Nvidia price prediction does not move just because the board approved the largest share-repurchase increase on record. At the 29 September close of $227.21, the $235 billion still authorized for buybacks would retire about 1.03 billion shares. That is 4.3 percent of the 24.285 billion diluted shares Nvidia reported for the quarter ended 26 July. Retiring 4.3 percent of the stock lifts earnings per share by about 4.5 percent if profits stay flat. It does not reprice a multiple. Reuters, citing LSEG, reported on 28 September that the shares traded at about 16.5 times forward earnings, the lowest since January 2015, against a 15-year average of 30. The cash return is enormous beside most of the S&P 500 and small beside Nvidia’s own equity. The board can retire shares. It cannot set the multiple.
The tape did not pay up for the announcement. Yahoo Finance closes were $230.36 on 4 September and $227.21 on 29 September, down about 1.4 percent. The later print sits 3.9 percent under the 52-week high of $236.54, and the stock is up 20.3 percent from the 2 January close of $188.85. Fiscal 2027 consensus earnings are $9.31 a share, against $4.77 in fiscal 2026, on S&P Global figures carried by StockAnalysis as of 29 September. The price rose about a fifth while the earnings denominator nearly doubled. That is why a stock near its high can still screen cheap, and why a roughly 4 percent share-count cut does not close the gap.
Key facts
$227.21, the Nasdaq regular-session close on 29 September 2026. The 52-week range is $164.27 to $236.54. The 30 September cash session had not opened when this series was pulled. — Yahoo Finance, 30 September 2026
$235 billion remains under the repurchase program after a $150 billion increase, to be used through fiscal 2028. Nvidia called the increase the largest on record. — NVIDIA newsroom, 28 September 2026
About 16.5 times 12-month forward earnings, the lowest multiple since January 2015, versus a 15-year average of 30. The $150 billion increase exceeds the market value of about 84 percent of S&P 500 constituents. — Reuters, 28 September 2026
Street average target $327.70, median $315, low $180, high $515, from 61 analysts. Fiscal 2028 consensus earnings are $15.68, on revenue of $682.73 billion. — StockAnalysis, citing S&P Global, 29 September 2026
July-quarter revenue was $96.2 billion, up 106 percent, with non-GAAP earnings of $2.22. The third-quarter guide is $108.0 billion, plus or minus 2 percent, at a 74 percent gross margin, and assumes no China data-center compute revenue. — NVIDIA, 26 August 2026
Cash and equivalents were $22.44 billion at the end of July. Free cash flow was $21.3 billion in that quarter, down from $48.6 billion in the April quarter. — Reuters, 28 September 2026; NVIDIA, 26 August 2026
Chinese suppliers are set to take nearly 80 percent of China’s AI server market this year. The foreign share is expected to fall to 21 percent from 34 percent. — Korea Times, citing TrendForce, 27 September 2026
Quick Take: At $227.21 the remaining buyback retires about 4.3 percent of the last reported diluted share count. That is a capital-return choice, not a new earnings forecast.
What is happening to the price
The three weeks since 7 September did not open a new range. Nvidia closed at $210.96 on 14 September, recovered to $228.86 on 28 September, the session of the buyback announcement, and finished at $227.21 the next day. That is a round trip toward the 4 September close of $230.36, on Yahoo Finance daily closes.
The action inside that flat patch is large in dollars and ordinary in share-count math. On 28 September the board added $150 billion, taking the remaining authorization to $235 billion through fiscal 2028. Reuters set the increase beside Apple’s $110 billion authorization from 2024 and noted that $150 billion exceeds the market value of about 84 percent of S&P 500 members. Reuters posted the same comparison. $235 billion divided by $227.21 is 1.03 billion shares, or 4.3 percent of the 24.285 billion diluted shares in the July quarter.
The cash balance does not fund it. Reuters reported $22.44 billion of cash and equivalents at the end of July, about 9.5 percent of the authorization. The earnings release shows free cash flow of $21.3 billion in the July quarter, after $48.6 billion in the April quarter. In the July quarter Nvidia also took in $24.9 billion of debt proceeds and spent $19.7 billion on stock. A $0.25 dividend, payable on 1 October to holders of record on 10 September, is about 0.11 percent of $227.21. It is not the return story.
