Nvidia Trades Flat at $225 on 75.68M Shares as AMD Jumps 6.5% and AVGO Drops 5.9%

Nvidia closed Friday at $225.16, down $0.14 or 0.06%, then eased to $224.72 in after-hours trade on a further $0.44 decline. Premarket Monday returned it to $225.16, unchanged on the session, on 75.68 million shares. That volume figure was the largest of any single name on the board, and it arrived alongside a session in which Nasdaq 100 futures climbed 139 points, or 0.5%.

The leadership sat elsewhere entirely. SanDisk jumped 7.39% to $1,641.11, Micron added 2.30% to $971.66, Seagate gained 5.65% to $973.44 and Advanced Micro Devices rose 6.50% to $514.39. Against those, Broadcom fell 5.94% to $392.99 and Applied Materials dropped 5.12% to $507.18. A 12-point spread between AMD and Broadcom inside one premarket session describes capital reallocation within semiconductors rather than a sector-wide move.

The catalyst was a private company's revenue disclosure. Anthropic reported preliminary revenue above $11.5 billion for its latest completed quarter, against $787 million a year earlier and $4.73 billion in the first quarter of 2026, alongside positive adjusted operating income. That reinforced the view that heavy artificial intelligence spending will be sustained, lifting the memory and storage complex that supplies the physical inputs for frontier model training. Nebius Group gained 8.88% to $277.68 on the same read.

Nvidia captured none of it. A stock that supplies the accelerators at the centre of that demand thesis finished flat on the heaviest volume in the market while its suppliers and its nearest competitor rallied between 2% and 7%. That combination is the defining fact of the session, and it has a specific interpretation: extreme turnover paired with negligible price change identifies Nvidia as the funding source for the rotation rather than a participant in it. Traders sold the crowded position to buy the memory names, and they did so nine sessions before Nvidia reports. Broad-based selling into a sector rally ahead of a binary event is position management, not a verdict on the business.

Institutional Flows Split as 13F Filings Land

The most recent quarterly disclosures document a genuine two-way market in Nvidia's shareholder base, which is unusual for a stock with near-unanimous analyst support. Mirova US cut its stake by 746,110 shares. Coatue Management trimmed 276,423. Gilder Gagnon Howe reduced its holding by 129,976 shares, Maverick Capital sold 118,134, and PRIMECAP Management trimmed roughly 109,000.

Buying came from a different cohort. Altimeter Capital's Brad Gerstner added 77,550 shares. Appaloosa Management boosted its position by 53,500. Clough Capital Partners added 4,975. Smaller reductions came from Chilton Investment at 1,079 shares and Driehaus Capital at 1,028. The aggregate across the disclosed names skews net negative by roughly a million shares, though the sample is not exhaustive and quarter-end snapshots reveal position size rather than intent.

The composition of that split matters more than the net figure. Reductions concentrated among long-only managers and multi-strategy funds with large existing positions, while additions came from concentrated growth investors adding to conviction holdings. That is the signature of a stock transitioning from a momentum vehicle to a fundamental one, where the marginal seller is rebalancing an oversized weight and the marginal buyer is adding on valuation.

Institutional ownership has historically run above 65% of the float, which means these disclosures capture the behaviour of the dominant holder class rather than a fringe. What they do not capture is the past six weeks, since the filings reflect positions as of quarter-end. Nvidia has traded from roughly $193 to $225.16 across the intervening period, a 16.7% advance, and the buyers who drove that move are not visible in the data. The 75.68 million shares changing hands premarket Monday will appear in the next round of filings, and the direction of that flow is what determines whether the current level holds through the report.

Nine Sessions to Earnings: August 26 After the Close

Nvidia reports second-quarter fiscal 2027 results on Wednesday, August 26, after the market closes, with the date confirmed by the company. The fiscal quarter ended July 26, 2026. The press release is scheduled for approximately 1:20 p.m. Pacific Time with the conference call following at 2:00 p.m. Pacific, which places the release at 4:20 p.m. Eastern and the call at 5:00 p.m. Eastern.

