Nvidia Trades at 28.54 Times Earnings With Revenue Doubling and a $5.4T Market Cap

NVDA closed Wednesday at $223.67, down $2.06 or 0.91%, inside a session range of $223.46 to $226.18 on 66.3 million shares against a 134.1 million three-month average. The stock traded $222.96 after the bell, off another $0.71 or 0.32%.

It went into Thursday's session against the worst backdrop the AI complex has seen in weeks. The Nasdaq Composite fell 185.48 points to 26,067.86, a 0.71% decline. The S&P 500 lost 0.61% to 7,590.14. The VIX jumped 9.23% to 17.98. And the 10-year Treasury yield climbed 8 basis points to 4.90%, its highest since November 2023, after August producer prices came in at 5.4% annually against a 5.3% forecast.

The damage inside the AI infrastructure cohort was concentrated and severe. Astera Labs dropped $16.84 to $283.70, down 5.60% on a $49.22 billion market cap and a 142.28 trailing multiple. CoreWeave lost $5.21 to $89.73, off 5.49% on 8.15 million shares against a 28.32 million average. Intel fell $6.04 to $100.20, down 5.69% on 26.5 million shares. Nebius dropped 4.06% to $230.60. Broadcom slipped 1.13% to $364.38.

AMD was the outlier, gaining 3.04% to $521.10.

Nvidia's beta of 2.22 is the number that matters in that environment. A 0.71% Nasdaq decline implies a 1.58% move for a stock with that sensitivity, and the AI cohort delivered considerably worse. Market capitalization sits at $5.401 trillion across 24.10 billion shares outstanding.

There is a mechanical wrinkle in Thursday's price that most screens will not flag. September 10 is the ex-dividend date for Nvidia's $0.25 quarterly payout, payable October 1 to shareholders of record on the same day. That removes a quarter from the reference price at the open before a single share trades on fundamentals.

The 52-week range runs $164.27 to $236.54. Nvidia sits 36% above the low and 5.5% below the high, up 27.41% over twelve months.

Over the August 10 to September 8 window the stock traded between $207.25 and $234.76, averaging $220.42 with a $27.51 spread and a net change of just 1.43%.

A $1.00 Dividend Nobody Notices on a $5.4 Trillion Company

The dividend deserves a paragraph because of what it says about the balance sheet rather than what it delivers to holders.

Nvidia pays $0.25 quarterly, $1.00 annually, for a yield of 0.45%. On a $222.96 share price that is a rounding error, and no investor owns this stock for income.

What it represents is a company generating so much cash that returning capital has become routine. Nvidia returned roughly $26.0 billion to shareholders through buybacks and dividends in the fiscal second quarter alone, with approximately $99.0 billion remaining under its repurchase authorization. Across fiscal 2026 the company returned $41.1 billion, with $58.5 billion remaining at that time — meaning the authorization has been substantially expanded since.

Run the buyback math. $99 billion of remaining authorization against a $5.401 trillion market capitalization retires 1.8% of the company if fully executed. At the $26 billion quarterly pace, Nvidia is repurchasing roughly $104 billion annually — about 1.9% of shares outstanding per year — while simultaneously funding a capital-intensive supply chain.

That combination separates Nvidia from every other name in the AI infrastructure trade. CoreWeave is burning cash to build capacity. Astera Labs trades at 142 times earnings. Alphabet is spending $195 billion to $205 billion in capex and has watched free cash flow margin fall to 9.2%.

Nvidia sells the picks. It does not dig.

The ex-dividend date landing on a day when the Nasdaq fell 0.71% and the AI cohort dropped 5% is coincidental, but it compounds the optical damage. Anyone reading a screen Thursday morning saw a lower reference price that owes $0.25 to mechanics rather than to selling.

Shares outstanding at 24.10 billion, ten-day average volume at 158.08 million, and a three-month average of 134.14 million describe a stock that trades roughly $30 billion of notional per day. Liquidity is not the constraint here.

The constraint is what the market is willing to pay for growth that everyone now agrees is real.

Technicals: $236.54 Above, $207.25 Below, and a $220 Magnet

The chart has been range-bound for a month and the boundaries are precise.

The 52-week high sits at $236.54, roughly 6% above the last print. The recent swing high is $234.76, established during the August 10 to September 8 window. Between those two levels is the resistance band that has capped every advance since the August earnings report.

