NVDA — NVIDIA Corporation
Price when we picked it (Oct 5, 2026)
$236.07
Target: $328.72 (+39.2%)
P/E Ratio
29.8
P/E Forward
14.9
Dividend
0.42%
Market Cap
$5700.4B
EPS
$7.91
Consensus
Strong Buy
What they do
NVIDIA Corporation (Nasdaq: NVDA; headquartered at 2788 San Tomas Expressway, Santa Clara, California; founded in 1993; co-founder Jensen Huang has been CEO since day one; CFO Colette Kress; ~42,000 employees in 38 countries at the end of its fiscal 2026, 31,000 of them in research and development) invented the graphics processor (GPU) in 1999 for video games and became the company that designs the infrastructure used to train and run artificial intelligence. It does not manufacture its chips: it designs them, and TSMC and Samsung produce them; it buys memory from SK hynix, Micron and Samsung.
It reports two big businesses. Data Center ($89.0 billion in May–July 2026, 92.5% of its sales): the Blackwell chips and the new Vera Rubin generation (already in full production), the Grace and Vera processors, the networking gear that links thousands of chips together (Spectrum-X) and the CUDA software, around since 2006, used to program its chips. Of that $89.0 billion, $49.0 billion was bought by the big clouds (such as Amazon, Microsoft, Google and Oracle) and $40.0 billion by newer AI-specialized clouds (such as CoreWeave and Nebius), governments, industry and businesses. Edge Computing — everything that doesn't live in a data center — ($7.2 billion, +27%): GeForce and RTX cards for gaming and PCs, workstations, and the computers inside self-driving cars and robots.
A note on the calendar: its fiscal year ends in late January, so 'the second quarter of fiscal 2027' is May–July 2026. And the price already reflects the 10-for-1 split of June 2024.
Why we like it
Three reasons.
- NOBODY THIS BIG GROWS THIS FAST: in May–July 2026 (reported on August 26) it sold $96,221 million, more than double a year earlier (+106%), and it was the fourth quarter in a row in which its growth accelerated. It beat what analysts expected (~$92.2 billion in sales and $2.22 per share vs. ~$2.10). In February–July alone it sold $177,837 million — 82% of everything it sold in its previous fiscal year. For August–October it expects $108 billion, and its CFO, Colette Kress, forecast ~70% growth in fiscal 2028, when analysts expected ~44%. She clarified it is a supply-constrained forecast: its customers' forecasts point to its sales doubling, and the five biggest clouds would invest nearly $800 billion in 2026 and $1.3 trillion in 2027.
- A CASH MACHINE THAT HANDS BACK MORE AND MORE: of every dollar it sells, it keeps 75 cents after manufacturing and 66 after paying for research and salaries. In February–July it generated $69,895 million of free cash, and in May–July alone it returned ~$26 billion to its owners (~$20 billion in buybacks and ~$6 billion in dividends). In May it multiplied its dividend by 25 (from $0.01 to $0.25 per quarter) and on September 28, 2026 it authorized another $150 billion for buying back its own shares — a record — bringing the total available to $235 billion, which it plans to use by the end of fiscal 2028 at the latest: ~4% of what the whole company is worth.
- NOT AS EXPENSIVE AS IT LOOKS: at $236.07 it costs ~30 times the profit of the last twelve months, but ~25 times what analysts expect for the fiscal year ending January 2027 ($9.31 per share) and ~15 times the one ending January 2028 ($15.80). After the August report, analysts raised their estimate for that year by ~23% (in early August 2026 they expected $12.84). Consensus Strong Buy: of 61 analysts, 58 say buy, 2 say hold and 1 says sell; the average target of 59 of them is $328.72 (+39%).
