Semiconductor stocks June 2026 — sources and research notes

This is the sources companion to the 30-stock semiconductor guide for June 2026. Every quantitative claim in the main guide is grounded in the data below — pulled from the latest filings, earnings calls, and reputable secondary sources as of early June 2026. Use this page to verify any specific figure against its original source before acting on it.

Data ages. Multiples and stock prices move daily. Treat the per-name numbers as point-in-time references for relative ordering — for current trading levels, pull the latest 10-Q and the live market quote.

---

Part 1 — Per-name data sheets

AI accelerators & GPU

NVDA — NVIDIA

• Latest quarter: Q1 FY27 (quarter ended April 26, 2026).
• Total revenue: $81.6 billion (+85% YoY).
• Data Center revenue: $75.2 billion (+92% YoY, +21% QoQ) — ~92% of total.
• Within Data Center: Hyperscale $37.9 billion + AI Clouds / Industrial / Enterprise $37.4 billion.
• Forward P/E: approximately 25× on FY27 estimates — below the 5-year trailing average around 70×. Earnings have outrun price.
• Moat: CUDA software ecosystem + Mellanox networking + Blackwell/Rubin roadmap.
• Catalyst: Blackwell B200 + GB200 NVL72 systems shipping; Rubin GR200 in 2027.
• Bear case: ASIC threat from Google TPU, Meta MTIA, AWS Trainium compressing merchant-silicon share.
• Sources: NVDA Q1 FY27 8-K (SEC) · NVDA Q1 FY27 CFO commentary (SEC) · NVDA Q1 FY27 10-Q (SEC)

AMD — Advanced Micro Devices

• Latest quarter: Q1 FY26 (calendar Q1 2026), reported May 2026.
• Revenue: $10.3 billion (+38% YoY). EPS $1.37.
• Non-GAAP gross margin: approximately 54%.
• Segment mix: Data Center $5.8 billion (~56% of revenue, +57% YoY) — EPYC CPUs + Instinct MI3xx/MI400 GPUs · Client + Gaming $3.6 billion (~35%, +23% YoY) · Embedded $873 million (~8%, +6% YoY).
• Forward P/E: approximately 45–69× depending on EPS basis; most defensible read ~50× on FY27e EPS.
• EV/Sales: approximately 9× FY26e.
• Customer concentration: Microsoft, Meta, Oracle disclosed multi-billion-dollar MI3xx commitments in 2025–26. Top hyperscalers cluster ~30–40% of Data Center revenue.
• Moat: Only credible #2 to NVIDIA in merchant data-center GPUs. x86 EPYC + ROCm software ecosystem creates platform lock-in for hyperscalers seeking second-source.
• Catalyst: MI400 series ramp H2 2026; OpenAI 6 GW deployment (announced 2025) at scale.
• Bear case: NVDA's CUDA + Blackwell/Rubin pace means AMD perpetually plays catch-up; margin pressure from custom-ASIC encroachment.
• Sources: AMD Q1 FY26 8-K (SEC) · Stocktitan AMD Q1 8-K · AMD forward P/E (GuruFocus)

AVGO — Broadcom

• Latest quarter: Q1 FY26.
• AI revenue: $8.4 billion (+106% YoY); Q2 guide $10.7 billion.
• AI backlog: $73 billion.
• 2027 AI revenue target: above $100 billion (CEO Hock Tan public guidance).
• Custom AI accelerator market share: approximately 70%.
• Meta MTIA deal: announced April 14, 2026; multi-generation through 2029; 1 GW initial deployment; "industry's first 2nm AI compute accelerator."
• Anthropic / Google TPU: Anthropic has access to 3.5 GW via the Google–Broadcom TPU capacity.
• VMware: integration tracking ahead of plan.
• Forward P/E: approximately 31×.
• Sources: Broadcom Meta MTIA partnership (CNBC) · Broadcom Q1 FY26 (Futurum) · Custom AI ASIC state of play (Tom's Hardware) · Broadcom $100B 2027 target (TIKR)