The low of that same window is $207.25, roughly 7% below current price, and it marks the floor of the post-earnings consolidation. The average across the period was $220.42 — a level the stock has repeatedly gravitated back toward, which makes it the effective centre of gravity for this range.

The full annual range runs $164.27 to $236.54, a 44% spread. Nvidia is currently trading in the upper quarter of it.

Near-term levels are tighter. Wednesday's low at $223.46 held on the close. Below it, $220 is the mean-reversion magnet and $215 is the next shelf. A break of $207.25 would end the consolidation and open the mid-$190s.

Above, $226.18 was Wednesday's session high and the first hurdle. Then $234.76, then $236.54.

The net change of 1.43% across a month with a $27.51 trading range is the definition of a stock going nowhere loudly. Volume tells the same story — Wednesday's 66.3 million shares against a 134.1 million three-month average is roughly half-normal turnover, indicating disinterest rather than distribution.

That is a market waiting for a catalyst rather than one making a decision.

The macro variable is what tips it. At a beta of 2.22 with the 10-year at 4.90% and 62% to 64% odds on a Federal Reserve hike at the September 15–16 meeting, Nvidia carries more rate sensitivity than almost any mega-cap. Friday's consumer price index — consensus 0.4% monthly, 3.4% annually, core at 2.4% — decides whether $207.25 or $236.54 gets tested first.

Historically the stock has been capable of far larger moves. A single session in a prior rate-driven selloff removed $153 billion of market value.

Q2 FY2027: $96.22 Billion and a Fourteenth Consecutive Beat

The fundamental picture from the August 26 report is the strongest in the company's history, and it is worth laying out precisely because the stock has not moved since.

Revenue came in at $96.22 billion, up 106% year over year. That beat the $92.07 billion consensus from 41 covering analysts by 4.5% and beat the company's own $91.0 billion guidance midpoint by 5.7%. It marked the fourteenth consecutive quarter that Nvidia has printed above its own outlook.

Non-GAAP earnings per share reached $2.22 against a $2.09 to $2.10 estimate. GAAP diluted EPS hit $2.46, up 128% from the prior-year quarter. Both GAAP and non-GAAP gross margins came in at exactly 75.0%.

Hold on that margin figure. A company doubling revenue year over year while holding a 75% gross margin is not experiencing competitive pressure in any measurable form. Margin is where competition shows up first, and it has not appeared.

The prior fiscal year provides the trajectory. Fiscal 2026 revenue reached $215.9 billion, up 65%, with fourth-quarter revenue of $68.1 billion representing 20% sequential and 73% annual growth. Full-year GAAP EPS was $4.90 and non-GAAP $4.77.

Compare the fourth quarter of fiscal 2026 at $68.1 billion to the second quarter of fiscal 2027 at $96.22 billion. That is 41% growth in two quarters at a revenue base approaching $100 billion per quarter.

Annualising the current run rate puts Nvidia near $385 billion of revenue before any of the guided sequential step is counted.

The full release carries the segment detail, and the earnings call framing from the chief executive was blunt: AI has reached its inflection point, it is doing useful work, its tokens are productive and profitable, and compute is now revenue.

That last phrase is the argument that keeps a $5.4 trillion valuation intact.

Data Center at $89.02 Billion Is 92% of the Company

The concentration is extreme and it cuts both ways.

Data Center revenue reached $89.02 billion in the quarter, up 117% year over year and representing roughly 92% of total company revenue — unchanged from the prior quarter's share. Data Center Networking rose 138% year over year, growing faster than the compute business itself.

That networking figure is the underappreciated one. It signals that the buildout extends well beyond chips: switches, interconnect, NVLink and the fabric that ties GPUs into a coherent system are growing faster than the GPUs. A company selling only accelerators would show networking lagging. Nvidia's shows the opposite.

Blackwell Ultra drove most of the acceleration, with hyperscale customers ramping deployments at a pace that more than doubled the segment's revenue year over year.

The concentration risk is straightforward. A business where 92% of revenue comes from one segment serving a handful of hyperscale customers has no diversification if that spending pauses. Gaming, professional visualisation and automotive together represent 8% of the company and cannot offset a Data Center deceleration.

