How NVIDIA's Valuation Changed 2016 to 2026

Direct Answer

NVIDIA's valuation changed between 2016 and 2026 because two different things happened at the same time: the business became dramatically larger and more profitable, and the market became willing to pay more for each dollar of sales than it did at the start of the period. The second effect mattered, but the first ultimately mattered more.

NVIDIA reported $5.01 billion of revenue and $614 million of net income for fiscal 2016. For fiscal 2026, it reported $215.94 billion of revenue and $120.07 billion of net income. That is roughly 43 times the revenue and 196 times the net income in ten fiscal years. Over roughly the same period, NVIDIA's market capitalization rose from about $57.5 billion at the end of 2016 to more than $5 trillion in 2026, an increase of about 91 times. Meanwhile, its calendar-year price/earnings multiple did not simply rise forever: Morningstar shows approximately 54.7x in 2016, 90.6x in 2021, and about 33x on its August 2026 snapshot.

The valuation story moved through three distinct phases: optionality (investors paid for possibilities not yet visible in the income statement), re-rating (accelerated computing and data center became increasingly material), and earnings catch-up (AI infrastructure revenue and profit grew fast enough that the P/E ratio could fall even while the market capitalization rose). This is a historical case study, not a forecast or a recommendation.

Key Takeaways

• NVIDIA's fiscal-year revenue increased from $5.01 billion in FY2016 to $215.94 billion in FY2026, about 43x.
• Net income increased from $614 million to $120.07 billion, about 196x.
• Market capitalization increased from roughly $57.5 billion at year-end 2016 to more than $5 trillion in 2026, about 91x using the cited market-cap series.
• The market did award NVIDIA a substantially higher sales multiple over much of the period, but the multiple was not a one-way climb.
• Calendar-year P/E moved from roughly 54.7x in 2016 to a peak around 90.6x in 2021, then declined to roughly 33x by August 2026 in Morningstar's historical table.
• The biggest analytical mistake is to describe the entire decade as "multiple expansion." That misses the much larger change in the business itself.
• The second-biggest mistake is to read a split-adjusted stock chart as though each historical quoted price represented the same number of shares. NVIDIA completed a four-for-one split in 2021 and a ten-for-one split in 2024.
• The most transferable lesson is to separate what changed in the company from what changed in the price investors were willing to pay for the company.

The 10-Year Valuation Snapshot

The table below intentionally combines several lenses. Fiscal revenue is tied to NVIDIA's fiscal-year reporting. Market capitalization is a calendar-year market value series. P/E and P/S are calendar-year or current valuation observations from Morningstar. Those bases do not line up perfectly, so they should not be multiplied together as though they were a single synchronized dataset. They are shown together to reveal the shape of the decade, not to manufacture false precision.

* The cited market-cap source reports approximately $5.253 trillion as of August 2026 rather than a completed 2026 year-end value.
** Morningstar's "Current" valuation snapshot was dated August 19, 2026.

Sources: SEC: NVIDIA FY2016 Form 10-K, SEC: NVIDIA FY2026 Form 10-K, Morningstar: NVIDIA valuation history, CompaniesMarketCap: NVIDIA market-cap history.

Why 2016 Is a Useful Starting Point

NVIDIA was already a successful public company in 2016. The point of starting there is not to pretend it was undiscovered. The point is that the company's economic identity was materially different.

For fiscal 2016, NVIDIA reported $5.01 billion of revenue, a gross margin of 56.1%, operating income of $747 million and net income of $614 million. GPU revenue was $4.19 billion. The company was benefiting from GeForce gaming products, while Tesla products for data centers were growing from a much smaller base. Cash, cash equivalents and marketable securities were about $5.04 billion.

Those figures matter because they show what an investor could actually see. It was reasonable in 2016 to describe NVIDIA as a graphics-chip company with promising adjacent opportunities. It was not yet reasonable to treat a global AI-infrastructure buildout as a reported operating fact.

The valuation was already demanding. Morningstar's calendar-year table shows NVIDIA at about 10.85 times sales and 54.74 times earnings in 2016. That means the market was not valuing the company as a sleepy semiconductor manufacturer. Some future success was already embedded in the price.