Key Risk
The #1: it depends on a handful of giants continuing to spend on AI — and NVIDIA finances part of that spending. 92.5% of its sales are data centers; the big clouds bought $49.0 billion of the quarter's $96,221 million, and a single direct customer accounted for 16% of sales (in February–July, three customers accounted for 16%, 15% and 13%). On top of that, NVIDIA already has $99 billion invested in companies in its own ecosystem — AI model makers, clouds and financiers — plus $25 billion committed, and in August 2026 it signed guarantees of up to $105 billion for a data center campus that SB Energy is building for OpenAI in Ohio. If AI spending slows, its sales and the value of those bets would fall at the same time. Others:
- memory got more expensive: its CFO speaks of 'extreme pricing conditions'; gross margin will go from 75% to 74% in August–October and to 71%–72% in November–January, before settling at 72%–73% in fiscal 2028 with the price increases it has already made;
- its customers make their own chips: Google announced its eighth-generation TPU chip and already sells it directly to certain clients, and Amazon's in-house chip business (Trainium) is growing at triple-digit rates;
- China: it is not counting on selling data center chips there; the US gave it licenses to sell H200 chips to certain Chinese customers, but the Chinese government is holding back those purchases, the shipments pay a 25% tariff that NVIDIA absorbs, and it took a $400 million charge for H200s that no longer have demand;
- it collects more slowly: its customers owe it $63,059 million (in January it was $38,466 million) and take ~60 days to pay because it gave longer terms on large purchases — that is why in May–July free cash ($21,341 million) was much lower than profit ($59,688 million);
- huge commitments: it committed to buy $279 billion in manufacturing and memory (it was $119 billion at the end of April 2026); if demand falls, those contracts weigh;
- we bought at its all-time high and it moves hard: it fell 20% between October 2025 and March 2026 and another 19% between May and July 2026; it moves ~2.2 times as much as the market;
- symbolic dividend: ~0.42% a year — the return depends on the company continuing to grow.
This is educational and informational content, not financial advice. Always consult a qualified financial advisor before making investment decisions.
Vectorial Data picked NVDA on Oct 5, 2026 at $236.07.
Full Research
What it does. NVIDIA Corporation (Nasdaq: NVDA; headquartered at 2788 San Tomas Expressway, Santa Clara, California; incorporated in California in April 1993; co-founder Jensen Huang has been President and CEO since its founding; CFO Colette Kress; ~42,000 employees in 38 countries at the end of fiscal 2026, 31,000 of them in research and development) describes itself as a data-center-scale AI infrastructure company. It invented the GPU in 1999, launched CUDA in 2006 — the software platform that runs on all its GPUs and is used to program them — and in 2012 the AlexNet neural network, trained on its GPUs, marked the start of modern AI (fiscal 2026 10-K). It has invested more than $76.7 billion in research and development since its founding. It does not manufacture: TSMC and Samsung produce its wafers, it buys memory from SK hynix, Micron and Samsung, and Hon Hai (Foxconn), Wistron and Fabrinet assemble and test its products. Its fiscal year ends in late January: the second quarter of fiscal 2027 is May–July 2026. The price already reflects the 10-for-1 split of June 2024.
Two platforms. Data Center ($89.0 billion in May–July 2026; +18% vs. the previous quarter and +117% in a year; 92.5% of sales): Blackwell and Vera Rubin GPUs (in full production, with racks running at CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius), the Grace and Vera processors (Vera is 'the first CPU built for AI agents'; according to Kress, Grace already brings in more than $5 billion over twelve months and she expects that figure to more than double in fiscal 2028), Spectrum-X and Spectrum-6 networking gear, the Groq 3 LPX accelerator and software. By type of customer: big clouds $49.0 billion (+13% quarter on quarter) and AI-specialized clouds, industry and businesses $40.0 billion (+25% quarter on quarter, +138% year on year). AI for governments, sold mostly through regional clouds, grew 35% in the quarter and more than tripled in a year; networking had another record quarter (+18% quarter on quarter). Edge Computing ($7.2 billion; +13% quarter on quarter, +27% year on year): RTX PCs (including a partnership with Microsoft for RTX Spark on Windows), DGX stations, the DRIVE Hyperion platform for robotaxis (with Foxconn, VinFast, Uber and HUMAIN) and robotics (Jetson Thor, Isaac GR00T).