The offsetting fact is that hyperscale capital expenditure has kept rising rather than plateauing. Alphabet raised full-year 2026 capex guidance to $195 billion to $205 billion, from $180 billion to $190 billion, and committed a further €13 billion to Finnish AI infrastructure this week. Oracle reports Thursday evening with a $638 billion remaining performance obligation. These are Nvidia's customers, and their spending plans are getting larger.

Blackwell revenue is projected to jump from roughly $86.4 billion last year to $135.7 billion this year. Estimates for the B-series in the second quarter alone ranged from $6.3 billion to $76.3 billion before the print, with consensus at $38.4 billion — a spread that captures how little visibility the covering community has into product-level mix.

China is effectively gone. Export controls reduced Nvidia's revenue share from that market to a single-digit percentage, and the current guidance assumes zero China compute entirely.

A $108 Billion Guide With Zero China Assumed

The forward number is what moves the stock, and it was the release.

Nvidia guided fiscal third-quarter revenue to $108.0 billion — a sequential step of $11.8 billion from the $96.22 billion just reported. That guide came in 4.0% above the street, and it assumes no compute revenue from China whatsoever.

Sequential growth of 12.3% at a $96 billion base is the part that should command attention. Companies of this scale do not typically add the equivalent of a large-cap company's annual revenue in a single quarter.

The guide is also conservative by construction. Zero China assumed means any policy shift that reopens that market is pure upside not reflected in the number. Nvidia's own beat cadence — 5.7% above the guide midpoint this quarter, against a 4% to 5% recent cadence and an 8.1% average across the full run — implies a realistic fourth-quarter print somewhere between $112 billion and $117 billion if the pattern holds.

Fourteen consecutive quarters above guidance is not luck. It is a management team that sets a number it intends to clear.

The risk sits in the reaction function rather than the results. The stock has a documented history of falling after earnings even on beats — four consecutive post-earnings declines at one stretch, and an 8% drop after a quarter that beat on revenue, earnings and guidance. At a $5.4 trillion market capitalization, breaking records is no longer sufficient; the market prices perfection and sells anything less.

That is precisely why the stock has gone nowhere since August 26 despite the strongest quarter in its history. Revenue doubled. EPS grew 128% on a GAAP basis. Margin held at 75%. And NVDA is up 1.43% over the following month.

The market has stopped paying for beats and started demanding evidence that the guidance is sustainable past 2027.

$279 Billion of Supply Commitments Already Locked

The single most underdiscussed line in the quarter is the balance sheet obligation, and it reframes the entire debate.

Supply commitments have swelled to $279 billion, predominantly memory for Vera Rubin.

That figure is the answer to the question the market keeps asking. Investors worry that AI infrastructure spending may be peaking faster than expected and that near-term revenues cannot justify unprecedented spending levels. A company that has contractually committed $279 billion to its own supply chain is not a company hedging against a demand collapse — it is a company that has seen enough forward orders to underwrite the working capital.

The obligation cuts both ways, which is the honest framing. $279 billion of committed purchases is a fixed cost against variable revenue. If hyperscale spending decelerates in 2027, Nvidia carries memory and wafer commitments it cannot cancel, and the 75% gross margin becomes considerably harder to defend.

The memory concentration is deliberate. High-bandwidth memory has been the binding constraint on GPU output for two years, and securing multi-year supply removed it. Reporting through the Asian supply chain indicated increased 4-nanometer wafer starts for Blackwell, suggesting current revenue models may be conservative on the supply assumption. HBM supply has been secured for 2026 and 2027.

With memory bottlenecks easing, the constraint shifts. Land, power and data center shells become the new limiting factor — a physical infrastructure problem rather than a semiconductor one, and one Nvidia cannot solve with a purchase order.

Separately, reporting indicated Nvidia was working with major financial firms to structure a $500 billion AI infrastructure funding package, a development that knocked $130 billion off market value in a single session on concentration-risk concerns. Vendor financing of customer purchases is the mechanism that turns a supply chain into a credit exposure.

The stock fell 2.4% to around $218 on that report.

$40 Billion Per Gigawatt: The Vera Rubin TAM Step

The clearest framework management has offered is revenue per gigawatt of data center capacity, and the progression is the bull case in one line.

