Folio · SPXSet Report · Digest
v0.7.0 STABLE · 2026-11-11 · r4

2026-W46 · Nov 9–Nov 15 · Company · Profile · published

Index >>> Company >>> Profile

NVIDIA Corporation (NVDA) — XLK Information Technology · S&P 500 Profile |輝達公司介紹

information-technologyS&P500SPX Top25xlkxlk-top-25
SPXSet Studio
IN-HOUSE RESEARCH · 站內獨立研究

NVIDIA Corporation (NVDA) — Business, History & S&P 500 Profile |輝達公司介紹

NVIDIA company profile · business segments · history · financial snapshot · S&P 500 rank #1 · 輝達公司業務歷史財務與標普500地位

中文版 ↓

PROFILE
Market data as of Oct 21, 2026 (ET) close
Financials: FY2026 (ended Jan 25, 2026)
MARKET CAP
$5,170B
Oct 16, 2026 · #1 S&P 500 by weight
S&P 500 WEIGHT
7.59%
Rank #1 · Oct 21, 2026 (Slickcharts)
FY2026 REVENUE
$215.94B
+65% YoY · record year
FY2026 NET INCOME
$120.07B
Net margin 55.6%

SNAPSHOT

TickerNVDA (NASDAQ)
CompanyNVIDIA Corporation
GICS sector / industryInformation Technology (XLK) / Semiconductors
S&P 500 rank (by weight)#1 · 7.59% (Oct 21, 2026)
Sector rank (by market cap)#1 in Information Technology (Oct 16, 2026)
HeadquartersSanta Clara, California
FoundedApril 5, 1993, Sunnyvale, California
CEOJensen Huang (co-founder; president & CEO since 1993)
Employees≈42,000 (≈31,000 in R&D) — Jan 25, 2026, FY2026 10-K
Market cap$5,170B (Oct 16, 2026)
SEC filingsEDGAR — CIK 0001045810

WHAT IT DOES

NVIDIA is a full-stack AI infrastructure company: it designs the accelerated-computing platforms (GPUs, CPUs, DPUs and networking) on which modern AI runs, and the CUDA software stack that makes those chips programmable. Its one-line model: sell the picks and shovels of the AI buildout — Blackwell/Hopper GPU systems and Mellanox networking — then deepen the moat with the CUDA developer ecosystem — millions of developers whose models are built for NVIDIA's software stack.

Revenue is reported in two segments. Compute & Networking (≈89.6% of FY2026 revenue) is the data-center business: Hopper, Blackwell and now Rubin AI accelerators sold to hyperscalers, cloud providers and enterprises, plus InfiniBand and Spectrum-X networking from the Mellanox acquisition. Graphics (≈10.4%) is the legacy consumer root — GeForce gaming GPUs and RTX workstation cards — still growing, but no longer the story.

The flywheel is platform lock-in: CUDA, cuDNN, TensorRT and NIMs mean models are trained and deployed on NVIDIA software first, which drives the next hardware purchase cycle and funds the R&D ($18.5B in FY2026, three-quarters of a 42,000-strong workforce in R&D roles) that keeps the one-generation lead.

BUSINESS SEGMENTS — FY2026 REVENUE

SegmentRevenueShareYoY
Compute & Networking$193.48B89.6%+67%
Graphics$22.46B10.4%+57%
Total$215.94B100%+65%
Source: NVIDIA FY2026 Form 10-K (fiscal year ended Jan 25, 2026). Segment operating income: Compute & Networking $130.14B · Graphics $9.16B. Compute & Networking is nine-tenths of the company.

Compute & Networking is data-center accelerated computing and networking — AI GPU platforms (Hopper, Blackwell, Rubin), Mellanox InfiniBand/Spectrum-X networking, AI software and automotive platforms. This is the AI buildout business, and it is effectively the company now.

Graphics is GeForce gaming GPUs and RTX workstation graphics. A cyclical, consumer-driven business that still grew 57% in FY2026 on the RTX 50-series cycle, but its strategic value is increasingly as the on-ramp to the CUDA ecosystem.