MRVL — Marvell Technology

• Latest quarter: Q1 FY26 (calendar Q1 2026), reported late May 2026.
• Revenue: $1.895 billion (+63% YoY). Non-GAAP EPS $0.62.
• Non-GAAP gross margin: approximately 60%.
• Segment mix: Data Center $1.8 billion (76% of revenue, +27% YoY) — custom AI silicon (AWS Trainium successor work, Microsoft custom, Meta MTIA-adjacent) + 800G/1.6T electro-optics. Enterprise networking + carrier + consumer + auto/industrial = remaining ~24%, mostly flat to soft.
• Forward P/E: approximately 50–55×.
• EV/Sales: approximately 10–12× FY27e revenue.
• Customer concentration: Top customer (almost certainly AWS) above 25% disclosed; top 3 likely ~50%+ on custom silicon.
• Moat: Co-design partner for hyperscaler custom AI silicon — multi-year, sticky NRE engagements. Optical DSP IP (Inphi acquisition) is structural in datacom.
• Catalyst: FY27 revenue guide raised to approximately $11.5 billion (+40%); FY28 outlook approximately $16.5 billion (+45%). Custom silicon ramp at Amazon and Microsoft Q3–Q4 CY26.
• Bear case: Custom silicon margins lower than merchant GPU; hyperscalers can in-source if Marvell stumbles on a node transition; concentration risk if AWS Trainium roadmap shifts.
• Sources: Marvell Q1 FY26 8-K (SEC) · Tech-insider Marvell custom AI chip · MRVL forward P/E (GuruFocus)

Edge / specialty silicon

QCOM — Qualcomm

• Latest quarter: Q2 FY26 (CQ1 2026), reported April 2026.
• Revenue: $10.6 billion. Non-GAAP EPS $2.65.
• Non-GAAP gross margin: approximately 56%. Operating margin approximately 30%.
• Segment mix: Handsets $6.02 billion (57%, −13% YoY) — memory constraints + China OEM softness. Automotive $1.33 billion (12.5%, +38% YoY) — record quarter. IoT $1.73 billion (16%, +9% YoY). Licensing (QTL) approximately $1.3 billion, flat.
• Forward P/E: approximately 18–22× — the cheapest large-cap semi in the universe.
• EV/Sales: approximately 4× FY26e.
• Customer concentration: Apple modem revenue rolling off (now ~10% and shrinking — last 5G iPhone modem socket lost). Samsung, Xiaomi, Honor account for bulk of remaining handset revenue.
• Moat: Snapdragon SoC + RF front-end integration; standards-essential 5G/6G patents (QTL); Snapdragon Auto Connect platform locked in at approximately 30 OEMs.
• Catalyst: Auto record continues; ByteDance edge-AI deal (signed May 2026, "biggest AI deal in QCOM history" per management) scales H2 2026; Snapdragon X PC traction.
• Bear case: Apple modem fade is permanent; handset re-rating cap; China handset cycle uncertain; QTL royalty pool secularly shrinking.
• Sources: Qualcomm Q2 FY26 8-K (SEC) · QCOM forward P/E (GuruFocus) · Qualcomm ByteDance AI deal

AMBA — Ambarella

• Latest quarter: Q1 FY27 (CQ1 2026), reported May 2026.
• Revenue: $100.4 million (+16.9% YoY).
• Gross margin: 58.4% GAAP.
• Segment mix: Edge AI SoCs = 80% of revenue (computer vision, robotics, physical AI). Auto record quarter — driven by commercial vehicle AI penetration. Remainder: legacy IP cameras / IoT security.
• EV/sales: approximately 10–12× FY27e (not yet GAAP profitable).
• Customer concentration: Tier-1 auto OEM and Tier-1 industrial customers — specific names not disclosed; top 5 customers historically ~50–60% of revenue.
• Moat: CV-class edge AI inference at ultra-low power (sub-5W envelopes) — proprietary CVflow architecture; >10 years of design-win lead on physical AI.
• Catalyst: Q2 FY27 guide $105–111 million; long-term customer agreements being signed — "new phase of market development." Commercial vehicle AI ramp continues.
• Bear case: Subscale vs Nvidia Jetson / Qualcomm Auto; lumpy quarterly print; cyclical exposure to consumer security cameras.
• Sources: Ambarella Q1 FY27 8-K (SEC) · Ambarella Q1 highlights (Yahoo)

