NVIDIA Earnings, August 2026: $96.22 Billion in Revenue, Weak Cash Collection and a Year-End Stock Outlook

A reading of NVIDIA’s second fiscal quarter (ended 26 July 2026): revenue versus consensus, gross margin, receivables, the FY2028 guide and share-price scenarios through 31 December 2026.

Published 27 August 2026 · Written before the regular session priced the print in full

Why this NVIDIA earnings report matters now

Johann Wolfgang von Goethe spent nearly sixty years on Faust. Part I appeared in 1808, while Europe was still living through the Napoleonic wars. Part II, begun much earlier, was finished in his last years and published in 1832, after his death. Between the two parts the subject itself shifts. It moves from one man’s soul to the soul of a state that has learned to live on credit.

The plot is famous and contract-simple. Heinrich Faust is a scholar who has exhausted his books and found no meaning. Mephistopheles offers him service on earth in exchange for his soul, if a single instant of delight makes him say: “Remain, you are so beautiful.” In Part I the wager plays out through Gretchen - love, guilt, ruin. In Part II the stage widens to the imperial court. The treasury is empty. The courtiers want a feast. Mephistopheles offers a solution that sounds like statecraft: issue paper money against the gold that belongs by right to the crown, but still lies in the ground. The banknote is covered. The cover has not been mined. The court feasts that same evening. The mine is left for tomorrow.

The metaphor is not that the paper is a lie. The metaphor is that recognition runs ahead of extraction. The revenue is already in circulation. The metal is not yet in the till.

The text below was written on the morning of 27 August 2026, before regular trading delivered the first full market verdict on the report. The official close on 26 August was $209.66. The $219.50 print after the session was a first reaction, not a settlement. The gap is not small. At a market capitalisation of about five trillion dollars, ten dollars a share is nearly $240 billion - the scale of a large energy company, gathered in the hours between the press release and the morning.

On 26 August NVIDIA beat analysts’ consensus on revenue and earnings per share for the fifteenth consecutive quarter. Immediately after the numbers the stock faded. It turned higher only when chief financial officer Colette Kress offered something the company had not given before: a guide for an entire next fiscal year. About 70 per cent revenue growth in fiscal 2028. Street consensus had been holding around 44 per cent. The real news was not in the July table. It was in the sentence about 2028.

Capital markets know the Goethe court manoeuvre. When management stops talking only about the next ninety days and starts talking about the next eighteen months, it is either sure of the seam under its feet, or it wants the conversation moved from the quality of what has already been dug to the size of what has not yet been opened. On this morning both readings of NVIDIA are possible.

The argument is no longer whether artificial intelligence buys chips. The argument is three questions. First: what share of reported operating profit turns into cash. Second: how concentrated the ultimate payer remains after the names on the invoices have been spread out. Third: what price-to-earnings multiple a business can support when it itself is calling for a lower gross margin and itself is extending payment terms.

The market surprise is no longer a $108 billion guide for the next quarter. After the analyst call that is the starting point. A surprise would be a result above the top of the stated range - or the same revenue with weaker collection of receivables.

NVIDIA’s fiscal year versus the calendar: what the July 2026 report covers

NVIDIA’s fiscal year does not match the calendar year. It ends on the last Sunday in January. Fiscal 2027 runs from February 2026 to January 2027. The second quarter ended on 26 July 2026. The third quarter ends towards the close of October; results are expected around 18 November. 31 December 2026 falls inside the fourth fiscal quarter - before the annual report for fiscal 2027. “By year-end” and “fiscal 2028” are not the same horizon. Fiscal 2028 is February 2027 through January 2028.

Revenue, profit and margins versus consensus

Table 1. NVIDIA’s key second-quarter results versus consensus

Adjusted earnings is a company-defined measure that adds back or strips out selected accounting items to show what management treats as “run-rate” profitability. It is comparable with consensus, but after a methodology change it now includes stock-based compensation, so the year-on-year comparison is not fully clean. At this scale of sales this is not merely a stable company. It is an operating-profit rate markets rarely see outside government contracts and owners of an industrial standard.

What is still a forecast, not a fact

Table 2. NVIDIA management statements that are not yet facts in the filing

GAAP net income is $59.69 billion - only 2 per cent higher than the prior quarter. It includes $7.77 billion of net income from the revaluation of equity investments. That is not a sold and paid chip. That is a portfolio that has marked higher. For an investment decision, GAAP net income is the wrong compass.