With Hopper, Nvidia's revenue opportunity ran roughly $18 billion per gigawatt. With Grace Blackwell it rose to about $25 billion. With Vera Rubin — spanning the Vera CPU, Rubin GPU, NVLink and networking — it reaches approximately $40 billion.

That is a 122% increase in monetisation per unit of power delivered, across two product generations, without requiring a single additional gigawatt to be built. Every existing customer commitment measured in megawatts converts to more Nvidia revenue simply because the platform captures more of the rack.

Vera Rubin is now in full production, with partners supplying systems during the second half of 2026 and a larger ramp expected through the fourth quarter and into 2027. The company claims Rubin reduces inference token costs by up to ten times and trains large mixture-of-experts models using four times fewer GPUs than Blackwell — performance claims that require commercial validation.

Early adopters named at announcement include AWS, Google Cloud, Microsoft, Oracle, CoreWeave and multiple AI laboratories. Third-quarter Rubin shipments have been projected around 150,000 units, worth roughly $9 billion, with full-year Rubin consensus near $45.3 billion.

The transition risk is real and specific. If customers defer Blackwell purchases while waiting for Rubin, a generational air pocket opens in the current quarter. Blackwell revenue is forecast to jump to $135.7 billion this year and then drop 71% next year — the mechanical consequence of a product cycle handoff.

The chief executive's framing of the demand side has shifted too. The move to agentic AI requires 15 to 100 times more compute than the previous generation of workloads, and where one lab drove the buildout a year ago, multiple frontier labs are now scaling in parallel alongside an open-model ecosystem and physical AI.

Whether that is analysis or salesmanship is the question the next two quarters answer.

AWS, Two Million GPUs, and Visibility Into 2028

The customer commitments announced alongside the quarter extend visibility further than any prior cycle.

AWS plans to deploy 2 million additional Blackwell Ultra, Rubin and Rubin Ultra GPUs across its global infrastructure during 2027 and 2028, including dedicated AI factories. AWS demand has already exceeded the 1 million GPUs pledged at GTC 2026, meaning the new commitment doubles an already-exceeded target.

The scope goes beyond chip volume. The collaboration covers U.S. government AI factories with 100,000 GPUs on secure AWS infrastructure, Vera CPU-based infrastructure on AWS, extended NVLink Fusion with custom high-bandwidth memory inside Trainium racks, integration with AWS Nitro and EFA, expanded Nemotron model support, GPU-accelerated data processing on Amazon EMR and OpenSearch, and physical AI robotics through Amazon Robotics adopting Jetson, Omniverse and Isaac.

The Trainium detail is the significant one. AWS builds its own custom silicon and has positioned Trainium as an Nvidia alternative — and it is now putting NVLink Fusion with Nvidia memory technology inside those racks. The competitor is licensing the interconnect.

The pipeline elsewhere is broad. A $3.1 billion GPU deal backing Indonesia's AI expansion. AI infrastructure capacity expansion across Australia through local partnerships, targeting a significant buildout by 2027. A deepened MediaTek partnership spanning edge to cloud. An enterprise AI platform collaboration with HP and Red Hat. Jetson Orin Nano 2 for entry-level edge AI, and Groq 3 LPX in full production for agentic workloads.

And the largest acquisition in company history: Hugging Face, at roughly $12.9 billion to $14 billion including a $1 billion employee retention package. Hugging Face is where developers share and access open-source AI models — the distribution layer for the ecosystem.

Buying the model repository after selling the compute is a move to control adjacent layers of the stack. It is also an acknowledgment that hardware alone does not build a moat when custom ASICs are improving.

Valuation: 28.54 Times Earnings on $5.4 Trillion

The multiple is the most surprising number in this entire analysis, and it is the strongest argument for the stock.

Nvidia trades at a trailing price-to-earnings ratio of 28.54 with a market capitalization of $5.401 trillion. Enterprise value to EBITDA runs approximately 24.5 times.

For context, Apple trades at 36.28 times trailing earnings while guiding to 9% to 11% revenue growth and margin compression. Nvidia trades at 28.54 times while growing revenue 106% year over year at a 75% gross margin with a guide implying 12.3% sequential growth.

The multiple has compressed materially. Nvidia previously traded at forward multiples exceeding 35 times, and the rate-driven selloffs of the past two months have brought the valuation down without the earnings following. Growth accelerated. The multiple fell. That is the definition of de-rating on macro rather than fundamentals.