MPWR — Monolithic Power Systems

• Latest quarter: Q1 2026, reported early May 2026.
• Revenue: $804 million (+26% YoY). Adjusted EPS $5.10 (+26% YoY).
• Non-GAAP gross margin: approximately 55% (recent quarters); Q1 showed a margin step-down.
• Segment mix: Enterprise Data +97.7% YoY — AI accelerator power, server VRMs, optical module power. Storage + Computing solid double-digit. Auto, Industrial, Comm mid-single to low-double.
• Forward P/E: approximately 38–42× (directional).
• EV/Sales: approximately 13–15× FY26e.
• Customer concentration: NVIDIA disclosed as material customer in 2024 (estimated >15% of revenue); top 3 likely >35%.
• Moat: Monolithic process design — integrates analog + digital on one die at lower BOM cost than peers; dominant share of vertical power stage in Hopper/Blackwell platforms.
• Catalyst: Enterprise Data growth floor raised from +50% to +85% YoY; manufacturing capacity target raised to $6 billion; Q2 guide $890–910 million.
• Bear case: GM step-down in Q1 — possibly mix-driven, possibly Renesas/Infineon winning share back. AI cycle dependency is now extreme.
• Sources: MPWR Q1 2026 (Simply Wall St) · Kavout MPWR surging

ALAB — Astera Labs

• Latest quarter: Q1 2026, reported May 2026.
• Revenue: $308.4 million (+93% YoY, +14% QoQ).
• Non-GAAP gross margin: approximately 74%. Non-GAAP operating margin 36.2% on $308 million ($111.7 million operating income).
• Segment mix (product line): PCIe retimers (Aries) · Smart Cable Modules (Taurus, Ethernet scale-out) · Smart Fabric Switches (Scorpio P-Series, X-Series 320-lane) · CXL memory controllers (Leo). PCIe 6 contributed above one-third of Q1 revenue.
• Forward P/E: approximately 80–100× (high-growth specialty connectivity multiple).
• EV/Sales: approximately 30× FY26e.
• Customer concentration: Heavy NVIDIA + AWS + Microsoft exposure; top 3 likely 60–70%. NVLink Fusion custom solutions tie ALAB to NVIDIA roadmap.
• Moat: PCIe 6 retimer IP lead; first-mover on AI fabric switches; NVLink Fusion design partner — locked in at silicon level on Blackwell / Rubin platforms.
• Catalyst: Scorpio X-Series 320-lane fabric switch ramp; NVLink Fusion scale-up content per system; PCIe 6 broader adoption.
• Bear case: Single-thesis stock — if AI fabric topology evolves (e.g., direct optical from chip), ALAB is the displaced layer. Valuation leaves no margin for execution stumbles.
• Sources: Astera Labs Q1 2026 8-K (SEC) · Astera Labs IR Q1 release

CRDO — Credo Technology

• Latest quarter: Q4 FY26 (CQ1 2026), reported late May 2026.
• Revenue: Q4 $437 million (+157% YoY); FY26 full year $1.3 billion (+206% YoY).
• Gross margin: 68.2% GAAP / 68.3% non-GAAP — at the top end of fabless semi.
• Segment mix (product line): Active Electrical Cables (AEC) — core, scales with AI cluster rack-to-rack interconnect · Optical DSPs + SiPho + Zero-Flap Optics — expected to deliver above $600 million in FY27 · PCIe retimers, line-card PHYs — minority but growing.
• Forward P/E: approximately 60–80× (extreme growth premium).
• EV/Sales: approximately 25× FY27e.
• Customer concentration: Historical concentration extreme — top customer reportedly above 40% (Microsoft AEC); top 3 around 70%.
• Moat: AEC technology — proprietary mixed-signal architecture that delivers optical-like reach at copper economics for sub-7m intra-rack links.
• Catalyst: Optical DSP ramp + new SiPho transceivers + Zero-Flap Optics diversify away from AEC concentration. 1.6T product cycle 2026–27.
• Bear case: Hyperscaler customer concentration; if Microsoft AEC ramp plateaus, FY27 optical contribution must come through cleanly. Margins peak.
• Sources: Credo Q4 FY26 8-K (SEC) · Credo Q4 FY26 highlights (GuruFocus)