Results (May–July 2026, reported on August 26, 2026). Revenue $96,221 million (+18% quarter on quarter, +106% year on year), above the ~$92.2 billion analysts expected; fourth straight quarter of accelerating growth. Gross margin 75.0% (72.4% a year earlier). Operating expenses $8,408 million (+55%), of which $7,054 million in research and development. Operating income $63,734 million (+124%): 66 cents of every dollar sold. Net income $59,688 million (+126%) and $2.46 per share (+128%), which includes $7,771 million of gains on the value of its investments in other companies; without those items (adjusted figure, which since this year already deducts what it pays employees in stock) it was $53,954 million and $2.22 per share (+120%; analysts expected ~$2.10). Operating cash $24,077 million and free cash $21,341 million (in February–July, $69,895 million): less than profit because accounts receivable rose $22,346 million and inventory $5,784 million. February–July half: revenue $177,837 million (+96%) and net income $118,010 million. Baseline: fiscal 2026 (to January 2026) closed with revenue of $215,938 million (+65%) and profit of $120,067 million (+65%); in all of fiscal 2023 it had sold $26,974 million. The stock closed at $209.66 on August 26 and at $227.98 on August 27 (+8.7%).
What the company expects. August–October 2026: revenue of $108 billion ±2% (its first quarter above $100 billion; analysts expected ~$104.2 billion), gross margin of 74.0% ±0.5 points and operating expenses of ~$9.2 billion, without counting data center chip sales in China. On the call with analysts, Kress forecast revenue ~70% higher in fiscal 2028 (February 2027–January 2028), versus ~44% analysts expected, and called it 'a supply-constrained outlook': she expects supply to remain the bottleneck at least through the end of that year, and its customers' forecasts point to its sales doubling. Gross margin would bottom at 71%–72% in November 2026–January 2027 and settle at 72%–73% in fiscal 2028, when price increases already made take effect; the cause is 'extreme pricing conditions' in memory. Context the company gave: the cloud industry's backlog of pending orders exceeds $2 trillion; the five biggest clouds would invest nearly $800 billion in 2026 and $1.3 trillion in 2027; and AWS (Amazon) will deploy 2 million additional GPUs between August 2026 and July 2028.
Balance sheet and commitments (as of July 26, 2026). $56.6 billion in cash and debt investments, plus $42,783 million in shares of listed companies and $51,157 million in private companies ($22,251 million in January). In total it has $99 billion invested in companies in its ecosystem and another $25 billion committed (AI model makers, infrastructure financiers and other private companies). Debt: $33.5 billion in bonds (it issued $24,896 million in the quarter) and an unused $25 billion commercial paper program. Accounts receivable: $63,059 million ($38,466 million in January); five customers account for 22%, 14%, 13%, 11% and 10% of that balance, and the average collection period rose to 60 days due to longer terms on large purchases. Inventory: $31,575 million, preparing for the Vera Rubin ramp. Purchase commitments for manufacturing and memory: $279 billion ($119 billion at the end of April 2026). Guarantees: in August 2026 it signed guarantees of up to $105 billion for the land, power and buildings of a ~4.25-gigawatt campus that SB Energy is building in Pike County, Ohio, for an OpenAI affiliate (first phase expected in fiscal 2029; in exchange, the site will host only NVIDIA infrastructure); with other guarantees to AI clouds, its maximum exposure adds up to $108.5 billion. And it launched a model with partner AI clouds that buy its equipment while NVIDIA commits to renting capacity from them: $36 billion in commitments, typically for six years.
Dividend and buybacks. Until March 2026 it paid $0.01 per share per quarter; on May 18, 2026 it raised it to $0.25 (25 times more). In May–July it paid $6,047 million in dividends (~10% of the quarter's profit) and $19,732 million in buybacks; in total it returned ~$26 billion. The latest dividend ($0.25; record date September 10, paid October 1, 2026) no longer applies to someone buying on October 5; the next one depends on the board declaring it. At $0.25 per quarter that is $1.00 a year: ~0.42% on $236.07. Buybacks: $99 billion remained authorized as of July 26; on September 28, 2026 the board authorized $150 billion more — a record — for a total of $235 billion that the company expects to execute by the end of fiscal 2028 at the latest: ~4% of its market value. Taxes: NVIDIA is a US company; if you don't live in the US, 30% of the dividend is withheld, or less if your country has a tax treaty with the US (with a 0.42% dividend, the effect is minimal).