The timing sits inside a dense macro window. Minutes from the July 28-29 FOMC meeting arrive Wednesday of this week, and Federal Reserve Chair Kevin Warsh speaks at the Jackson Hole Economic Policy Symposium running August 27 to 29, which begins the day after Nvidia reports. That sequencing means Nvidia's guidance lands into a market simultaneously repricing monetary policy, with the federal funds target at 3.50% to 3.75% and September hike odds having collapsed to roughly 25% from near 50% a week ago.

Positioning into the print reflects the calendar rather than conviction. Nvidia's flat performance on record volume, against a memory complex up 7% and Nasdaq 100 futures up 0.5%, is what a market does when it wants exposure to the AI theme without carrying single-name event risk into a binary release. Capital rotates to the suppliers, which participate in the same demand thesis without the earnings date.

The information asymmetry heading into the report is smaller than usual, which raises rather than lowers the risk. Management has already guided to approximately $91.0 billion in revenue with a plus or minus 2% band, published gross margin targets of 74.9% GAAP and 75.0% non-GAAP, and specified operating expenses of $8.5 billion GAAP and $8.3 billion non-GAAP. With the quarter's revenue effectively pre-announced, the next-quarter outlook, Blackwell Ultra execution, gross-margin durability, supply availability and hyperscaler demand commentary all carry more weight than whether the company narrowly clears its own number.

The $91 Billion Guide Against $91.90 Billion Consensus

Consensus sits marginally above guidance, which is the standard configuration for Nvidia and the source of its recurring post-earnings problem. Wall Street expects earnings of $2.08 per share against $1.05 in the year-ago quarter, nearly double, on revenue of $91.90 billion representing roughly 97% annual growth. Management guided to $91.0 billion at the midpoint. Some estimates run higher, with a range of $93 billion to $95 billion circulating and implying approximately 96% growth.

The comparison base is what generates those growth rates. Q2 FY2026 revenue was $46.7 billion. Doubling that figure in twelve months on a base measured in tens of billions of dollars is the achievement embedded in the consensus, and it is why the implied year-over-year growth reads at 96%. It also means the comparison hardens dramatically from here. Every subsequent quarter laps a base above $80 billion.

Full-year estimates frame the multi-year picture. FY2027 consensus revenue sits at $391.3 billion with EPS estimates spanning $8.79 to $9.34, the former representing 92% growth over FY2026. FY2028 EPS is projected near $12.12, implying a further 38% advance. At $225.16, Nvidia trades at roughly 25.6 times the $8.79 FY2027 estimate and approximately 18.6 times the FY2028 figure.

Those multiples are the strongest element of the bull case and the least discussed. A company growing earnings 92% this fiscal year and 38% next, at a 25.6 times forward multiple, is not priced for perfection by any conventional measure. The market is applying a discount, and the components of that discount are identifiable: China export restrictions, execution risk across the Blackwell-to-Rubin transition, and questions about whether hyperscaler capital expenditure now approaching $730 billion annually can sustain its growth rate. Achieving the Q2 guide and raising Q3 guidance toward or above $100 billion would keep the annual consensus intact and force the market to justify the discount rather than the multiple.

Q1 FY2027 Delivered $81.6 Billion With Data Center at $75.2 Billion

The prior quarter established the run rate the August report must extend. Nvidia reported Q1 FY2027 revenue of $81.6 billion on May 20, 2026, up 85% year over year and 20% sequentially, driven by demand for AI infrastructure and particularly Blackwell systems. Earnings for the quarter reached $58.32 billion. Both figures set company records.

Data center supplied nearly the entire business. That segment generated $75.2 billion, up 92% year over year and 21% sequentially, which is 92% of total revenue. The internal composition has shifted meaningfully: hyperscale and AI cloud plus enterprise segments now split roughly 50/50. That balance matters because it reduces dependence on a handful of hyperscaler capital budgets and broadens the customer base toward neoclouds, sovereign programmes and enterprises.