That distinction prevents hindsight from doing too much work. A good historical case study should not ask, "Why didn't everyone know NVIDIA would become a $5 trillion company?" The better question is, "What evidence existed at each stage, and how much future success was the market already charging for?"

Source: SEC: NVIDIA FY2016 Form 10-K.

Phase One: 2016 to 2018 - The Market Starts Paying for More Than Gaming

The first major change in NVIDIA's valuation was not the generative-AI boom. It was the recognition that GPUs could address workloads beyond traditional graphics.

Revenue increased from $5.01 billion in fiscal 2016 to $6.91 billion in fiscal 2017 and $9.71 billion in fiscal 2018. That business growth was accompanied by a valuation re-rating. Morningstar's price/sales figure increased from 10.85x in 2016 to 13.94x in 2017.

This is the first place where separating fundamentals from multiple matters. A stock can rise because revenue grows, because investors pay more for each dollar of revenue, or both. NVIDIA experienced both.

Then 2018 demonstrated the opposite. NVIDIA's market capitalization fell from about $117 billion at the end of 2017 to about $81 billion at the end of 2018. Morningstar's P/S fell to 6.74x and P/E to 17.85x. The business had not disappeared; the market's willingness to pay for the growth story had changed.

NVIDIA later described significant volatility in gaming demand associated with cryptocurrency mining and its after-effects. In its fiscal 2019 filing, the company specifically said crypto-related demand had distorted gaming trends. That matters because it is a classic case of a secular thesis being temporarily mixed with a cyclical or speculative demand shock.

A weak historical narrative says, "NVIDIA fell in 2018 because the market was wrong." A better one says, "The market had capitalized a high-growth path, one component of demand turned out to be less durable than assumed, channel conditions worsened, and the multiple reset."

Source: SEC: NVIDIA FY2019 Form 10-K.

Phase Two: 2019 to 2021 - NVIDIA Becomes a Long-Duration Growth Asset

By 2019 through 2021, NVIDIA was increasingly valued less like a conventional cyclical chip vendor and more like a long-duration platform company.

The key phrase is long duration. When investors expect a large share of a company's value to come from profits far in the future, changes in discount rates and confidence about long-run growth can move the present value dramatically. That is why valuation multiples can expand much faster than current-year revenue.

Morningstar's calendar-year P/S history moved from 14.52x in 2019 to 22.08x in 2020 and 30.69x in 2021. P/E moved from roughly 60x to 85x and then about 91x.

The company's operating profile was also changing. Fiscal 2020 revenue was $10.92 billion, with Data Center revenue of $2.98 billion. The company was building a larger position in accelerated computing, and the subsequent Mellanox acquisition strengthened networking exposure around high-performance and data-center workloads.

But the valuation ran ahead of the then-reported earnings. That is not automatically irrational; price is forward-looking. It does mean the investment case became more sensitive to future execution.

At a 90x earnings multiple, a company does not merely need to be good. It needs to deliver enough growth, margins and durability for the denominator to catch up before the market changes its mind about the numerator. That is exactly why 2021 is best understood as a valuation-risk peak, even though it was not the peak in NVIDIA's long-run business performance.

Sources: SEC: NVIDIA FY2020 Form 10-K; Morningstar: NVIDIA valuation history.

Phase Three: 2022 - The Business Grows While the Valuation Falls

NVIDIA's 2022 market-cap decline is one of the most useful years in the case study because it breaks the lazy assumption that business growth and stock valuation move together.

Fiscal 2022 revenue reached $26.91 billion. Yet NVIDIA's calendar-year market capitalization fell to about $364 billion from about $735 billion in 2021.

Several things were happening at once.

First, interest rates rose sharply. Higher real and nominal yields reduce the present value of distant cash flows, which hits long-duration growth assets especially hard.

Second, gaming demand and channel inventory were resetting. In the fiscal 2023 filing, NVIDIA said gaming revenue fell 27% year over year and that it was undershipping GPU supply to help normalize channel inventory. The company also recorded significant inventory charges.

Third, the failed Arm transaction produced a large termination charge.

The result was a valuation reset in a company whose long-term strategic position had not disappeared. Morningstar's P/S fell from 30.69x in 2021 to 12.91x in 2022.

This is a clean example of why "the company grew" is not enough to explain a stock. The market was simultaneously revising the discount rate, the near-term earnings path, the durability of some demand, and the price it was willing to pay for future growth.