On the call, analysts returned to two themes the filing itself had put forward: the price of memory and the way NVIDIA takes part in financing its customers’ infrastructure. Management’s answer was blunt on the first and institutional on the second. Memory is “extremely” expensive and above their prior estimates; gross margin will fall. External capital with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR was presented as a way for customers to build without that construction sitting as debt on NVIDIA’s balance sheet. Both themes remain November tests: the first in the gross-margin percentage; the second in receivables and in the notes on guarantees.

Jensen Huang framed the era this way: artificial intelligence has passed the inflection point; tokens are productive and profitable; compute is already revenue. As a message to investors who doubt the return on AI capital spending, the sentence is clear. As proof it is not enough. A stronger argument would be cash following operating profit with the same consistency.

NVIDIA revenue by product and segment: Data Center, cloud and Graphics

The company presents the same $96.22 billion under three classifications. The rows do not add. Anyone who sums them counts the same revenue three times.

Table 3. NVIDIA revenue by market platform

Table 4. NVIDIA revenue by accounting segment

Graphics and the edge overlap. They are not two separate businesses.

During the quarter one customer was reclassified from the industrial line to the large cloud operators, and comparable periods were restated. The 25 per cent growth versus 13 per cent does not, by itself, prove new end customers. Part of the “industrial” line may be the same laboratory or the same cloud buying through a smaller tenant of compute capacity - a neocloud. A neocloud is a company that buys or leases entire halls of accelerators and resells compute time to third parties, often without the balance sheet or credit rating of Amazon, Microsoft, Google or Meta. NVIDIA itself notes that one AI laboratory generated “significant” revenue by buying cloud services through NVIDIA customers.

When the end user buys through an intermediary, the direct shares on the invoices look more dispersed than the actual concentration of demand.

Geography of sales: United States, Taiwan and China

Table 5. NVIDIA revenue by invoice address

Taiwan is mainly a channel to integrators and makers of full systems. Compute revenue for data centres in China is not in the guide. Hopper shipments to China are under 1 per cent of Data Center revenue. In the first half a $0.4 billion provision was taken on obsolete H200 product. Over a four-month horizon China barely moves the revenue line. Over a longer horizon NVIDIA’s absence from that market gives local competitors domestic room in which to build their own software layer and later offer it abroad. That is a strategic risk, not a near-term price risk.

The company does not publish unit shipments or average selling price per rack. The mix of sales is visible only indirectly - through the margin and through the statement that Blackwell Ultra still accounts for the bulk of shipments. Without unit figures, any claim of a “richer mix” remains a partial inference.

NVIDIA gross margin: from 75% toward 71-72%

A 75 per cent gross margin is high. The path management now describes is no longer rising.

Table 6. Stated trajectory of NVIDIA’s gross margin

A basis point is one-hundredth of a percentage point. The net effect of inventory provisions in the reported quarter was about minus 0.8 percentage points on gross margin. Management describes memory prices as extremely high and above prior estimates. That does not exhaust the margin risk. If cheaper configurations or lower-margin processors rise in the mix, a 72 per cent level will not be memory alone. It will also reflect a change in the structure of sales.

Three percentage points of lower gross margin on $100 billion of revenue cuts gross profit by $3 billion. With operating expenses almost unchanged, the 66 per cent operating margin moves toward the high fifties or low sixties. The economic model remains highly profitable. The idea that 75 per cent is a durable norm is not supported by the company’s own guide.

The historical parallel here is not a speculative bubble. It is a mix shift at the owner of a standard. When Intel set the architecture of the personal computer, its gross margin also looked durable - until customers began to design systems around it, not solely through its products. NVIDIA is not Intel in 2004. The margin path, however, is no longer one-way.

Operating expenses and the effective tax rate

Operating expenses were $8.41 billion against the company’s prior guide of about $8.5 billion. For the next quarter it is pointing to about $9.2 billion on a GAAP line and $9.0 billion on an adjusted line. For the whole of fiscal 2027, operating expenses are expected to grow in the low fifties of per cent. Tax on profit was $11.8 billion on $71.5 billion of pre-tax profit - an effective rate of about 16.5 per cent, inside the annual 16-18 per cent guide.