Analog incumbents

TXN — Texas Instruments

• Latest quarter: Q1 2026, reported April 2026.
• Revenue: $4.83 billion (+19% YoY). EPS $1.68.
• Operating margin: Analog OPM 41.7%; Embedded improving sharply.
• Segment mix: Analog $3.92 billion (81%, +22% YoY) · Embedded Processing $723 million (15%, +12% YoY) · Other approximately $190 million.
• End-market signal: Industrial +30% YoY; Data Center approximately +90% YoY — TI is now AI-exposed through power management, signal chain, sensors.
• Forward P/E: approximately 26–35× depending on EPS basis.
• EV/Sales: approximately 10× FY26e.
• Customer concentration: None material; tens of thousands of customers. Distribution-heavy. Apple historically below 10%.
• Moat: 300mm internal manufacturing scale + portfolio breadth (above 80,000 SKUs) + direct distribution. Capex cycle (committed above $30 billion over 2023–26) now winding down, FCF inflection ahead.
• Catalyst: Q2 guide $5.0–5.4 billion (above consensus); $7.5 billion Silicon Labs acquisition closes H1 2027.
• Bear case: Mature industrial / auto cycle — margin compresses as utilisation underwhelms on heavy new fab capacity; AI exposure is real but only approximately 10% of revenue.
• Sources: TXN Q1 2026 8-K (Stocktitan) · TXN Q1 highlights (Futurum) · TXN forward P/E (GuruFocus)

ADI — Analog Devices

• Latest quarter: Q2 FY26 (CQ1 2026), reported May 2026.
• Revenue: $3.62 billion (+37% YoY) — record.
• Non-GAAP gross margin: 73.0% (+360 bps YoY).
• Non-GAAP operating margin: approximately 42%.
• Segment mix: Industrial $1.80 billion (50%, +56% YoY) — aerospace + defense + automation + test. Automotive $872 million (24%, +2% YoY) — BMS double-digit YoY for first time in two years. Communications approximately 13%. Consumer approximately 13%.
• End-market signal: Data center power + signal chain growing; management ties lift to "AI-linked data center and power demand."
• Forward P/E: approximately 32–35× (directional).
• EV/Sales: approximately 12× FY26e.
• Customer concentration: Diversified; no customer above 10%.
• Moat: Precision analog + RF (Hittite legacy) + Maxim integration — highest-margin analog franchise; industrial / aerospace certifications create long product cycles.
• Catalyst: Industrial cycle re-acceleration; auto BMS recovery; defense electronics spend rising.
• Bear case: Industrial cycle could roll over; defense spend politically sensitive; consensus already reflects strong recovery.
• Sources: ADI Q2 FY26 8-K (SEC) · ADI Q2 highlights (GuruFocus)

ON — Onsemi

• Latest quarter: Q1 2026, reported May 2026.
• Revenue: $1.513 billion (above guidance mid). Non-GAAP EPS $0.64.
• Non-GAAP gross margin: approximately 38.5% (expanding sequentially).
• Segment mix: Automotive $797 million (53%, +5% YoY — first YoY gain after 7 declining quarters) · Industrial + Cloud Power: balance — AI data center +30% QoQ · SiC: largest pure-play merchant SiC ex-Wolfspeed; Treo platform revenue +2.5× QoQ.
• Forward P/E: approximately 22–28× (directional based on consensus EPS recovery).
• EV/Sales: approximately 4–5×.
• Customer concentration: Auto OEM heavy — Tesla, BYD, Hyundai-Kia, German big-3 disclosed in past filings. No single above 10%.
• Moat: SiC vertical integration (substrate + epitaxy + device + module) — only Western pure-play competitor to STMicro / Infineon in SiC; design-locked in EVs.
• Catalyst: Q2 guide $1.535–1.635 billion; H2 above H1 visibility; AI data center power becoming material; SiC share of 2026 Beijing Auto Show new EVs approximately 55%.
• Bear case: SiC pricing under pressure from China entrants; Sony winning share back in CIS image sensors; EV cycle still uncertain post-Tesla deliveries plateau.
• Sources: Onsemi Q1 2026 8-K (SEC) · Onsemi Q1 (BigGo) · 24/7 Wall St Onsemi SiC slowdown