The operating facts under the share price are still the 26 August print, which FinanceFeeds covered in the Q2 results note. Revenue was $96.2 billion, up 106 percent. Data-center revenue was $89.0 billion, up 117 percent. Gross margin was 75.0 percent. The third-quarter guide is $108.0 billion, plus or minus 2 percent, at a 74.0 percent gross margin, plus or minus 50 basis points, with no China data-center compute revenue assumed. Vera Rubin was already described that day as in full production. The buyback decides what to do with cash. It does not replace the guide.
Our cash generation gives us the capacity to invest in the technologies that advance this transformation and return capital to shareholders.
— Jensen Huang, founder and CEO of NVIDIA, NVIDIA newsroom, 28 September 2026
Quick Take: The shares are about 4 percent under the 52-week high and up about 20 percent this year. The multiple is low because earnings estimates rose faster than the price.
Who is responding
Nvidia’s response is the authorization and the reason Huang gave for it: cash from a platform shift, and confidence in the duration of demand. He did not publish a new revenue figure on 28 September. The figure Reuters attributed to the company, from the prior month, is about 70 percent revenue growth in fiscal 2028. The street is close and a little lower. S&P Global, via StockAnalysis on 29 September, has fiscal 2028 revenue at $682.73 billion, up 65.9 percent from $411.56 billion, and fiscal 2028 earnings at $15.68, up 68.5 percent from $9.31.
Customers have not walked back the orders under those figures. In late August, Amazon Web Services agreed to deploy an additional 2 million Nvidia GPUs in 2027 and 2028, on top of more than 1 million already committed, including Rubin-generation parts and a 100,000-GPU government slice. Amazon’s own budget is the $220 billion 2026 capital-expenditure guide FinanceFeeds reported when the capex figure was raised. A cloud budget is not an Nvidia revenue line. It is the pool Nvidia sells into.
Since 7 September the major desks have held targets rather than rebuilt them around the buyback. On StockAnalysis, using TipRanks for the names and S&P Global for the average, Atif Malik at Citi maintained a Buy and $315 on 29 September. Stacy Rasgon at Bernstein reiterated a Buy and $400 on 21 September. Timm Schulze-Melander at Rothschild & Co Redburn maintained a Buy and $325 the same day. James Schneider at Goldman Sachs maintained a Buy and $300 on 11 September. The average of 61 targets is $327.70. Those are bank targets, not the scenarios in the next section.
The dissent arrived with the headline. Jacob Bourne, an analyst at Emarketer, told Reuters the buildout will not last forever at this pace, and that the buyback signals staying power in demand. A company can be tight on supply this year and still be priced for a slower one. Hyperscalers are also paying for custom accelerators, the competitive question in the Broadcom scenarios. The multiple is where that risk is already being charged.
The AI buildout won’t continue at its current pace forever, but Nvidia is signaling confidence that demand for its hardware and services has staying power.
— Jacob Bourne, analyst at Emarketer, Reuters, 28 September 2026
Quick Take: Citi, Bernstein, Redburn, and Goldman Sachs kept Buy ratings and targets from $300 to $400. The $327.70 street average is a consensus of bank targets, not a FinanceFeeds scenario.
Market data and the scenarios
The Nvidia price prediction in this table is a set of scenarios. Each FinanceFeeds row is an earnings figure times a multiple, and the product is shown. Bank targets sit in separate rows. They are not forced into the same formula. A $400 Bernstein target is not a model output, and a model output is not something Citi published.
NVDA daily closes from the Yahoo Finance chart API, 30 September 2025 through 29 September 2026. The $169, $259 and $383 lines are FinanceFeeds scenario estimates. The $327.70 line is the street average target, not a FinanceFeeds scenario.
The chart is 251 Yahoo Finance daily closes from 30 September 2025 through 29 September 2026, ending at $227.21.