HISTORY

YearMilestone
1993Founded April 5 in Sunnyvale, California by Jensen Huang, Chris Malachowsky and Curtis Priem
1998RIVA TNT establishes NVIDIA in 3D graphics for the PC
1999GeForce 256 — billed as the world's first GPU; IPO on NASDAQ (NVDA) in January (Jan 22, at $12 per share)
2000Acquires the assets of rival 3dfx, ending the early 3D-chip wars
2006CUDA parallel-computing platform launches — the software bet that becomes the moat
2012AlexNet trained on NVIDIA GPUs ignites the deep-learning revolution
2016Pascal / Tesla P100 and the DGX-1 — the first AI supercomputer in a box
2020Completes Mellanox acquisition ($6.9B) for data-center networking; announces Arm bid ($40B)
2022Arm acquisition terminated on regulatory opposition; Hopper H100 unveiled
2023First $1T market-cap close (May 30) on the generative-AI wave
202410-for-1 stock split (June); Blackwell platform announced; $2T (Feb) and $3T (Jun) market caps; joins the Dow Jones (Nov), replacing Intel
2025Rubin platform roadmap detailed at GTC; US export-license regime forces a $4.5B H20 China write-off (Q1 FY2026)
2026Record Q2 FY2027: revenue $96.22B (+106% YoY); Blackwell Ultra and Rubin ramp; Anthropic, Meta and CoreWeave partnerships announced

COMPETITIVE POSITION

  • ~90% of desktop discrete GPUs — NVIDIA shipped ≈11.25M of 12.5M desktop add-in-board GPUs in Q2 2026 (Jon Peddie Research, via Tom's Hardware, Sep 2026); AMD ≈8%, Intel ≈2%.
  • The AI-accelerator standard — data-center revenue of $193.7B in FY2026 (≈90% of total revenue) reflects an overwhelming share of AI training and inference compute; no rival ships at comparable scale.
  • CUDA is the moat — two decades of libraries, compilers and developer tooling make rival chips a rewrite, not a swap; hyperscaler custom silicon (Google TPUs, Amazon Trainium, Microsoft Maia) is the only structural threat.
  • Full-stack pricing power — 71.1% gross margin and 60.4% operating margin in FY2026; systems (DGX/HGX) sell the networking and software attach, not just the die.
  • Competition set — AMD (Instinct accelerators, Radeon), Intel (Gaudi, Arc), hyperscaler custom chips, and Broadcom/Marvell in AI networking.

FINANCIAL SNAPSHOT — FY2026

Revenue$215.94B (+65% YoY; FY2025 $130.50B)
Gross margin$153.46B · 71.1% (FY2025 75.0%)
Operating income$130.39B · 60.4% margin
Net income$120.07B · 55.6% margin
Diluted EPS$4.90 (vs $2.94 in FY2025)
Operating cash flow$102.72B (FY2025 $64.09B)
Free cash flow≈$96.6B (OCF $102.72B − capex $6.1B)
Cash & marketable securities$62.56B (Jan 25, 2026)
Total debt$8.47B — net cash position
Capital returned$40.4B buybacks (282M shares) · $974M dividends paid in FY2026
R&D expense$18.50B
Latest quarter (Q2 FY2027, ended Jul 26, 2026)Revenue $96.22B (+106% YoY) · net income $59.69B (+126%, incl. $7.8B equity-securities gains) · gross margin 75.0% · diluted EPS $2.46
Source: NVIDIA FY2026 Form 10-K and Q2 FY2027 10-Q / earnings release. FY2026 includes a $4.5B H20 inventory charge from the China export-license regime.

KEY RISKS

  • Hyperscaler customer concentration — a handful of cloud customers drive the $193.7B data-center business; any capex pause or share shift to in-house silicon hits the whole P&L.
  • China is effectively closed — the $4.5B H20 write-off in Q1 FY2026 and a guided zero for China data-center compute revenue; export-license policy can change the addressable market overnight.
  • Custom-silicon substitution — Google TPUs, Amazon Trainium and Microsoft Maia reduce hyperscaler reliance on NVIDIA GPUs over time, even if volumes still grow.
  • Financier-of-the-boom exposure — growing equity stakes and commitments to neocloud buyers (e.g., CoreWeave) plus large guarantees tied to OpenAI-linked data-center projects turn customer credit risk into NVIDIA's balance-sheet risk.
  • Fabless supply dependence — leading-edge supply rests on TSMC; supply-related purchase commitments reached $119B (Q1 FY2027 disclosure), a fixed bet on continued demand.
  • Premium expectations — priced for flawless AI-capex growth; any deceleration in the buildout would re-rate the multiple long before revenue actually fell.