Run a simple forward calculation. Four quarters at the guided $108 billion run rate produces $432 billion of annual revenue. At a 75% gross margin and the operating leverage demonstrated this year, forward earnings support a multiple in the low twenties on next year's numbers.

The bear response is that trailing multiples on a hypergrowth cycle are meaningless because the denominator is unsustainable. If revenue growth decelerates from 106% to 20% in fiscal 2028, a 28.54 multiple on peak earnings is expensive rather than cheap. The $279 billion of supply commitments becomes the fixed cost that turns a margin story into a margin problem.

Both arguments are defensible. What is not defensible is calling Nvidia expensive on the current numbers — at 28.54 times with 106% growth, the PEG ratio is a fraction of the market's.

The Philadelphia semiconductor index has risen 92% over twelve months despite a near-20% drawdown at one point, and 20 of the world's most valuable chip stocks shed $1.3 trillion in a single July week. Volatility in this sector is not a signal about the underlying businesses.

A $327 Average Target Against a $180-to-$515 Range

Sell-side positioning is near-unanimous and the dispersion is extraordinary.

Fifty-eight analysts recommend buying the stock and one recommends selling, producing an aggregate Strong Buy. The average twelve-month price target sits at $327.18, implying 46.28% upside from $223.67. The high estimate is $515. The low is $180.

A $335 spread between the highest and lowest target — 186% of the low estimate — on a $5.4 trillion company is not a research disagreement about a rounding error. It is a fundamental split on whether AI infrastructure spending is a decade-long buildout or a two-year capex cycle.

The $515 target implies a market capitalization above $12 trillion. The $180 target implies $4.3 trillion, a 19% decline from current levels. Both come from analysts covering the same disclosed financials.

Compare that to Apple, where 44 analysts produce an average target of $324.53 sitting 2.89% above spot with a $290 to $400 range. Apple's covering community broadly agrees. Nvidia's does not.

The 46.28% implied upside is also unusual for a company this large. Sell-side targets on mega-caps typically cluster within 10% to 15% of spot because the analysts anchor to price. A 46% average gap means the models are running well ahead of what the market will pay.

That is either a signal that the stock is materially mispriced or a signal that the models have not been marked to a higher discount rate. With the 10-year at 4.90% and Nvidia carrying a beta of 2.22, the second explanation deserves weight.

Historical target revisions offer a caution. Following an earlier quarter, one target was raised to $220 from $215 with a Hold rating maintained — acknowledging fundamental strength while flagging the debate over sustainability of elevated growth rates. That framing has recurred throughout the cycle: nobody doubts the numbers, everybody doubts the durability.

Competition, Rental Economics, and the Bear Case

The credible bear arguments are worth stating precisely because they are not about demand.

The first is unit economics downstream. A prominent short-seller has publicly challenged Nvidia GPU rental economics as H100 hourly rates jumped 22% — the argument being that if rental prices are rising while newer, cheaper-per-token silicon arrives, the depreciation assumptions embedded in every neocloud's business model are wrong. If GPU renters cannot earn their cost of capital, they stop buying regardless of how good the chip is.

The chief executive has answered this directly, arguing that because Nvidia compute is so productive, the tokens customers generate and the GPU hours they rent out are highly profitable with strong margins. That is an assertion about customer economics from the supplier, and it requires independent verification.

The second is competitive. MLPerf benchmark results have positioned AMD's Instinct line and custom silicon against Blackwell, and AMD gained 3.04% to $521.10 Thursday while Nvidia fell — the only meaningful outperformance in the cohort. AMD's data center segment expanded 107% year over year to $6.7 billion in a recent quarter. Custom ASICs from hyperscalers remain the structural threat.

The third is geopolitical. Export controls cut Nvidia's China revenue to single digits, and U.S. officials have said China is using American AI models to close the AI gap — a framing that could tighten restrictions further rather than loosen them.

The fourth is the physical constraint. With memory secured for 2026 and 2027, land, power and data center shells become the bottleneck. Nvidia can produce chips faster than customers can energise buildings to house them.

The counter to all four is the same figure: $279 billion of supply commitments and a $108 billion quarterly guide with zero China. Management has committed real money against real orders.