Source: SEC: NVIDIA FY2023 Form 10-K.

Phase Four: 2023 - The Generative-AI Re-Rating Arrives Before the Full-Year Numbers

The 2023 re-rating is where hindsight becomes especially dangerous.

NVIDIA's fiscal 2023 revenue was essentially flat at $26.97 billion. Net income was only a fraction of what it would become a few years later. Yet the market capitalization ended 2023 at roughly $1.22 trillion.

Why would valuation rise so dramatically before full-year financial statements showed the eventual scale? Because markets discount expectations.

The release of generative-AI products and the rapid buildout of large-language-model infrastructure changed expectations for accelerator demand. NVIDIA's Hopper architecture and H100 product ramp became central to that buildout. Investors were no longer treating data-center acceleration as one attractive vertical among several. They were beginning to price NVIDIA as critical infrastructure for a new computing cycle.

Morningstar's P/S moved back to about 27.5x for 2023. That multiple looked extraordinary against the trailing fiscal 2023 revenue base. It looked less extraordinary after the next fiscal year arrived.

This is the denominator-lag problem. At an inflection point, trailing valuation ratios can make a rapidly changing business look more expensive precisely when the denominator is most stale. That does not make a high multiple safe. It means the analyst must ask whether the denominator is representative. The correct lesson is not "ignore P/E and P/S for fast growers." The lesson is to know what period the denominator represents and what assumptions are needed to replace it with something forward-looking.

Phase Five: 2024 - Data Center Stops Being an Adjacent Business

Fiscal 2024 is the year the business mix visibly broke with the past.

NVIDIA reported $60.92 billion of total revenue, up 126%. Data Center revenue was $47.53 billion, up 217%. Gross margin increased to 72.7%. Operating income reached $32.97 billion, up 681%, and net income reached $29.76 billion.

Those figures changed the meaning of NVIDIA's historical valuation. In 2016, investors could debate whether data-center acceleration would become a large adjacent market. By 2024, Data Center represented roughly 78% of annual revenue. The question was no longer whether the segment mattered. The questions were how large AI infrastructure spending could become, how durable NVIDIA's economics would be, how much competition or custom silicon could capture, and how long unusually high growth and margins could persist.

The market capitalization rose to about $3.29 trillion by the end of 2024. Morningstar shows the P/S ratio around 29.48x and P/E around 53.04x. The apparently paradoxical part is that the P/E was already below its 2021 level even though the company was worth several times more. Earnings were catching up.

Source: SEC: NVIDIA FY2024 Form 10-K.

Phase Six: 2025 to 2026 - Earnings Catch Up With the Market Cap

Fiscal 2025 and 2026 changed the character of NVIDIA's valuation again.

NVIDIA reported $130.50 billion of revenue for fiscal 2025 and $215.94 billion for fiscal 2026. Fiscal 2026 net income reached $120.07 billion.

The fiscal 2026 revenue mix:

• Data Center: $193.74 billion
• Gaming: $16.04 billion
• Professional Visualization: $3.19 billion
• Automotive: $2.35 billion
• OEM and Other: $0.62 billion

Data Center represented almost 90% of fiscal 2026 revenue. This is why the company's 2016 and 2026 valuations are not simply two points on the same business. The ticker is the same. The economic engine is not.

By August 2026, CompaniesMarketCap's series placed NVIDIA above $5.2 trillion. Yet Morningstar's current P/E was roughly 33x, lower than the 2016 figure and dramatically lower than the 2021 figure. A company can become vastly more valuable while its earnings multiple declines if earnings grow faster than market value.

NVIDIA's August 26, 2026 fiscal-Q2 2027 release reinforced the point. Quarterly revenue was $96.2 billion, up 106% from a year earlier, and Data Center revenue was $89.0 billion. A valuation that once required imagining a future data-center business was now being measured against quarterly data-center revenue larger than NVIDIA's entire annual company revenue only a few years earlier.

Sources: SEC: NVIDIA FY2026 Form 10-K, NVIDIA: Q2 FY2027 results, Aug. 26, 2026.

The Valuation Bridge: What Actually Drove the 2016 to 2026 Change?

