026 Form 10-K
These are not the financial characteristics of a speculative company surviving on promises and motivational presentations. Nvidia is producing extraordinary revenue, profit, cash flow, and margins right now.
That matters because expensive stocks become dangerous when the underlying business consists mostly of a story. Nvidia has a story, certainly, but it also has enough cash flow to make the story sit down and behave itself.
Nvidia Is Selling More Than Chips
Reducing Nvidia to “a company that makes AI chips” misses why it has been so difficult for competitors to displace.
The hardware is essential, but the moat extends beyond the GPU.
Nvidia has built a computing platform involving chips, networking equipment, complete systems, software libraries, development tools, and an enormous community of developers trained to work within its ecosystem. CUDA, Nvidia’s parallel-computing platform, has had years to become embedded in research, enterprise applications, cloud infrastructure, and AI-development workflows.
Customers are not simply comparing one chip with another on a specification sheet. They are comparing entire computing platforms.
A competitor can produce a capable accelerator and still face a painful question: How easily can developers move existing workloads onto it? If switching creates delays, retraining costs, software problems, integration risks, or weaker support, the cheaper chip may become the more expensive decision.
This is where Nvidia’s advantage becomes self-reinforcing.
Developers build around Nvidia because its platform is widely available. Cloud providers offer Nvidia systems because customers request them. Customers request them because developers know the tools. More usage generates more optimization, more software support, and more reasons for the next customer to select Nvidia.
That does not make the company invincible. Technology history contains a large cemetery filled with businesses once described as invincible. It does mean competitors must attack more than a single product.
Nvidia is also expanding the definition of what it sells. It is moving beyond individual accelerators toward full-scale AI factories—massive systems that combine compute, networking, storage, software, and power-efficient architecture. Its Blackwell platform is being followed by Blackwell Ultra and the Vera Rubin architecture. The company is positioning itself not merely as a component supplier but as the infrastructure layer for training, inference, agentic AI, robotics, autonomous systems, scientific computing, and national AI projects.
That is an ambitious collection of markets. It also creates a risk that every corporate presentation begins to sound as though Nvidia plans to sell intelligence directly to civilization.
For the moment, however, customers are buying.
The Next Phase Is About Inference
The original AI investment boom centered heavily on training large models. Companies needed enormous clusters of GPUs to train increasingly capable systems.
The bear argument is that this spending cannot continue indefinitely. Eventually, the major models are trained, the infrastructure is built, and demand slows.
I think that argument underestimates inference.
Training creates the model. Inference is what happens every time someone actually uses it.
When a person asks an AI assistant a question, generates an image, summarizes a document, writes software, analyzes medical data, operates an autonomous system, or deploys an AI agent, computing resources are consumed. More users, more applications, more complex reasoning, and more generated tokens all increase the need for inference infrastructure.
The shift toward reasoning models could be especially important. These systems may use considerably more computation to work through complex tasks before producing an answer. An AI system that responds instantly with a short sentence consumes one level of resources. A system that conducts research, compares sources, writes and tests code, analyzes an entire database, or coordinates multiple agents can consume far more.
This changes the investment story.
The AI infrastructure buildout is not necessarily a one-time event in which a few giant companies purchase training clusters and then close their wallets. If AI usage continues expanding across search, advertising, business software, customer service, healthcare, manufacturing, robotics, autonomous vehicles, cybersecurity, entertainment, and scientific research, inference demand could become continuous.
That does not guarantee that Nvidia captures every dollar. Custom accelerators and competing GPUs will take portions of the market. But the total market may become large enough that Nvidia can lose some share while continuing to grow.
I would rather own a strong company in an expanding market with increasing competition than a monopoly in a shrinking one.
The Customers Have Money—But They Also Have Leverage
Nvidia’s largest buyers include hyperscale cloud providers and major technology companies. These customers are spending staggering amounts on AI infrastructure.
That is good news because they possess the balance sheets and cash flow necessary to fund the buildout.
It is also a risk because concentrated customers have leverage, engineering talent, and powerful incentives to reduce their dependence on Nvidia.
Nvidia reported that one direct customer represented 22% of fiscal 2026 revenue and another represented 14%. That means two direct customers accounted for more than a third of total revenue. The filing also noted that one AI research and deployment company contributed meaningfully to revenue by purchasing cloud services from Nvidia’s customers. Nvidia fiscal 2026 Form 10-K
Customer concentration does not automatically make the business unhealthy. Large infrastructure projects naturally flow through large cloud providers, original equipment manufacturers, and system builders. But it does mean a shift in spending from a few companies could noticeably affect Nvidia’s results.