Rubin's shipping cadence, customer feedback and competitiveness relative to ASICs and AMD still require data to validate. The fiscal third-quarter report is when that data arrives.

A 4.90% Ten-Year Against a 2.22 Beta

The macro is doing more to this stock right now than the fundamentals, and the mechanism is arithmetic.

Nvidia carries a beta of 2.22 on a five-year measure. That means the stock has historically moved 2.22 times the market. The 10-year Treasury yield at 4.90% and the two-year at 4.36% raise the discount rate applied to cash flows that arrive in 2028 and beyond — and the entire $327 average price target rests on cash flows that far out.

August producer prices at 5.4% annually pushed Federal Reserve hike odds to between 62% and 64% for the September 15–16 meeting. Core PPI came in softer at 0.2% monthly against a 0.3% forecast, and jobless claims printed 206,000 against 205,000 expected. Friday's consumer price index, with core expected at 2.4%, resolves it.

The precedent is documented. Rate-driven selloffs have removed $153 billion of Nvidia market value in a single session and $238 billion across a few days. Twenty of the world's most valuable chip stocks shed $1.3 trillion inside a week when investors reassessed whether near-term revenues justify unprecedented AI spending levels.

The framing that emerged from that episode is the right one: the selling was less about weakening AI demand and more about a repricing of expectations after an exceptionally strong rally.

Thursday's tape was the same pattern in miniature. Astera Labs at 142 times earnings fell 5.60%. CoreWeave, burning cash to build capacity, fell 5.49%. Intel fell 5.69%. Nvidia, at 28.54 times earnings with $99 billion of buyback authorization and 75% gross margins, held up better than all of them.

That relative resilience is the argument for owning the highest-quality asset in a leveraged sector during a rate shock. Nvidia is the only name in the AI complex that generates enough cash to buy back nearly 2% of itself annually while funding a $279 billion supply chain.

Oracle's report after Thursday's close, with a $638 billion RPO, is the immediate read-through on whether the customer spending underwriting all of it is converting.

Verdict and Forecast: The Cheapest Hypergrowth Stock in the Market

NVDA at $223.67 with a $222.96 after-hours print is a stock where the fundamentals have accelerated and the price has not moved for a month.

The evidence is not ambiguous. Revenue of $96.22 billion growing 106% year over year, beating consensus by 4.5% and the company's own guide by 5.7% for a fourteenth consecutive quarter. Data Center at $89.02 billion, up 117%, with networking growing 138%. Gross margin at exactly 75.0% on both GAAP and non-GAAP measures — no competitive erosion whatsoever. GAAP EPS up 128%. A third-quarter guide of $108.0 billion, a $11.8 billion sequential step, assuming zero China compute. $279 billion of supply commitments locked in. $26 billion returned to shareholders in one quarter with $99 billion of authorization remaining. Revenue per gigawatt rising from $18 billion on Hopper to $40 billion on Vera Rubin. AWS committing 2 million GPUs across 2027 and 2028.

And the stock trades at 28.54 times trailing earnings, up 1.43% over the past month, with an average sell-side target of $327.18 sitting 46% above spot.

What is holding it back is not the company. It is a 10-year at 4.90%, 64% odds on a Federal Reserve hike, a 2.22 beta, and a market that has stopped paying for beats and started demanding proof that fiscal 2028 looks like fiscal 2027.

The forecast splits by horizon. Into Friday's CPI and next week's FOMC, the range is $207.25 to $236.54, with $220.42 the mean the stock keeps returning to. A hot core CPI takes NVDA toward $210 and possibly through $207.25 given the beta; a soft print sends it at $234.76 and then the $236.54 high. The $0.25 ex-dividend on Thursday accounts for a quarter of any decline and should be netted out.

Beyond that, the fiscal third-quarter report is the event. A $108 billion guide delivered at or above plan, with Rubin shipping on schedule and margin holding at 75%, breaks $236.54 and puts $260 in play. A guide that decelerates, or a Rubin transition gap that opens as customers defer Blackwell, sends the stock to $195.

Buying weakness toward $210 with the October print as the catalyst is the trade. At 28.54 times earnings growing 106%, the valuation is not the risk — the discount rate is, and that resolves next Wednesday.

That's TradingNEWS