NVIDIA cash flow: why only 38% of operating profit reached the till

After the 2000-2001 collapse of the dot-com wave, the market adopted a blunt rule: revenue is an estimate, cash is the check. The rule is crude. At NVIDIA it needs a refinement, because GAAP net income includes revaluations. The right denominator is operating profit.

Table 7. How much of NVIDIA’s operating profit became cash

What has deteriorated is the share of profit that arrives as cash, not the existence of cash flow. The regime is not a shortage of money. The regime is a gap between the pace of recognised profit and the pace of collection.

Receivables, 60-day collection and inventories

Table 8. NVIDIA receivables, collection period and inventories

About $16 billion of receivables is explained by longer contractual terms. The rest of the $22.3 billion increase reflects a larger volume of sales. To ascribe the entire rise to worse payment terms is to mix volume with tenor.

For credit risk, the structure of receivables is an earlier signal than the 16 per cent revenue share of a single direct customer. Revenue shows who is buying. The receivable shows who has not yet paid. The jump in raw materials is preparation for Vera Rubin and is not, by itself, a warning. The warning is the lengthening of collection.

In the first quarter the company made no federal advance payments of corporate tax. In the second it made two. Part of the drop in operating cash flow is the tax calendar. The calendar does not explain about $16 billion of receivables tied to longer terms.

Buybacks, the dividend and new debt

Table 9. NVIDIA free cash flow, capital returned and new debt

Without the new issue and without drawing on the securities portfolio, returning $26 billion against $21 billion of free cash flow would have reduced the cash pile. NVIDIA returned more to shareholders than it generated as free cash in the quarter, and covered the gap with debt and portfolio assets.

For liquidity the sequence matters. For earnings per share the effect is negligible. At a five-trillion-dollar capitalisation a buyback is a signal to shareholders and a test of the cash balance. It is not a mechanism that materially lifts earnings per share.

In the late 1990s Lucent and Cisco extended trade credit to customers in order to support revenue. Later the revenue stayed in the accounts longer than the cash stayed in the till. The comparison is not a verdict on this filing. It is a test for the next ones. One quarter at 60 days outstanding does not repeat 2001. Two consecutive quarters of a lengthening collection period and of cash at about 38 per cent of operating profit will no longer be explained by the tax calendar.

Customer concentration and financing demand: neoclouds, guarantees and $500 billion

One direct customer accounted for 16 per cent of revenue in the quarter. For the half year three direct customers held 16, 15 and 13 per cent. Separately there are indirect end customers above 10 per cent and a laboratory that buys cloud services through NVIDIA customers. The names in the filing are spreading. The ultimate payer remains concentrated.

What share of the $89 billion in Data Center runs through a neocloud with an NVIDIA guarantee, through a contract with extended terms, or through an entity in which NVIDIA holds an equity stake, the filing does not say. The weight of that linkage cannot be calculated from public data. Its existence is visible.

Table 10. Visible ties from NVIDIA to the customer’s balance sheet and future commitments

The $500 billion intention with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR matters for the story told to investors. No capital has been received. Supply commitments are a booking with vendors. They are also a hard future payment if demand fades.

The mechanism is one. Sales are spread across more categories. The payer at the end of the chain remains a narrow circle. Part of that circle uses trade credit, a guarantee or an equity stake from NVIDIA. The broader classification in the filing and the tighter link between revenue and the customer’s balance sheet are not two separate trends.

Part of accounting profit comes from marking up stakes in the same ecosystem that buys the products. A rise in NVIDIA’s share price supports the valuations of related companies; higher valuations return as financial income in the filing. That is why GAAP earnings per share is not used as the primary compass.

Capex at Amazon, Microsoft, Alphabet and Meta: what it proves and what it does not

The guide of about 70 per cent growth in fiscal 2028 assumes durable capital spending at customers. The latest public outlooks from the four large cloud groups, as of late July and early August, do not show cuts.

Table 11. Capital expenditure of the large cloud groups in calendar 2026

The figures confirm that infrastructure construction has not stopped. They do not confirm what share of those budgets goes to NVIDIA products. Higher capex does not mean proportionally larger purchases from NVIDIA.

NVIDIA’s filing contains no data on the economic life of delivered systems at the customer. Vera Rubin is offered at a higher revenue per installed gigawatt precisely because the prior generation is being replaced faster.

Table 12. Indicative NVIDIA revenue per installed gigawatt, by generation

If the customer depreciates the machine over four years and the generation turns over in eighteen months, the economic return is lower than the accounting return. The gap will not show up in NVIDIA’s current revenue. It will appear in customers’ capital budgets with a lag of one or two quarters.