WHY IT MATTERS FOR THE S&P 500

NVIDIA is the index's largest constituent at a 7.59% weight (Oct 21, 2026) — a 1% move in NVDA alone moves the S&P 500 by roughly 7.6 basis points. Together with Apple (≈7.1%), two names carry about 14.7% of the index, so NVIDIA's print is the single most-watched earnings event in the US market: it reads as a verdict on the entire AI-capex cycle, not just one company.

For index and 0DTE traders specifically: NVIDIA's late-February / late-August earnings fall inside heavy options positioning windows, its weeklies are among the highest-premium single-stock options in the market, and its direction on mega-cap breadth days often decides whether an SPX move is broad or narrow.

SOURCES & METHODOLOGY

Financials, segments, employees, cash/debtNVIDIA FY2026 Form 10-K (filed Feb 25, 2026; fiscal year ended Jan 25, 2026; accession 0001045810-26-000021)
Latest quarter (Q2 FY2027)NVIDIA Q2 FY2027 10-Q (ended Jul 26, 2026) and company earnings release, Aug 2026
Index weight & rankSlickcharts S&P 500 weights, Oct 21, 2026 close (assignment baseline)
Market cap & sector rankOct 16, 2026 snapshot (SPXSet Top-25 dataset; assignment baseline)
GPU market shareJon Peddie Research Q2 2026 desktop AIB data, via Tom's Hardware (Sep 11, 2026)
LogoSPXSet logo library, compiled from company official websites, filings, or Wikipedia; as company logos are dynamic corporate information, no 100% guarantee is made for the graphics.

Price Chart — 10-Year Monthly Candles

NVIDIA

Source: Yahoo Finance monthly bars, split-adjusted · Data as of Sep 30, 2026 (last closed month) · SPXSet.com

FINANCIALS

繁體中文版

公司快照

股票代號NVDA(NASDAQ)
公司NVIDIA Corporation(輝達公司)
GICS 板塊/行業資訊科技(XLK)/半導體
標普500排名(按權重)#1 · 7.59%(2026年10月21日)
板塊內排名(按市值)資訊科技 #1(2026年10月16日)
總部加州聖塔克拉拉
成立1993年4月5日,加州桑尼維爾
執行長黃仁勳(共同創辦人;1993年起任總裁兼執行長)
員工約42,000名(研發約31,000名)——2026年1月25日,FY2026 10-K
市值$5,170B(2026年10月16日)
SEC 申報EDGAR — CIK 0001045810

業務概覽

輝達是一家全端 AI 基礎設施公司:設計現代 AI 運算所依賴的加速運算平台(GPU、CPU、DPU 與網路設備),以及讓這些晶片可被程式化的 CUDA 軟體堆疊。一句話商業模式:賣 AI 大建設的鏟子——Blackwell/Hopper GPU 系統與 Mellanox 網路設備——再用 CUDA 開發者生態把護城河越挖越深,數以百萬計開發者的模型都建構在輝達的軟體堆疊上。

收入按兩個板塊申報。運算與網路(佔 FY2026 營收約 89.6%)是資料中心業務:賣給超大規模雲端商、雲端業者與企業的 Hopper、Blackwell、Rubin AI 加速器,加上來自收購 Mellanox 的 InfiniBand/Spectrum-X 網路設備。圖形(約 10.4%)是消費端老本行——GeForce 遊戲 GPU 與 RTX 工作站顯卡——仍在成長,但早已不是故事主角。

飛輪就是平台鎖定:CUDA、cuDNN、TensorRT 與 NIMs 讓模型先在輝達軟體上訓練與部署,帶動下一輪硬體採購,再支撐研發投入(FY2026 研發 185 億美元,42,000 名員工中約四分之三從事研發),維持領先同業一個世代的優勢。