Instead of asking whether NVIDIA's stock was "cheap" or "expensive," this section splits the decade into three bridges.

Bridge 1: Revenue

Fiscal revenue increased from $5.01 billion to $215.94 billion. Growth factor: about 43.1x. That alone can support a radically larger market value even if the price/sales multiple does not change.

Bridge 2: Net Income

Net income increased from $614 million to $120.07 billion. Growth factor: about 195.5x. That is faster than the market-cap increase across the same rough decade.

Bridge 3: Market Capitalization

Market capitalization increased from approximately $57.5 billion to more than $5.2 trillion. Growth factor: about 91.3x. That is an extraordinary increase, but it sits between revenue growth and net-income growth.

What the Three Bridges Mean

If market cap grew about 91x while revenue grew about 43x, the market ended the period paying a materially higher value per dollar of revenue than at the beginning.

But if net income grew about 196x while market cap grew about 91x, the market ended the period paying a lower value per dollar of earnings than the full-year 2016 comparison suggests.

Those statements are not contradictory. NVIDIA's margin structure transformed. In fiscal 2016, gross margin was 56.1% and net margin was roughly 12%. In fiscal 2026, gross margin was 71.1% and net margin was roughly 56%. The valuation cannot be understood from revenue growth alone because each dollar of revenue in 2026 produced vastly more net income than each dollar in 2016.

That is the durable lesson: valuation multiples are summaries of a business model, not properties of a ticker symbol.

NVIDIA's Valuation Changed Because the Business Changed Categories

The market repeatedly changed the category it used to think about NVIDIA.

2016: High-quality gaming GPU company with adjacent optionality

The market already recognized a strong franchise. Data center and automotive were interesting but not dominant.

2017 to 2019: Accelerated-computing platform with cyclical noise

Data-center growth became more important, but gaming and crypto-linked demand could still distort the picture.

2020 to 2022: Long-duration computing platform

The market paid for a broader platform opportunity, networking, data center and future workloads. That made the stock more sensitive to discount rates and future expectations.

2023 to 2026: AI infrastructure system provider

The center of gravity moved from selling GPUs for discrete applications toward supplying large-scale accelerated-computing infrastructure, networking, systems and software around AI workloads.

The categories matter because a reasonable valuation method changes with them. A mature gaming-hardware multiple is not a sensible anchor for an AI-infrastructure business with materially different growth and margin economics. At the same time, changing the category does not eliminate valuation risk. It changes the questions an analyst must ask.

What Was Knowable at the Time?

Historical analysis is easy when every uncertain branch has already resolved. The purpose of this table is to reconstruct the information set before the outcome.

The Two Stock Splits Matter for Interpretation, Not Valuation

NVIDIA completed a four-for-one stock split in July 2021 and a ten-for-one split in June 2024.

The splits changed the number of shares and the quoted price per share. They did not, by themselves, change the proportionate economic value of a shareholder's ownership. That is why historical prices in this case study should always be split-adjusted.

A reader looking at an old headline quoting a pre-split NVIDIA share price and comparing it directly with a 2026 quote can reach a meaningless conclusion unless the series is adjusted.

Sources: SEC: NVIDIA 2021 four-for-one split filing, NVIDIA: 2024 ten-for-one split announcement.

What Could Go Wrong With a Valuation Analysis Like This?

A historical article can still be misleading even when every number is correct.

Mistake 1: Using only one multiple

P/E is powerful when earnings are representative. It becomes unstable when earnings are temporarily depressed or exploding upward. P/S ignores margin structure. EV/Sales ignores capital intensity. EV/EBITDA can hide stock compensation and reinvestment. No one multiple tells the decade.

Mistake 2: Treating high growth as proof that a high multiple was justified

Observed future growth does not retroactively remove the risk investors took before the growth happened. A 90x P/E can work out and still have been vulnerable to many adverse paths.

Mistake 3: Treating a low multiple as proof of cheapness

A multiple can fall because expectations deteriorate faster than price. It can also be low because earnings are temporarily near a cyclical peak.

Mistake 4: Ignoring business-mix change

NVIDIA's 2026 Data Center business is so large that comparing the company mechanically with its 2016 self can obscure more than it reveals.

