Builds high-speed networking fabric inside cloud and AI data centers. Revenue $2.3B (2020) → est. $8.5B (2025), ~4× in 5 years, ~30% CAGR. Net margins 27% → 42%+. FCF margins 50%+, CapEx under 1% of revenue, $10B+ cumulative FCF. $9.5B cash vs $60M debt. As AI clusters scale, networking becomes the bottleneck, not compute. Key metric: revenue growth in mid-20% range = hyperscaler spending intact.

构建云和AI数据中心内部高速网络结构。营收$2.3B(2020)→预估$8.5B(2025)，5年约4倍，~30% CAGR。净利润率27%→42%+。FCF利润率50%+，CapEx低于营收1%，累计FCF 100亿美元+。现金95亿美元 vs 债务6000万。AI集群扩展时，网络成瓶颈而非算力。关键指标：营收增长保持20%+区间=超大规模支出完好。

Builds large-scale AI data centers around compute density and power cost. Revenue $8.5M (FY2022) → $215M (FY2025), 194% CAGR — 7× faster than broader AI DC market. Builds in power-rich regions (North Dakota), locks low-cost kWh into 10–15-year leases. Ellendale campus online Q4, adding ~$50M/quarter initially. Targeting 45%+ gross margins once fully leased. Key metric: MW placed into service — ~$200M annualized revenue per 100MW.

围绕算力密度和电力成本建设大规模AI数据中心。营收$8.5M(FY2022)→$215M(FY2025)，194% CAGR——比AI DC市场快7倍。在电力丰富地区（北达科他）建设，锁定低成本千瓦时10-15年租约。Ellendale园区Q4上线，初期增~$5000万/季。目标45%+毛利率（满租后）。关键指标：投运MW——每100MW约$2亿年化营收。

Moved from niche drone supplier to broader defense tech platform. $4.1B BlueHalo acquisition tripled revenue, added lasers, space systems, counter-UAS. Switchblade production 6× in ~1 year, capacity for thousands annually. Funded backlog $400M → $750M+, total bookings >$1B. Long-dated Army/Air Force contracts extend visibility into next decade. Gross margin collapse to high teens = acquisition accounting, not demand problem. Key metric: gross margin recovery to mid-20% = integration progressing.

从利基无人机供应商变为更广泛的国防科技平台。41亿美元BlueHalo收购使营收三倍，增加激光、太空系统、反无人机。Switchblade产量约1年6倍，年产数千台能力。已签约积压$4亿→$7.5亿+，总预订超10亿。长期陆军/空军合同延续至下个十年。毛利率暴跌至高十位数=收购会计，非需求问题。关键指标：毛利率恢复至20%+中段=整合推进中。

Quarterly revenue low teens → ~$40M in Q3 2025 — almost all of 2024 in a single quarter. Gross margins below 10% by design — selling systems at near cost to get embedded early. Two-qubit gate fidelity at 99.99% — fewer physical qubits needed for useful work. Stock ran $8 → mid-$80s in 2025, pulled back ~40% to ~$50 — consolidation after repricing. Key metric: revenue trajectory step-ups = turning quantum into a real platform business.

季度营收低十数百万→2025Q3约$4000万——一个季度接近2024全年。毛利率低于10%是刻意为之——以近成本价出售系统提前嵌入。双量子比特门保真度99.99%——更少物理量子比特即可做有用工作。股价2025年从$8→80多美元，回调约40%至~$50——重新定价后盘整。关键指标：营收阶梯式增长=量子变成真正的平台业务。

Data center revenue $51B in Q3 FY26, up 66% YoY, now 90% of total. Vera Rubia platform: 5× inference, 3.5× training, tokens at 1/10 cost. Compute scales three ways: training + post-training + inference. Nvidia AI Enterprise at $2B+ annual pace, triple-digit growth. Most global computing infrastructure hasn't transitioned to accelerated compute yet. Key metric: data center revenue growth = confirms Nvidia's role at center of compute transition.

数据中心营收FY26Q3达$510亿，同比涨66%，占总营收90%。Vera Rubia平台：5倍推理、3.5倍训练、token成本1/10。算力三种扩展：训练+后训练+推理。英伟达AI企业年化20亿美元+，三位数增长。全球大部分计算基础设施尚未转向加速计算。关键指标：数据中心营收增长=确认英伟达在算力转型中的核心地位。

Applies etch and deposition tools that make advanced logic and memory possible. Revenue $1.6B (FY2006) → $18.4B (FY2025) — 11×, roughly double semiconductor market growth. Operating margins mid-20s → 32%, net margins 29%. HBM requires 2–3× more etch steps per wafer — new demand vector growing 50%+ annually. FCF $5.7B at 31% margin, returning 80% to shareholders. Drawdowns improved from -56% (GFC) to -15% (recent normalization). Key metric: operating margin expanding faster than revenue = operating leverage intact.

提供蚀刻和沉积工具使先进逻辑和内存成为可能。营收$1.6B(FY2006)→$18.4B(FY2025)——11倍，约为半导体市场增速2倍。运营利润率20%+中段→32%，净利率29%。HBM每晶圆需2-3倍更多蚀刻步骤——年增50%+的新需求向量。FCF $57亿利润率31%，80%回馈股东。回撤从-56%(GFC)改善至-15%。关键指标：运营利润率扩展快于营收=运营杠杆完好。

Amazon surges ahead, then goes sideways for years while business catches up. Since 2023: up 15–20% while S&P 500 up 45–50%. Two years of underperformance right on pattern. AWS growth accelerating back toward 20%. Operating income $12B (2022) → $69B (2024) — 5× in 2 years. FCF swung from negative to $30B+. Buildout funded by profits, not balance sheet stress. Stock compressed near highs, momentum constructive but not overheated. Key metric: returns as DC capacity fills = breakout signal.

Amazon先飙升再横盘数年等业务赶上。2023年以来：涨15-20%而标普涨45-50%。两年的跑输正合模式。AWS增速回到20%。运营利润$120亿(2022)→$690亿(2024)——2年5倍。FCF从负转$300亿+。扩建由利润而非负债表压力资助。股价在高位盘整，动量建设性但未过热。关键指标：DC容量填充时回报=突破信号。

Revenue $10B (2020) → $34B (2025) — 3× in 5 years. PEG ratio 0.47 — below 0.5 means market is heavily discounting growth even though earnings say otherwise. Data center revenue 60%+ YoY, now most important profit driver. $3B/yr non-cash amortization from Xilinx depresses GAAP earnings but not cash (FCF $2.4B, sharply up). Moving from chips to rack-scale AI systems — revenue per deployment rises without needing explosive unit growth. Key metric: PEG ratio staying this low = market betting growth won't last, even though it already is.

营收$100亿(2020)→$340亿(2025)——5年3倍。PEG比率0.47——低于0.5意味着市场严重低估增长即使盈利说相反。数据中心营收同比60%+，现在是最重要利润驱动。Xilinx每年30亿美元非现金摊销压低GAAP盈利但不影响现金流(FCF $24亿，大幅上升)。从芯片转向整机架AI系统——每次部署营收上升无需爆发性单位增长。关键指标：PEG如此低=市场赌增长持续不了，但增长已经在持续。