業務板塊 — FY2026 營收

板塊營收佔比同比
運算與網路$193.48B89.6%+67%
圖形$22.46B10.4%+57%
合計$215.94B100%+65%
來源:NVIDIA FY2026 Form 10-K(會計年度截至2026年1月25日)。板塊營業利潤:運算與網路 $130.14B · 圖形 $9.16B。運算與網路佔公司九成。

運算與網路是資料中心加速運算與網路——AI GPU 平台(Hopper、Blackwell、Rubin)、Mellanox InfiniBand/Spectrum-X 網路、AI 軟體與車用平台。這就是 AI 大建設本身,實際上等於整家公司。

圖形是 GeForce 遊戲 GPU 與 RTX 工作站顯卡。消費週期型業務,FY2026 靠 RTX 50 系列週期仍成長 57%,但戰略價值越來越是通往 CUDA 生態的入口。

歷史沿革

年份里程碑
19934月5日由黃仁勳、Chris Malachowsky 與 Curtis Priem 在桑尼維爾創立
1998RIVA TNT 在 PC 3D 顯卡市場站穩腳跟
1999GeForce 256——號稱全球首款 GPU;1月22日在 NASDAQ 上市(NVDA,發行價 $12)
2000收購對手 3dfx 資產,終結早期 3D 晶片大戰
2006CUDA 平行運算平台推出——後來成為護城河的軟體賭注
2012AlexNet 在輝達 GPU 上訓練成功,點燃深度學習革命
2016Pascal/Tesla P100 與 DGX-1——第一台「盒裝 AI 超級電腦」
2020完成收購 Mellanox(69億美元)補強資料中心網路;宣布收購 Arm(400億美元)
2022Arm 收購因監管反對終止;發表 Hopper H100
2023乘生成式 AI 浪潮,市值首度收盤突破 1 兆美元(5月30日)
20241 拆 10 股票分割(6月);發表 Blackwell 平台;市值先後突破 2 兆(2月)與 3 兆美元(6月);11月納入道瓊指數,取代英特爾
2025GTC 公布 Rubin 平台路線圖;美國出口許可制迫使 H20 中國業務認列 45 億美元減損(FY2026 第一季)
2026FY2027 第二季創紀錄:營收 962.2 億美元(同比 +106%);Blackwell Ultra 與 Rubin 放量;宣布與 Anthropic、Meta、CoreWeave 的合作

競爭地位

  • 桌上型獨立顯卡約九成市佔——2026 年第二季出貨約 1,125 萬張,佔 1,250 萬張桌上型 AIB 顯卡市場的九成(Jon Peddie Research,經 Tom's Hardware 2026年9月報導);AMD 約 8%,英特爾約 2%。
  • AI 加速器的事實標準——FY2026 資料中心營收 1,937 億美元(約佔總營收九成),反映其在 AI 訓練與推論算力的壓倒性份額;沒有對手能以同等規模出貨。
  • CUDA 就是護城河——二十年累積的函式庫、編譯器與開發工具,讓對手晶片意味著重寫而非替換;唯一結構性威脅是超大規模雲端商的自研晶片(Google TPU、亞馬遜 Trainium、微軟 Maia)。
  • 全端定價權——FY2026 毛利率 71.1%、營業利潤率 60.4%;賣的是整機系統(DGX/HGX),網路與軟體一起搭售,不只是晶片裸片。
  • 競爭集合——AMD(Instinct 加速器、Radeon)、英特爾(Gaudi、Arc)、雲端商自研晶片,以及博通/Marvell 的 AI 網路業務。