The competitive gap at that scale is difficult to overstate. Nvidia's $75.2 billion data center quarter compares against AMD's $6.72 billion and Intel's $6.3 billion in their respective periods. Nvidia's data center segment alone is roughly eleven times AMD's entire data center business. Market share arguments that treat AMD and Intel as immediate substitutes have to contend with an order-of-magnitude difference in delivered volume.

Sequential growth is the metric that determines whether the multiple holds. Twenty percent quarter-over-quarter expansion at an $81.6 billion base means adding roughly $16 billion of quarterly revenue in three months, which is more than AMD's and Intel's data center businesses combined. Guidance of $91.0 billion implies 11.5% sequential growth, a deceleration from 20%, and that deceleration is arithmetically unavoidable as the base compounds. The market's tolerance for slowing sequential growth at high absolute levels is the question the August 26 report answers, and it is a different question from whether demand remains strong.

Vera Rubin Enters Full Production With Q3 Shipments

The platform transition is the primary execution risk and the primary growth catalyst simultaneously. Nvidia introduced its Vera Rubin platform in Q1 FY2027 materials, including the Vera CPU and BlueField-4 STX, with Jensen Huang presenting the Vera Rubin AI system at Computex 2026 in Taipei on June 1. Management has indicated Vera Rubin is in full production with shipments beginning in the third quarter.

That timeline places the ramp directly inside the guidance window Nvidia issues on August 26. Blackwell delivers substantial performance improvements over Hopper with particular emphasis on inference workloads and has been adopted by major cloud providers, and the Hopper-to-Blackwell transition was the central supply chain story of the first half of 2026. Rubin now becomes the next transition, and the market has already priced execution risk around it.

Analyst focus reflects that. Goldman Sachs analyst James Schneider, who reiterated a Buy rating with a $285 price target implying 26.5% upside, will be looking for evidence that Rubin is gaining adoption among new customer groups including AI labs, neocloud providers and sovereign customers. A strong ramp would provide another leg of growth following the Blackwell rollout. Schneider also expects investors to focus on Nvidia's expanding CPU business as agentic AI drives demand for additional computing infrastructure.

Supply-side commentary carries asymmetric consequences. Confirmation that Blackwell system yields have improved and shipment velocity matches the demand signals hyperscalers communicated on their own calls would be a strong positive. Any indication of supply friction or customer delivery delays would be a clear negative, and would arrive at a moment when the stock trades 4.8% below its all-time high with a 25.6 times forward multiple that already embeds a discount. The company is also releasing a lightweight open-source AI model capable of running on a single GPU, which extends the platform argument down-market without moving near-term revenue.

Four Straight Post-Earnings Declines Despite Four Straight Beats

The most actionable pattern in Nvidia's recent history is the disconnect between results and price. The stock has fallen after each of its past four earnings reports despite beating Wall Street estimates every time. That includes the Q1 FY2027 print in May, when shares declined despite 85% revenue growth and a beat on both the top and bottom lines.

The mechanism is expectations rather than execution. When 43 of 47 analysts rate the stock a strong buy, the risk shifts from missing estimates to merely meeting them, and a company of Nvidia's scale that simply meets expectations tends to get sold. Investors appear to price the upside before results land, which converts a beat into a non-event and a slight miss into a repricing. The setup remains favourable, but high expectations raise the bar for another post-earnings rally.

The historical comparison illustrates how much the bar has moved. In August 2025, Nvidia reported Q2 FY2026 EPS of $1.05 against $1.01 consensus, a four-cent beat, on revenue of $46.74 billion versus $45.65 billion expected and 55.6% annual growth. Those numbers would have been extraordinary at almost any other company. This August the comparable expectation is $2.08 on $91.90 billion with 97% growth, and a four-cent beat on that base is unlikely to move the stock without accompanying guidance.

What breaks the streak is Q3 guidance rather than Q2 results. Raising the next-quarter outlook toward or above $100 billion, combined with gross margins holding near the 75% non-GAAP target and evidence of Rubin adoption across new customer categories, is the combination that would deliver a positive reaction. Absent that, the pattern of the past four quarters is the base case, and options markets appear to agree.