Stock and valuation. Bought at $236.07 (opening price on October 5, 2026 on Nasdaq). The previous record close was $235.74 (May 14, 2026) and the intraday high $237.88 (October 2, 2026): we bought it practically at its all-time high. 52-week range: $164.27 (March 30, 2026) to $240.10. In 2026 it is up ~27% (it closed 2025 at $186.50). Recent path: −20% from October 29, 2025 ($207.04) to March 30, 2026 ($165.17); +43% to May 14 ($235.74); −19% to ~$190 in late July; +24% since then. Beta of 2.2: it moves ~2.2 times as much as the market. With 24,147 million shares, at $236.07 it is worth ~$5.7 trillion. Earnings per share over the last twelve months: $7.91 (includes investment gains), price vs. profit (P/E) ~29.8x. Analysts: $9.31 per share for fiscal 2027 (51 analysts; ~25.4x) and $15.80 for fiscal 2028 (53 analysts; ~14.9x; range $9.80–$18.75); in early August 2026 they expected $12.84 for that year. For August–October they expect $2.47 per share (+90%) and $108,977 million in sales. Consensus of 61 analysts: 10 strong buy, 48 buy, 2 hold and 1 sell; average price target $328.72 (median $315; range $180–$515), 39% above the purchase price. Next report: November 17, 2026.
Risks. (1) Concentration in AI spending: 92.5% of sales are data centers; the big clouds bought $49.0 billion of the quarter; one direct customer was 16% of May–July sales (in February–July, three customers accounted for 16%, 15% and 13%), and the company estimates that one AI research company contributed a meaningful share of its sales by buying cloud services from its customers. It also finances part of that spending: $99 billion invested and $25 billion committed in its ecosystem, and guarantees of up to $108.5 billion. If AI spending slows, its sales and the value of those bets fall together. (2) Memory: 'extreme' prices will take gross margin from 75% to 71%–72% in November–January. (3) Competition from its customers' own chips: Google announced its eighth-generation TPU and already sells TPUs directly to certain clients, and Amazon's in-house chip business (Trainium) is growing at triple-digit rates (The Motley Fool, May 7, 2026). (4) China: it is not counting on data center chip sales there; the US granted licenses to sell H200s to certain Chinese customers, but the Chinese government is holding back those purchases and pushing its companies toward local competitors; shipments are less than 1% of data center sales, pay a 25% tariff when passing through the US that NVIDIA absorbs, and it took a $400 million charge for H200s without demand. (5) Collections: accounts receivable of $63,059 million and 60 days to collect; the quarter's free cash ($21,341 million) was far below profit ($59,688 million). (6) Commitments: $279 billion in manufacturing and memory purchases and $36 billion in cloud capacity; if demand falls, they weigh. (7) Supply chain concentrated in Asia, especially Taiwan and South Korea. (8) Competition regulators in the US, the European Union, the United Kingdom, China and South Korea are requesting information about its sales, how it allocates supply and its investments in AI model makers. (9) Price: we bought at its all-time high, after two ~20% drops in the last year. (10) Symbolic dividend (~0.42%): the return depends on growth.
In the portfolio. First purchase of NVIDIA. It completes the AI chain we already own: TSMC (TSM) makes its chips, Micron (MU) sells it memory, ASML and Applied Materials (AMAT) make the machines that manufacture chips, Cadence (CDNS) the software used to design them, and Broadcom (AVGO) and ARM are other pieces of the same boom. That is also a risk: with NVIDIA, the portfolio becomes even more exposed to AI spending continuing to grow.
This is not financial advice. Consult a certified financial advisor.
The author may hold positions in the securities discussed.
Past performance does not guarantee future results.