財務快照 — FY2026

營收$215.94B(同比 +65%;FY2025 $130.50B)
毛利$153.46B · 71.1%(FY2025 75.0%)
營業利潤$130.39B · 利潤率 60.4%
淨利潤$120.07B · 淨利率 55.6%
稀釋 EPS$4.90(FY2025 為 $2.94)
經營現金流$102.72B(FY2025 $64.09B)
自由現金流約$96.6B(經營現金流 $102.72B − 資本支出 $6.1B)
現金與有價證券$62.56B(2026年1月25日)
總債務$8.47B——淨現金狀態
資本返還FY2026 回購 $40.4B(2.82億股)· 已付股息 $974M
研發費用$18.50B
最新一季(FY2027 第二季,截至2026年7月26日)營收 $96.22B(同比 +106%)· 淨利潤 $59.69B(同比 +126%,含 78 億美元股權證券收益)· 毛利率 75.0% · 稀釋 EPS $2.46
來源:NVIDIA FY2026 Form 10-K 與 FY2027 第二季 10-Q/財報。FY2026 含中國出口許可制下 45 億美元 H20 存貨減損。

主要風險

  • 超大規模客戶集中——1,937 億美元的資料中心業務繫於少數雲端大客戶;任何資本支出暫停或轉向自研晶片,都直接衝擊整張損益表。
  • 中國市場實質關閉——FY2026 第一季 45 億美元 H20 減損,指引中中國資料中心算力營收為零;出口許可政策一變,可觸及市場規模隔夜改變。
  • 自研晶片替代——Google TPU、亞馬遜 Trainium、微軟 Maia 長期降低超大規模商對輝達 GPU 的依賴,即便總量仍在成長。
  • 成為 AI 熱潮的出資方——對新興雲端買家(如 CoreWeave)日益增加的股權投資與承諾,加上與 OpenAI 相關資料中心項目的巨額擔保,把客戶信用風險變成輝達自身的資產負債表風險。
  • 無廠模式的供應依賴——先進製程供應繫於台積電;供應相關採購承諾達 1,190 億美元(FY2027 第一季披露),是對需求持續的固定押注。
  • 高預期定價——股價已反映 AI 資本支出完美成長;建設放緩會先重估倍數,遠早於營收真正下滑。

對標普500的重要性

輝達是指數第一大成分股,權重 7.59%(2026年10月21日)——NVDA 單獨波動 1%,標普500 就大約動 7.6 個基點。加上蘋果(約 7.1%),兩家公司合計佔指數約 14.7%,因此輝達財報是全美市場最受矚目的財報事件:它被市場當作對整個 AI 資本支出週期的定調,而不只是一家公司的成績單。

對指數與 0DTE 交易者:輝達每年 2 月底與 8 月底的財報常落在期權倉位沉重的窗口,其週期權是市場上權利金最高的個股期權之一,而在大型股寬度主導的日子,輝達的方向往往決定 SPX 的走勢是全面還是狹窄。

來源與方法

財務、板塊、員工、現金/債務NVIDIA FY2026 Form 10-K(2026年2月25日申報;會計年度截至2026年1月25日;文號 0001045810-26-000021)
最新一季(FY2027 第二季)NVIDIA FY2027 第二季 10-Q(截至2026年7月26日)與公司財報,2026年8月
指數權重與排名Slickcharts 標普500權重,2026年10月21日收盤(任務基線)
市值與板塊排名2026年10月16日快照(SPXSet Top-25 數據集;任務基線)
GPU 市佔Jon Peddie Research 2026年第二季桌上型 AIB 數據,經 Tom's Hardware(2026年9月11日)報導
LogoSPXSet logo 庫,整理自企業官方網站、文件或維基百科;由於是企業自身動態資訊,不對圖形做百分百保證。

10年股價K線圖(月K)

NVIDIA

來源:Yahoo Finance 月 K(經拆股調整)· 數據截至 2026 年 9 月 30 日(最後已收月份) · SPXSet.com

財報

SPXSet Studio

SPXSet in-house research — market structure, options and volatility. Not investment advice.
SPXSet 站內獨立研究,聚焦 SPX 市場結構、期權與波動率;內容不構成投資建議。

SPXSet · Company Library · NVDA
Data as of Oct 21, 2026 · Financials FY2026 (10-K) · Market data Oct 2026
Previous · Apple Inc (AAPL) — XLK Information Technology · S&P 500 Profile |蘋果公司介紹

Entered 2026-10-22 22:20 · updated 2026-11-14 19:28 · rev 20