The price of credit works through the neocloud operators. A higher rate makes the guarantees more expensive for the party that wrote them. A threshold - at what yield the tenant stops adding capacity - is not in the public documents. The direction of the dependence is clear.

Vera Rubin, Blackwell Ultra and the competitors: AMD, TPU, Trainium, Maia, CUDA

Blackwell Ultra formed the revenue of the reported quarter. Vera Rubin has been in volume production since early August. The November test is double. First: whether its share is about one-fifth of Data Center revenue. Second: whether the margin on that share does not pull the corporate gross margin below the 74 per cent guide. Hitting only the first mark is not enough. An 18 per cent share at a thinner margin is formally close to the promise and economically weaker.

The Vera CPU, Spectrum-6 switches, the Groq 3 LPX accelerator and the DSX platform sketch the direction of the full stack - GPU, CPU, network, software and a ready hall. The guide of about $20 billion of demand for server CPUs is a management estimate, not a book of firm orders.

The constraint on growth is not single. Memory price and availability have been stated. Vendor commitments have been raised to $279 billion. Separately the company takes commitments on power, land and buildings. If the next bottleneck is permitting and grid capacity, the 70 per cent guide for fiscal 2028 will not fail because of a competitor. It will fail because of physical and regulatory limits. There is no breakdown in this filing of capacity by wafer, memory and packaging.

In the merchant channel for accelerators outside China, NVIDIA keeps the lead. Custom accelerators at the cloud groups - Google’s TPU, Amazon’s Trainium, Microsoft’s Maia, Meta’s MTIA - are growing faster than general-purpose GPUs. That is the structural competitor. AMD is a second supplier with large announced gigawatts toward laboratories and platforms. Until those gigawatts appear as revenue in someone else’s filing, the installed-base comparison still favours NVIDIA by construction. China, for this horizon, is zero in the Data Center compute guide.

The CUDA software environment, on observed practice, still holds the training of leading models. That is an inference from customers’ behaviour so far. It is not a measured cost of leaving for another environment. Industrial standards last a long time - Windows outlived more than one generation of hardware. The same standards give way when the customer is large enough to build its own architecture. The large cloud groups are already doing so.

Valuing the NVIDIA share: price-to-earnings, price-to-sales and free cash flow

At $209.66 the market capitalisation is about $5.08 trillion.

Table 13. NVIDIA valuation multiples at $209.66

The second forward P/E is conditional. It has meaning only if the 70 per cent growth materialises. The multiple on free cash flow is the most stretched valuation.

The free-cash-flow yield at a $5.08 trillion capitalisation and $21 billion for one quarter is negligible on an annualised basis if cash stays near 38 per cent of operating profit. To bring the cash multiple back into the high twenties or low thirties, free cash flow over twelve months has to approach, roughly, $150-200 billion. That is the arithmetic of the scenario in which revenue hits the guide and the market still compresses the multiple: the top line is delivered, cash yield does not recover, the P/E contracts.

Against Broadcom, NVIDIA’s trailing earnings are cheaper. Against AMD they are cheaper as well, at a substantially higher margin. A lower multiple versus peers is not a low multiple versus its own shrunken cash flow and versus the stated decline in gross margin.

If the next quarter is about $108 billion and the fourth records another step up, fiscal 2027 revenue falls in a wide $400-430 billion range. That is arithmetic on public guides, not an annual company forecast. Seventy per cent on that base gives fiscal 2028 in a $680-730 billion range. Consensus before the call was materially lower. The gap is the argument over the conditional forward P/E. It is not a basis for a discounted-cash-flow valuation built on one sentence from a conference call.

At this scale of sales historical comparisons are scarce. Companies with tens of billions of revenue rarely double the base more than once. NVIDIA is doing it on a foundation that already exceeds the annual sales of most large technology companies of the prior generation. The market has no settled way to price that pace, which is why the P/E moves sharply on every new fact about the margin and about receivables.

The average twelve-month analyst targets - about $310-317, in a wide band from about $180 to $515 - are not a 31 December 2026 compass. They point toward the summer of 2027, and only if the oral FY2028 guide is confirmed after two further reports.