Case
What it is
Earnings used
Multiple
Math
Price
Bear
FinanceFeeds scenario
$12.10, fiscal 2028
14 times
$9.31 times 1.30, then times 14
$169
Base
FinanceFeeds scenario
$15.68, fiscal 2028 consensus
16.5 times
$15.68 times 16.5 = $258.72
$259
Bull
FinanceFeeds scenario
$15.68, fiscal 2028 consensus
24.4 times
$15.68 times 24.4 = $382.59
$383
Street average
Bank consensus, not a FinanceFeeds scenario
Not reconstructed
Not reconstructed
Mean of 61 targets, S&P Global via StockAnalysis, 29 September
$327.70
Citi
Bank target
Not reconstructed
Not reconstructed
Atif Malik, Buy maintained, 29 September
$315
Bernstein
Bank target
Not reconstructed
Not reconstructed
Stacy Rasgon, Buy reiterated, 21 September
$400
Goldman Sachs
Bank target
Not reconstructed
Not reconstructed
James Schneider, Buy maintained, 11 September
$300
The base case is the street’s $15.68 at the 16.5 times LSEG showed on 28 September. The product is $258.72, rounded to $259. LSEG did not publish $259. At $227.21 that multiple implies about $13.77 of earnings. Thirty times $15.68 would be about $470, and that old average is not a scenario here. The bull case puts 24.4 times, $227.21 divided by the $9.31 fiscal 2027 consensus, on the same $15.68. That is $382.59, rounded to $383, below Bernstein’s $400 bank target. The bear case grows $9.31 by 30 percent instead of 68.5 percent, to $12.10, then applies 14 times. The product is $169.44, rounded to $169. Thirty percent is not Nvidia’s forecast. It is the path if revenue does not step from $411.56 billion to $682.73 billion. The street low is $180. The $169 row sits under that low. It is not a desk’s number.
The buyback is not in those earnings. A 4.5 percent lift is about $0.70 on $15.68, or about $12 at 16.5 times, which moves a base case toward $271 rather than $383. The gross-margin guide is already a point under the 75 percent just printed, which is why the Micron comparison still belongs next to the multiple.
Quick Take: $259 is $15.68 times 16.5. $383 is $15.68 times 24.4. $169 is $12.10 times 14. The $327.70 average is an average of analyst targets, not one of those products.
Export controls, and what the summit did not do
The guide Nvidia is managing to still assumes no data-center compute revenue from China. That sentence is in the 26 August outlook, and no later release has replaced it. Jensen Huang sat at the head table of the 24 September White House dinner for Xi Jinping, with AMD’s Lisa Su, the Korea Times reported. A seat at the dinner is not a license, and it did not put a China number into the model.
The opening that already exists is the H200. On 13 January the Bureau of Industry and Security moved that chip, and similar parts, from a presumption of denial to case-by-case review if U.S. supply is protected, the buyer screens customers, and the chip is tested in the United States. Jeffrey Kessler, Under Secretary for Industry and Security, said controlled H200 sales would strengthen the American technology ecosystem. Blackwell and Rubin are not in that rule. Leaving China compute revenue out of the $108 billion guide is Nvidia treating the licensed slice as something other than a forecast.
The summit did not trade the controls away. The Korea Times reported that Trade Representative Jamieson Greer told CNBC the U.S. side had taken “off the table” the export controls “that are national security issues.” Beijing’s eight deliverables skipped export controls and included an AI-risk dialogue. A few Chinese firms may buy capped volumes of H200. The Korea Times said the delay produced a $400 million charge for excess H200 inventory in the first half. That figure is the newspaper’s, not a line re-read from the 10-Q.
TrendForce, via the Korea Times, expects domestic suppliers led by Huawei and Cambricon to take nearly 80 percent of China’s AI server market this year, and the foreign share to fall to 21 percent from 34 percent. Kyle Chan, a fellow at Brookings, called confidence in that supply the main change. The $259 case does not need a reopening. Buyers can still slow fiscal 2028 spending, the risk in the capex-slowdown note.
Neither side will acknowledge progress on any of these issues, given domestic sensitivities around anything that could be perceived as a concession.
— Paul Triolo, partner and technology policy lead at DGA-Albright Stonebridge Group, Korea Times, 27 September 2026
Quick Take: The summit took national-security export controls off the trade table. It did not put China data-center revenue into the $108 billion guide. The H200 remains a capped, licensed exception.