Options Price a 6.99% Move Against a $180 to $500 Target Range

Derivatives markets are pricing a substantial post-earnings dislocation. Options imply a 6.99% move following the August 26 report, which on a $225.16 share price translates to roughly $15.74 in either direction, framing a range of approximately $209 to $241. The upper boundary of that range sits above the all-time high of $236.54 reached in May 2026, which means a positive reaction of the implied magnitude would establish a record.

Analyst targets show extraordinary dispersion. The average sits between $302.83 and $309.94 depending on the compiler, with Barchart tracking $304.32. That implies upside between 34.5% and 37.6% from current levels. The low target is $180, representing 20% downside, and the high is $500, representing 122% upside. A $320 spread between the extremes on a $225 stock reflects genuine disagreement about the durability of the AI capital expenditure cycle rather than modelling noise.

Rating distribution is close to unanimous. TipRanks counts a Strong Buy consensus based on 36 Buys and one Hold. Barchart tracks 43 of 47 analysts at strong buy. Wells Fargo reiterated Overweight on August 11. Goldman Sachs holds Buy at $285. When coverage is this one-sided, the rating itself carries little information, and the target dispersion becomes the more useful signal.

Relative performance explains part of the discount the low targets imply. Nvidia has returned roughly 20.88% over the trailing twelve months against 13.74% for the S&P 500, and roughly 20% over the past year while trailing the Technology Select Sector SPDR Fund. Over three years the stock has gained 415.62% against 73.41% for the index, and over five years 1,019.17% against 74.26%. Outperforming the index while lagging its own sector, at a $5.3 trillion market capitalisation, is the profile of a stock digesting prior gains rather than one in distribution.

The $500 Billion Wall Street Financing Platform

Nvidia's most consequential recent announcement had nothing to do with silicon. The company signed memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish independent compute financing platforms aimed at mobilising more than $500 billion of third-party capital for AI infrastructure over time. Each firm will assemble capital pools at rates Nvidia characterised as attractive, with intended beneficiaries spanning frontier AI labs, enterprises and cloud providers.

The strategic reframing is explicit. Huang described Nvidia's AI factory platform as an investable infrastructure asset, arguing it is productive, revenue-generating, fungible, used by nearly every cloud service provider and capable of running every AI model. He characterised the moment as the first time technology chips have become an investable asset class, and wrote that the industry has moved from buying chips and building data centres project by project to financing AI factories as productive infrastructure.

The purpose is to widen the customer base beyond entities with investment-grade balance sheets. The platforms are partly intended to make it easier for smaller, start-up AI companies to borrow to buy compute, addressing companies that lack the credit rating or cash to purchase silicon outright. Apollo President Jim Zelter described modern compute as a scarce, mission-critical asset class with compelling investment characteristics. Huang stated he approached only those six firms and none declined.

The qualifications matter as much as the headline. Preliminary agreements are not contracts. Nvidia disclosed no financial terms, no investment commitments from individual firms and no schedule for deploying capital. The $500 billion figure is a target rather than committed money. The context is that the first three years of the AI expansion ran on corporate balance sheets, Big Tech capital spending is set to pass $730 billion this year, and several of the largest spenders have turned cash-flow negative reaching that level. Extending the demand story into credit markets is a response to that constraint rather than proof it has been solved.

The $125 Billion Backstop Transfers Credit Risk Onto Nvidia

The detail that changes Nvidia's risk profile appeared after the initial announcement. Huang wrote that Nvidia holds an option to backstop up to $125 billion of the $500 billion total, roughly a quarter of potential deals. That means Nvidia can guarantee up to 25% of qualifying loans, exposing the company to approximately $125 billion if the underlying projects underperform.

That is a material change in what Nvidia is. A chip supplier sells hardware and recognises revenue. A supplier that guarantees a quarter of the debt financing its customers use to buy that hardware has taken on credit exposure to its own demand. The structure resembles vendor financing, and vendor financing has a documented history of inflating reported demand during expansions and accelerating losses during contractions.