Mistake 5: Mixing fiscal and calendar years without disclosure

NVIDIA's fiscal year ends in January. Calendar-year market-cap and valuation data do not line up exactly with fiscal financial statements. That is acceptable for a historical orientation table only if the mismatch is disclosed.

Mistake 6: Confusing a split with value creation

A stock split changes unit size, not enterprise value.

Risks That Matter to the 2026 Valuation

A historical case study should end where the evidence ends, not turn into a price target. The observable risks in NVIDIA's own filings include:

Customer concentration

NVIDIA reported that one direct customer represented 22% of fiscal 2026 revenue and another represented 14%, primarily in Compute & Networking. Concentration does not mean demand will disappear. It means a small number of purchasing decisions can materially affect results.

Export controls and China exposure

NVIDIA's filings describe U.S. export controls as a material constraint on access to the China data-center-compute market. Regulation can change the addressable market even when end demand exists.

Product-transition and supply risk

The company is moving through rapid architecture transitions and operates a complex supply chain. Fiscal 2026 gross margin was affected by a $4.5 billion charge associated with H20 excess inventory and purchase obligations. That is a reminder that even a supply-constrained growth company can mis-estimate product-specific demand.

Competitive and custom-silicon risk

Cloud providers and semiconductor peers have incentives to build alternatives, reduce cost or optimize particular workloads. The relevant question is not whether alternatives exist; it is whether they change NVIDIA's pricing, volume, margin or platform advantage.

Capital-spending durability

A large share of current demand is tied to an enormous infrastructure buildout. If customers slow capital spending, improve utilization, shift to internally designed accelerators or require fewer GPUs per unit of useful compute, growth can decelerate even if AI usage continues to expand.

What This Case Study Does Not Prove

It does not prove NVIDIA was undervalued in 2016. It does not prove a high P/E is harmless. It does not prove AI infrastructure spending will continue at current growth rates. It does not prove another company can repeat the same path. And it does not prove that knowing the winning technology theme is enough to know which security will produce the best return.

The decade shows something narrower and more useful: a security's valuation can move through completely different regimes as its revenue mix, margins, competitive position, growth rate and perceived duration change.

The Practical Lesson for Investors

When reviewing a ten-year winner, do not ask only, "How much did the stock go up?" Build four lines:

1. Revenue growth.
2. Earnings or free-cash-flow growth.
3. Share-count change.
4. Valuation-multiple change.

Then ask what information was available before each re-rating. That process turns a victory story into a reusable research method.

For NVIDIA, the decade is especially instructive because all four lines moved dramatically. Revenue grew. Margins expanded. Earnings grew even faster. The share count was affected by splits but not economically multiplied by them. The market paid changing multiples through multiple booms and resets. The outcome was not one uninterrupted valuation expansion. It was a sequence of expansion, compression, expansion and earnings catch-up. That is more complicated than a simple story. It is also much more useful.

Frequently Asked Questions

References

1. SEC: NVIDIA Corporation Form 10-K for fiscal year ended Jan. 31, 2016
2. SEC: NVIDIA Corporation Form 10-K for fiscal year ended Jan. 27, 2019
3. SEC: NVIDIA Corporation Form 10-K for fiscal year ended Jan. 26, 2020
4. SEC: NVIDIA Corporation Form 10-K for fiscal year ended Jan. 29, 2023
5. SEC: NVIDIA Corporation Form 10-K for fiscal year ended Jan. 28, 2024
6. SEC: NVIDIA Corporation Form 10-K for fiscal year ended Jan. 25, 2026
7. NVIDIA: Second Quarter Fiscal 2027 Results, Aug. 26, 2026
8. NVIDIA: Annual Reports and Proxies
9. Morningstar: NVIDIA Valuation History, snapshot dated Aug. 19, 2026
10. CompaniesMarketCap: NVIDIA Market Capitalization History
11. SEC: NVIDIA 2021 stock-split disclosure
12. NVIDIA: 2024 ten-for-one split announcement

Data as of: Sept. 5, 2026. NVIDIA fiscal years end in January; calendar-year market data do not align perfectly with fiscal reporting periods. Historical market capitalization is sourced to CompaniesMarketCap for consistency. Historical P/E and P/S are sourced to Morningstar's published calendar-year table. No forward price target appears on this page. This page is educational and does not constitute individualized investment advice.