The hyperscalers are also developing custom chips.
Alphabet has its tensor processing units. Amazon has Trainium and Inferentia. Microsoft has developed AI accelerators. Meta has internal silicon projects. These companies are not designing chips as a weekend craft activity. They want better control over cost, supply, power consumption, and workloads.
I do not expect custom silicon to eliminate Nvidia. I do expect it to capture portions of predictable, high-volume workloads where customers can justify developing specialized hardware.
Nvidia’s defense is that AI computing remains difficult, rapidly evolving, and increasingly system-level. A custom accelerator may perform well on a specific task, but customers still need networking, software, flexibility, development tools, and support for new model architectures. Nvidia’s full platform can remain valuable even when a customer uses alternative chips elsewhere.
The likely future is not “Nvidia wins everything” or “custom chips destroy Nvidia.” The likely future is a complicated market in which Nvidia remains the leading general-purpose AI platform while customers use custom silicon to lower costs where it makes sense.
That future can still be extremely profitable.
It may not justify every valuation investors are willing to assign.
Nearly $5 Trillion Changes the Math
At approximately $203 per share, Nvidia’s market capitalization is around $4.96 trillion. Its trailing price-to-earnings ratio is roughly 31 based on current market data.
On the surface, a P/E near 31 does not look absurd for a company that recently grew quarterly revenue by 85% and guided toward another sequential increase. Plenty of slower businesses have traded at similar or higher multiples while accomplishing far less.
The problem is not simply the current P/E ratio. The problem is the scale embedded in the market capitalization.
For a $5 trillion company to double, investors must eventually value it near $10 trillion. That is possible over a long enough period if earnings continue compounding, but the sheer size reduces the probability of repeating Nvidia’s historical stock performance.
Nvidia does not need merely to remain a great company. It needs to generate enough additional profit to make today’s valuation look modest.
Suppose Nvidia eventually earns $200 billion annually. A valuation of $5 trillion would still represent 25 times those earnings. If net income grows beyond $250 billion, the current price begins looking much more comfortable. If earnings stall closer to current levels, the stock can spend years digesting the valuation even if the company remains dominant.
This is where I separate business quality from expected return.
I can believe Nvidia is one of the finest companies in the market while also believing the stock may deliver less spectacular returns from here. A company does not have to fail for an investment to disappoint. It merely has to perform below the expectations built into the price.
That is the danger of priced-in perfection.
What Does “Priced In” Actually Mean?
People often say good news is “priced in” as though the market has completed a neat calculation and placed the correct answer in a sealed envelope.
In reality, the market is a constantly changing argument.
Investors disagree about future revenue, margins, competition, interest rates, regulation, product cycles, and the duration of AI spending. Those disagreements create the price.
At $203, the market clearly expects Nvidia to remain the center of AI infrastructure. It expects Blackwell demand to remain strong. It expects the transition to Rubin to go well. It expects the major cloud providers to continue spending. It expects inference to expand. It expects high margins and enormous profits.
Those expectations are not unreasonable. Nvidia’s actual results support them.
Perfection would mean assuming that nothing material goes wrong.
That assumption would be dangerous.
Nvidia has already shown how quickly government policy can affect the business. U.S. export restrictions related to advanced AI chips contributed to a $4.5 billion charge for H20 excess inventory and purchase obligations in fiscal 2026. Nvidia has warned that future controls could affect demand beyond China, disrupt supply and distribution, encourage international customers to remove U.S. semiconductors from product designs, and support the development of foreign competitors. Nvidia fiscal 2027 first-quarter Form 10-Q
The company’s guidance for the second quarter assumes no data-center compute revenue from China. That reduces one uncertainty in the immediate forecast, but it also highlights a market Nvidia is effectively unable to serve.
I cannot model export policy with confidence because governments have a habit of changing complex rules at moments that are not convenient for my spreadsheet.
The Supply Chain Remains a Vulnerability
Nvidia is a fabless semiconductor designer. It relies on manufacturing partners, particularly Taiwan Semiconductor Manufacturing Company, along with advanced packaging, memory suppliers, system manufacturers, and other parts of a vast supply chain.