NVIDIA share-price outlook to 31 December 2026: four scenarios

The order below is not a statistical probability. It ranks outcomes by how much of the information is already public. The price ranges are a compass around the official $209.66 close and around the first after-hours reaction. They are not a forecast. Regular trading on 27 August can move the starting point.

Table 14. NVIDIA share-price scenarios through 31 December 2026

A level near $310 matches twelve-month targets after two further reports. It is not a year-end compass. The first technical cap overhead is the yearly high of $236.54. Below $200 the market would be paying a lower P/E at the same level of demand.

A framework for a position in NVIDIA stock

A horizon that can absorb the swings in this stock without a forced sale is twelve to twenty-four months. On a shorter horizon the weight in a portfolio should be limited. The drawdown accepted in advance, which the position must be able to stand, is 20 to 25 per cent.

The sessions of 27 and 28 August are not the moment to add. Adding makes sense only if the November report shows, together, a collection period back toward 45-50 days and operating cash that is again a material share of operating profit. Vera Rubin’s slide share is not, by itself, a reason to add.

Without a current market price of volatility, option structures are not justified here. A naked short after an oral 70 per cent guide for the next fiscal year has an unfavourable payoff. A comparison with other links in the same investment cycle - power, memory, networking gear - makes sense only after November, and only if cash remains structurally weak. Specific issuers and hedge ratios are not fixed before that report.

Position size is set by the willingness to absorb a 20-25 per cent decline, not by the force of the stated 70 per cent. That guide has already been spoken. The first after-hours reaction included it in the price only in part and only for a time.

What to watch in NVIDIA’s November 2026 report

Table 15. Five checks for NVIDIA’s November report

If the first two checks are met and the fourth and fifth are not, the probable range is $200-220. If the third, fourth and fifth arrive together with the revenue, the $230-255 range becomes realistic. If the second check misses on share or on margin, the oral FY2028 guide weakens first and the P/E compresses before the revenue line does.

The November report will not close every open question. It will not give the share of revenue running through guarantees and extended terms. It will not measure the cost of leaving CUDA. It will not break manufacturing capacity into wafer, memory and packaging. It will not state a rate threshold at which a neocloud operator stops expanding. The FY2028 guide will keep resting on those unknowns as well. The reasonable behaviour is for the unknowns to limit the weight of the position, not to remain a footnote.

The ledger is posted. The cover is still being mined

Demand is confirmed by the filing and by the fact that the large cloud groups are not cutting their capital programmes. NVIDIA’s share inside those programmes is not confirmed. The quality of each unit of growth is weaker on three lines at once: a stated decline in gross margin, extended receivables concentrated in a handful of names, and the manufacturer’s participation in financing and in the customer’s power infrastructure.

Until November the investment question is whether demand holds - at a price that, after the 26 August session, already reflects part of the optimism on revenue but has not yet been confirmed by regular trading. Adding to a position is justified by cash. Not by the new platform’s share on a slide.

In Part II of Faust the court is not deceived in the narrow sense: the gold in the ground may well exist, and the crown may well have a right to it. The deceit, if there is one, is in time. The banknote is issued before the face is opened. The feast is tonight. The extraction is tomorrow.

The 70 per cent guide for fiscal 2028 is such a banknote. Not because the growth is impossible - the July report showed a plant that is still digging, and customers that are still building. But because part of what has already been recognised lives on tenor, guarantee and a seam not yet opened: Vera Rubin as a share, collection at 38 per cent of operating profit, the five hundred billion as a preliminary agreement.

Goethe leaves Faust searching for an instant he wants to hold. The market after the session was tempted to hold the sentence about 2028. November is not that instant. November is the test of whether the cover has been mined - whether receivables shorten and cash enters the till - or whether the court has received only a new issue of paper against the same unopened mine.

The revenue is on the books. The cover is still being mined.

Frequently asked questions on NVIDIA’s report and the stock

1. Should NVIDIA stock be bought immediately after the report?

Not in the sessions of 27 and 28 August, if the aim is to add to an existing holding. The report confirmed demand and raised the bar for the next quarter to about $108 billion. Collection of profit into cash deteriorated: operating cash is about 38 per cent of operating profit, against about 94 per cent in the prior quarter. The after-hours price included part of the optimism on the fiscal 2028 guide, but regular trading has not yet delivered a verdict. Adding makes sense if, in November, the collection period returns toward 45-50 days and cash is again a material share of operating profit. A horizon shorter than twelve months requires a limited weight. A 20-25 per cent decline accepted in advance is the condition on which the position is held without a forced sale.