What happens next
Three paths follow from the guide, the multiple, and the authorization.
If revenue lands inside $105.8 billion to $110.2 billion and gross margin lands inside 73.5 to 74.5 percent, the street’s $15.68 does not have to be rebuilt. Holding 16.5 times on that figure is the base case of $259, about 14 percent above the 29 September close. The cause is delivery of a guide that is already public. A faster buyback adds only about $12 at that multiple, nearer $271 than $383.
The bull case is a decision by holders, not by the board. Fiscal 2028 earnings are already $15.68. Capitalizing them at 24.4 times, the multiple on this year’s $9.31, is $383. That requires the multiple to stop falling as the earnings arrive. The drop from a 15-year average of 30 times to 16.5 times is that markdown. A record authorization can sit beside it. It does not reverse it. The check is whether the 65.9 percent revenue step is still the consensus after the third-quarter call, and whether the shares still trade at a mid-teens multiple on that year.
The bear case is a growth cut, not a new ban. The practical ban is already in the guide. If the next outlook still assumes zero China data-center compute revenue, and if gross margin is guided below 74 percent, the earnings line is what moves. A step from 68.5 percent growth to 30 percent is $12.10. At 14 times, that is $169. The $0.25 dividend on 1 October changes none of the three. The next Nvidia price prediction checkpoint is the quarter, not the coupon.
Quick Take: $259 follows if the guide is met and 16.5 times holds. $383 follows only if the multiple rises toward 24 times next year’s earnings. $169 follows if fiscal 2028 growth is cut to 30 percent and the multiple slips to 14 times.
FAQ
What is the Nvidia price prediction from the 29 September close?
The FinanceFeeds scenarios are $169, $259, and $383, against a $227.21 Nasdaq close on 29 September 2026. They are not the street average. S&P Global’s mean of 61 targets was $327.70 that day, and the median was $315. The base case is $15.68 times 16.5. The bull case is $15.68 times 24.4. The bear case is $12.10 times 14. Each price is the product in the table, rounded to the nearest dollar.
Does the $235 billion buyback produce the bull case?
No. At $227.21 the authorization retires about 1.03 billion shares, 4.3 percent of the July quarter’s diluted count. Unchanged profits would lift earnings per share by about 4.5 percent, roughly $12 on the base multiple, not the gap to $383. The bull case is a higher multiple. July free cash flow was $21.3 billion and cash was $22.44 billion, so the authorization spends future cash flow.
Did the Xi-Trump summit add China revenue to the outlook?
No. The 26 August outlook still assumes no data-center compute revenue from China, and Nvidia has not replaced that sentence. The Korea Times reported that national-security export controls were taken off the trade-talk table, and that Beijing’s summit deliverables did not mention export controls. The H200 can be licensed case by case under the January rule. That channel is capped. It is not the data-center forecast inside the $108 billion guide.
How do these scenarios differ from bank targets?
A bank target is a figure a named analyst publishes. Citi’s is $315, Bernstein’s is $400, and Goldman Sachs’ is $300, on the dates in the table. The $327.70 average is the mean of 61 such targets. A FinanceFeeds scenario is a stated earnings figure times a stated multiple, so the multiplication can be repeated. The rows are separate because they answer different questions: what a desk has published, and what a chosen multiple does to a chosen earnings year.
What is the dividend worth at this price?
Nvidia pays $0.25 a share on 1 October 2026 to holders of record on 10 September 2026. At $227.21 that coupon is about 0.11 percent of the price. The rate moved from $0.01 in May. Four payments would be $1.00 a share, under half of 1 percent a year. The capital-return argument is the repurchase authorization, not this coupon.
Which price is the spot in this note?
The spot is $227.21, the Nasdaq close on 29 September 2026, from Yahoo Finance. The 30 September cash session had not opened when the series was pulled, so there is no official close for that day. A premarket print near $228 is not used. The same feed shows a 52-week high of $236.54 and a low of $164.27. A later close replaces the spot. It does not reset the earnings or the multiples.