The circularity concern is already live in the market. Investors have grown uneasy about the volume of debt-financed investment moving through the AI sector, and the circular character of many arrangements, where one AI company invests in a second on the understanding that the second will buy the first's products, has raised questions about whether demand is artificially inflated. A backstop from the primary supplier to the lenders funding purchases of that supplier's product sits squarely inside that concern.

The timing was deliberate. The announcement followed market volatility in July 2026, when investors began questioning whether Big Tech's AI investments would generate adequate returns. Providing an external financing mechanism validates the deployment thesis while reducing balance sheet pressure on Nvidia's largest customers, and it aligns the interests of Nvidia, its customers and capital providers. It also means that if the AI capital cycle turns, Nvidia absorbs losses on two fronts simultaneously: falling hardware orders and impairment on guaranteed loans. Investors will want detail on the structure when management takes questions on August 26, and Goldman Sachs has flagged it as a priority topic.

GPU Residual Value Is the Question Asset-Backed Lending Cannot Answer

The financing structure rests on an assumption that has never been tested at scale. In standard asset-backed finance, a lender advances money because a default allows repossession and resale of the underlying asset. Buildings, warehouses and cargo ships have established secondary markets and useful lives measured in decades. The productive lifespan of cutting-edge GPUs is far from settled.

The depreciation path is the specific problem. New chips power frontier model training, then after a few years get relegated to lower-margin inference work. That migration directly affects the collateral value underpinning any loan written against them. A lender underwriting a five-year facility secured by Blackwell systems is making an implicit forecast about what those systems fetch in year four, at which point Rubin and its successor will both have shipped.

Nvidia's own product cadence works against the collateral thesis. Annual platform transitions are the company's competitive strategy and the source of its pricing power, and each transition compresses the residual value of the prior generation. The faster Nvidia innovates, the worse the collateral ages. That tension sits at the centre of the $500 billion structure and explains why Nvidia's willingness to backstop 25% was necessary to bring the capital pools together.

The broader concern extends past collateral. Combined hyperscaler capital expenditure estimates reached approximately $730 billion by July 2026, and investors have increasingly focused on free-cash-flow pressure and the possibility of slower spending growth. High spending supports near-term Nvidia demand while raising longer-term questions about debt financing, depreciation schedules, AI monetisation and whether annual spending can keep growing at the current rate. Anthropic's $11.5 billion quarter with positive adjusted operating income addresses the monetisation question directly, which is why that disclosure moved the entire supply chain on Monday. It does not address depreciation.

China: A $4.5 Billion Charge, H200 Licensing, and a 25% Tariff

Export restrictions have already cost Nvidia real money and continue to cap the addressable market. The company incurred a $4.5 billion charge in the first quarter of fiscal 2026 associated with H20 excess inventory and purchase obligations as demand for that product diminished under licensing requirements. In August 2025 the US government granted licences permitting shipment of certain H20 products to certain China-based customers, and Nvidia generated approximately $60 million in H20 revenue under them.

The subsequent programme has produced nothing. In February 2026 the US government granted a licence allowing shipment of small amounts of H200 products to specific China-based customers. Nvidia has generated no revenue under the H200 licensing programme and does not yet know whether any imports will be permitted into China. The licence requires H200s to pass a US inspection process before shipment, and any unit shipped under the programme is subject to a 25% tariff upon importation into the United States.

The economics of that structure explain the zero revenue. A product that must clear US inspection, absorb a 25% import tariff, and then obtain Chinese permission to enter carries a cost stack that undermines its competitiveness against domestic alternatives. Nvidia operates as a data centre scale AI infrastructure company across the United States, Taiwan, China, Hong Kong, Europe and internationally, and the China component of that footprint currently contributes close to nothing at the accelerator level.

Part of the valuation discount reflects this overhang directly, and part reflects execution risk around the Blackwell-to-Rubin transition. The strategic dimension is larger than the revenue line. Huang's bet on financing AI infrastructure hinges on outpacing AI development in China, which means the export restriction that removes Chinese revenue also creates the competitive window the entire $500 billion structure assumes. If Chinese domestic accelerators close the capability gap while Nvidia is locked out, the collateral value of Western AI factories and the growth trajectory underpinning a $5.3 trillion valuation both come under pressure simultaneously.