This model allows Nvidia to focus resources on design and software rather than building fabrication plants at enormous cost. It also creates dependence on outside capacity and geopolitically sensitive regions.
Demand for leading-edge manufacturing, high-bandwidth memory, advanced packaging, power equipment, networking components, and data-center construction can create bottlenecks.
A delay in one component can slow an entire system.
The AI boom also requires immense electricity, cooling, land, and construction. Nvidia can have customer orders ready while utilities, data centers, or supply chains struggle to support deployment.
These constraints do not necessarily destroy demand. They may merely delay revenue. But when a stock carries high expectations, a delay and a cancellation can look remarkably similar during earnings season.
Wall Street rarely responds to the sentence “The long-term opportunity remains intact” with quiet maturity.
Competition Is Coming From Every Direction
Advanced Micro Devices is improving its data-center accelerator offerings. Cloud providers are developing internal chips. Startups are building specialized systems. Chinese companies are investing in domestic alternatives. Intel remains involved in the broader computing ecosystem. Application-specific chips may capture workloads that do not require Nvidia’s flexibility.
Nvidia’s advantage remains considerable, but high profit margins act like an engraved invitation to competitors.
A 75% gross margin tells the world two things:
First, Nvidia has extraordinary pricing power.
Second, there is an enormous financial reward available to anyone capable of weakening that pricing power.
Competition does not need to produce a chip that is better in every category. A rival product can win by being adequate, available, compatible, or cheaper for a specific workload.
I expect Nvidia’s market share to face pressure over time. The key question is whether the AI-computing market grows fast enough to offset that pressure.
My base case is that it does.
I believe demand for accelerated computing, inference, robotics, simulation, sovereign AI, scientific research, and enterprise automation will expand for years. I also believe Nvidia will remain a central supplier because its hardware, networking, software, and developer ecosystem create a moat that cannot be replicated quickly.
But “remain a central supplier” is not the same as “maintain every current margin and share level forever.”
My valuation must leave space between those statements.
The AI Spending Cycle Could Eventually Pause
The largest risk may not be a rival chip. It may be a change in customer economics.
The companies buying Nvidia systems eventually need to earn satisfactory returns on their AI investments. Massive capital spending can continue while investors believe future revenue will justify it. At some point, however, cloud providers and enterprises will ask whether AI products are producing enough revenue, savings, or strategic value.
If the answer is yes, spending can continue or accelerate.
If the answer is “we have gained tremendous insights and remain excited about the long-term opportunity,” I will begin checking whether anyone has located the profits.
Infrastructure booms often move through cycles. Demand exceeds supply, capacity expands aggressively, customers double-order to secure equipment, supply catches up, and growth slows. Semiconductors are especially familiar with this pattern.
AI may be a generational platform shift and still experience cyclical corrections. The internet transformed the economy, but it did not prevent technology companies from becoming overvalued or capital spending from collapsing.
I do not assume Nvidia’s revenue will rise smoothly every quarter for the next decade. I assume there will eventually be a digestion period. The stock market may discover this possibility with all the emotional stability of someone finding a spider in the shower.
The Bull Case
My bull case begins with the possibility that AI demand remains underestimated.
In this scenario, agentic AI becomes integrated into corporate workflows faster than expected. Reasoning models require substantially more computation. Inference demand expands across billions of users and devices. Sovereign governments build domestic AI infrastructure. Robotics and autonomous systems become meaningful businesses. Nvidia’s networking revenue continues surging as larger AI clusters require faster interconnects.
Nvidia maintains a dominant platform position because customers prefer the flexibility, performance, software ecosystem, and time-to-market advantages of its systems. Custom chips grow but primarily expand the overall market rather than displacing Nvidia.
Annual revenue could move well beyond $400 billion, with enormous operating leverage and cash generation. If earnings continue compounding at a strong rate, today’s valuation could look reasonable in hindsight.
Under this scenario, Nvidia remains a strong investment even near $203. The stock may not repeat its earlier multiplication, but it could still outperform the broader market over a five-year horizon.
The Bear Case
My bear case does not require AI to disappear.
It requires expectations to come down.
Hyperscalers could slow capital spending after several years of aggressive infrastructure construction. Customers could struggle to monetize AI services. Custom accelerators could take a larger share of inference workloads. AMD and other competitors could improve enough to pressure Nvidia’s pricing. Export restrictions could expand. A product transition could face delays. Supply or power constraints could prevent systems from being deployed.