2. Why, on such a report, does the market not re-rate NVIDIA at once?

Because the revenue surprise was already partly in the price, and the new information weakens the quality of each unit of growth. A 75 per cent gross margin is high, but the company’s own guide takes it to 74 per cent next quarter and to 71-72 per cent in the winter. Collection lengthens from 45 to 60 days; about $16 billion of receivables is due to longer contractual terms. Five customers hold about 70 per cent of receivables. Free cash flow for the quarter is about $21 billion against about $26 billion returned to shareholders and a new debt issue of about $25 billion. A price-to-free-cash-flow multiple of about 40 times is stretched. The market can accept $108 billion of next-quarter revenue and still pay a lower P/E if cash does not follow operating profit.

3. How much weight does NVIDIA’s oral guide of about 70 per cent growth in fiscal 2028 carry?

It moves the conversation and the conditional forward P/E. It is not an annual forecast in the results table and it is not a basis for a discounted-cash-flow valuation built on one sentence. Fiscal 2028 is February 2027 through January 2028; by 31 December 2026 the market will see at most one further full report - the third quarter, around 18 November. Arithmetic on $108 billion next quarter and another step in the fourth gives fiscal 2027 in a wide $400-430 billion revenue range; 70 per cent on that base points to $680-730 billion in fiscal 2028. Consensus before the call was about 44 per cent. The gap is the argument. Support from the customer side is that Amazon, Alphabet, Microsoft and Meta are not cutting calendar-2026 capital programmes (together about $730 billion). That does not prove NVIDIA’s share inside the budgets: at Microsoft more than half is already described as power, cooling and buildings, Alphabet is developing its own accelerators, and Amazon’s increase is tied in part to memory.

4. What is the main risk for an NVIDIA shareholder through the end of calendar 2026?

Not a lack of demand, but the quality of growth already recognised. The lines are three and they act together. First, the stated decline in gross margin amid extremely expensive memory and a possible cheaper mix. Second, concentration and extended tenor of receivables - five names with about 70 per cent of $63.1 billion. Third, the manufacturer’s participation in financing the buyer: $36 billion of cloud-service agreements, $20 billion of leases for transfer, guarantees for 4.25 gigawatts to SB Energy in Ohio, $25 billion of committed equity investments, and preliminary agreements for more than $500 billion with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR, which according to the filing may not become final contracts. The filing does not give what share of $89 billion in Data Center runs through a guarantee, an extended term or an entity in which NVIDIA holds a stake. The presence is visible. The weight is not. The structural competitor over a longer horizon is custom silicon at the cloud groups (TPU, Trainium, Maia, MTIA), not so much AMD, whose announced gigawatts are not yet revenue in someone else’s filing. China is zero in the Data Center compute guide and does not move the price through year-end; strategically it gives room for a local software layer.

5. What exactly must the November report show for the thesis to change?

Five checks together, not one slide on Vera Rubin. Revenue versus the $108 billion range plus or minus 2 per cent - landing inside is the expectation, the miss is the news. Vera Rubin at about one-fifth of Data Center revenue at a margin that does not pull corporate gross margin below the 74 per cent guide. Gross margin itself versus 74 per cent and the tone on a 71-72 per cent winter trough. Collection toward 50 days or stretching again, plus the names in the receivables book. Operating cash versus operating profit materially above 38 per cent. If revenue and the Vera Rubin share are delivered and collection is not, the probable range remains $200-220. If margin, tenor and cash improve together with revenue, the $230-255 range becomes realistic. If Vera Rubin misses on share or on margin, the FY2028 guide weakens first and the P/E compresses before the revenue line does. November will not measure the share of tied revenue, the cost of leaving CUDA, capacity by wafer, memory and packaging, or the rate threshold for a neocloud operator. Those unknowns continue to limit the weight of the position.

Important notice

This article is for information and education only. It is not financial advice, an investment recommendation or professional advice of any kind. The views expressed are the author’s personal opinions and are not recommendations to buy, sell or hold securities, assets or investments. Investing involves substantial risk, including the possible loss of all capital. Past performance is not a guarantee of future results. Each reader should conduct independent research, consult a qualified financial adviser and weigh financial circumstances, risk tolerance and investment objectives before taking any decision. The author and borsite.net accept no liability for any loss arising from the use of the information in this article.