Competitive Encirclement From AMD, Intel, TPUs and Inference Startups

The competitive set has broadened even as the delivered volume gap has widened. AMD has positioned itself as the primary alternative with its MI350 GPU series, and Intel competes with its Gaudi platform. Specialised inference chip startups including Cerebras, SambaNova and d-Matrix target the workload segment where Blackwell's architectural advantages are least decisive.

Custom silicon from Nvidia's own customers is the more consequential threat. Alphabet's TPU activity demonstrates that hyperscalers will develop internal alternatives for specific workloads, and every dollar of TPU deployment is a dollar not spent on Nvidia accelerators. The counterargument is that Nvidia's addressable opportunity can expand even while customers build alternatives for some workloads, because total compute demand is growing faster than any single supplier can serve.

Monday's tape complicated the competitive read. AMD gained 6.50% to $514.39 while Nvidia held flat, which on its face suggests share-shift positioning. The more probable explanation is that AMD, Micron, SanDisk and Seagate all lack an earnings date in the next nine sessions, making them the preferred vehicles for expressing the Anthropic-driven demand thesis. Broadcom's 5.94% decline to $392.99 argues against a simple rotation-into-competitors reading, since Broadcom's custom accelerator franchise is the most direct challenger to Nvidia's position.

Scale remains the durable defence. Nvidia's $75.2 billion data centre quarter against AMD's $6.72 billion and Intel's $6.3 billion means competitors would need to multiply their businesses by an order of magnitude to threaten Nvidia's position, and the supply chain capacity to do so does not currently exist. What competitors can do is cap pricing power at the margin, which shows up in gross margins rather than revenue. Guidance of 75.0% non-GAAP gross margin is therefore the single most informative number in the August 26 release, because it measures competitive pressure more precisely than any revenue figure.

The Forecast: $236.54 Decides the Trend, $91 Billion Decides the Quarter

The bullish path requires three confirmations in sequence. First, Nvidia holds $225.16 through the run-up while the memory complex rallies, which would establish that Monday's flat print reflected event-risk management rather than distribution. Second, Q2 revenue above the $91.0 billion guidance midpoint and above the $91.90 billion consensus, with non-GAAP gross margin at or above the 75.0% target. Third, Q3 guidance at or above $100 billion alongside evidence of Rubin adoption among AI labs, neoclouds and sovereign customers. That combination clears $236.54 and takes the stock to a record, with the $285 to $309.94 target band as the next objective.

The bearish path requires two. Revenue merely meeting the $91.0 billion guide without a Q3 raise, or any commentary indicating supply friction or delivery delays across the Blackwell-to-Rubin transition. Given four consecutive post-earnings declines on four consecutive beats, an in-line print is the higher-probability negative outcome than an actual miss. The 6.99% implied move places the downside boundary near $209, and beneath that the $193 level from two weeks ago becomes the reference. The $180 low analyst target marks the floor of published expectations.

The base case into August 26 is continued range trade between $220 and $236.54 with volume concentrated and price static, precisely the pattern Monday delivered. A 25.6 times forward multiple on 92% FY2027 earnings growth is not an expensive stock, and the discount to the sector reflects three identifiable overhangs rather than a deteriorating business: China producing no accelerator revenue under the H200 programme, execution risk across the platform transition, and $730 billion of hyperscaler capital expenditure whose growth rate is now questioned.

The asymmetry favours Nvidia on fundamentals and disfavours it on positioning. Data centre revenue at $75.2 billion growing 92% annually against competitors at $6.72 billion and $6.3 billion is a structural moat that no announcement this quarter can alter. Against that, 43 of 47 analysts at strong buy, a $309.94 average target implying 37.6% upside, and options pricing a 6.99% move describe a market where the good news is already in the price. The $500 billion financing platform with its $125 billion backstop is the variable that could reprice the story in either direction, and management's answers on that structure will matter more than the revenue line.

That's TradingNEWS