Gross margins could decline from the mid-70% range. Revenue growth could fall much faster than analysts expect. Even if earnings continue rising, investors might assign Nvidia a lower multiple as the business matures.
A company can grow earnings while its stock falls if the valuation contracts faster.
If Nvidia’s multiple dropped from roughly 31 times earnings to the low 20s during a growth scare, the share price could experience a decline of 25% to 35% without requiring a corporate disaster.
That is not a prediction. It is a reminder that high-quality stocks do not come with shock absorbers.
My Valuation Range
I do not pretend I can produce an exact fair value for Nvidia. Too many variables depend on the pace of AI adoption, future chip cycles, customer spending, regulation, and margins.
Instead, I use a range.
At approximately $203, I view Nvidia as fairly valued to moderately attractive if earnings continue growing strongly through fiscal 2027 and 2028. The valuation is not cheap in absolute terms, but it is supported by exceptional growth, margins, cash generation, and competitive positioning.
My preferred buying range would be approximately $170 to $190. In that range, I would feel more adequately compensated for the risk of a spending pause or valuation contraction.
Below $170, assuming the long-term business thesis remained intact, I would become considerably more interested.
Between $200 and $220, I would be comfortable making smaller purchases or continuing to hold, but I would not chase the stock with money I might need soon.
Above $230 without a corresponding increase in earnings expectations, I would become more cautious.
For a 12-to-18-month outlook, my base-case price range is approximately $220 to $250. My stronger five-year view matters more: if Nvidia remains the leading AI-computing platform and earnings continue compounding, I believe the stock can produce attractive returns from today’s level, though probably with several stomach-testing declines along the way.
These ranges are estimates, not promises delivered from the financial heavens.
My Final Decision: Cautious Buy
If I did not own Nvidia, I would begin with a partial position rather than buying everything at once.
I might invest one-third of my intended amount now, another third during a meaningful market pullback, and the final portion after a future earnings report confirms that the thesis remains intact. This approach reduces the temptation to guess the perfect entry price—a skill I have noticed everyone claims shortly after the price has already moved.
If I already owned Nvidia at a much lower cost, I would hold.
I would consider trimming only if the position had grown so large that one company could dictate the financial mood of my household. Position sizing is not an insult to Nvidia. It is an acknowledgment that no company, however successful, deserves complete control over my future.
I would watch five things closely:
Data-center revenue growth, especially sequential growth.
Gross margins during the transition from Blackwell to Rubin.
Hyperscaler capital-spending plans and evidence of AI monetization.
Competitive adoption of custom accelerators and rival GPUs.
Export restrictions and Nvidia’s ability to serve international markets.
If those factors remain supportive, I am willing to tolerate a premium valuation.
A Great Company With a Demanding Price
Nvidia is still a buy in my view, but it is not an effortless buy.
The business is extraordinary. The growth is extraordinary. The margins are extraordinary. The cash generation is extraordinary.
The expectations are also extraordinary.
At nearly $5 trillion, Nvidia cannot rely on investors discovering the AI story. It must keep delivering on it. The company must turn unprecedented infrastructure spending into sustained revenue, protect its software and systems advantage, navigate product transitions, manage a geopolitically exposed supply chain, and remain ahead of some of the wealthiest competitors in corporate history.
That is a demanding assignment.
Fortunately, Nvidia has spent the last several years making demanding assignments look strangely routine.
I am not buying because the stock has gone up. I am not avoiding it because the stock has gone up. I am evaluating whether the future earnings power of the company can justify the current price.
My conclusion is that it can—but the margin of safety is thinner than the excitement surrounding AI might suggest.
Nvidia may not be priced for literal perfection, but it is priced for sustained excellence. Those are not the same thing. Sustained excellence is achievable. Perfection is what investors demand right before discovering that companies are operated in the physical universe, where governments interfere, customers negotiate, products experience delays, and competition refuses to remain decorative.
I would buy Nvidia gradually, hold it patiently, and expect volatility.
The company does not need to be perfect.
At this price, however, it needs to remain very, very good.
Disclosure: This article reflects a general investment opinion and is not individualized financial advice. Stock prices and company fundamentals can change quickly. Investors should consider their time horizon, financial situation, diversification, and tolerance for loss before buying any security.
Comments
Post a Comment