MCHP — Microchip

• Latest quarter: Q4 FY26 (CQ1 2026), reported May 2026.
• Revenue: Q4 $1.311 billion (+35.1% YoY, +10.6% QoQ). FY26 $4.713 billion (+7.1% YoY).
• Non-GAAP gross margin: 61.6%. Non-GAAP operating margin: 30.6%. EPS $0.57.
• Segment mix: Microchip reports by end-market (Industrial, Automotive, Computing, Consumer, Comm, Defense + Aerospace). Industrial + Auto historically approximately 55% combined; data center / computing approximately 15–20%.
• Forward P/E: approximately 30–35× (recovery-cycle multiple).
• EV/Sales: approximately 8×.
• Customer concentration: None material; distribution-heavy.
• Moat: Broadest MCU + analog catalog among mid-cap analog incumbents; sticky industrial design wins; nine-point recovery plan now visibly working.
• Catalyst: Distribution inventory below normal — restocking orders accelerating; broad-based end-market recovery confirmed; approximately CY26 second-half growth visibility strong.
• Bear case: Inventory normalisation is one-time; the real question is post-restock unit growth, which is industrial/auto cycle dependent.
• Sources: MCHP Q4 FY26 8-K (SEC) · MCHP Q4 (BigGo)

Foundries / IDMs

TSM — TSMC

• Q4 2025 foundry share: 70.4%, projected approximately 75% in 2026.
• Leading-edge (3nm / 2nm) share: above 90%.
• Customer concentration: NVIDIA overtook Apple in 2025 as #1 at approximately 19% of revenue. Apple #2 at 17%. Top 10 customers approximately 76% of revenue.
• 2026 capex: raised to high end of $52–56 billion range (Q1 2026 earnings).
• FY26 revenue: guided above +30% YoY (approximately $158 billion).
• N3 (3nm): end-2026 capacity target approximately 180K wafers/month (+40% YoY). N3 fully booked through 2026.
• N2 (2nm): volume production started late 2025; end-2026 target approximately 100K wafers/month; 2027 target approximately 200K.
• CoWoS: 2026 capacity ramping ~80K → 115–140K wafers/month. Lead times 50+ weeks.
• Forward P/E: approximately 24–28×.
• Sources: TSMC raises 2026 CapEx (Tom's Hardware) · TSMC Q1 2026 6-K (SEC) · Apple no longer biggest TSMC customer (wccftech) · NVDA overtakes Apple (CNBC) · TrendForce TSMC 3nm 180K wafers by 2026 · Silicon Analysts TSMC foundry allocation Q1 2026

INTC — Intel

• 18A external customers confirmed: Amazon, Microsoft, US Department of Defense.
• 18A tested but not committed: NVIDIA, Broadcom.
• Apple status: Has 18A-P PDK and is running simulations (Ming-Chi Kuo).
• 2026 strategy shift: CEO Lip-Bu Tan publicly embraced 18A for external customers; 18A-P getting "inbound interest."
• Forward P/E: approximately 102–117× (some analysts cite 143×) — extreme even for a turnaround.
• Sources: Intel 18A external customers (Tom's Hardware) · NVIDIA tested 18A but did not commit (TechPowerUp) · INTC forward P/E (GuruFocus)

GFS — GlobalFoundries

• Latest quarter: Q1 2026, reported May 2026.
• Revenue: $1.634 billion (+3.1% YoY). EPS $0.40 (beat $0.35).
• Gross margin: 27.6% GAAP / 29.0% non-IFRS.
• Segment mix: Automotive 23% (−11% QoQ but +24% YoY) · Communications infrastructure + datacenter: growing · Smart mobile devices: dropped to approximately one-third of revenue (non-mobile + services = two-thirds).
• Forward P/E: approximately 25–30× (directional).
• EV/Sales: approximately 4×.
• Customer concentration: Qualcomm + Apple historically approximately 30–40% combined (RF, PMIC); now diversified as smart-mobile fell.
• Moat: Mature-node specialty (RF SOI for handset front-end, FDX 22nm) — defensible niches where TSMC doesn't fight; US/Singapore/Germany capacity advantage under CHIPS Act geopolitics.
• Catalyst: Auto low-double-digit growth FY26; comms infra + datacenter PSU + silicon photonics ramp; CHIPS Act-fuelled US capacity 50% complete.
• Bear case: Not on the AI compute bandwagon — silicon photonics is the only AI-adjacent line; mature-node margins structurally lower than TSM N3/N2.
• Sources: GFS Q1 2026 6-K (SEC) · GFS Q1 (Investing.com)

UMC — United Microelectronics

• Latest quarter: Q1 2026, reported April 2026.
• Revenue: NT$61.04 billion (approximately US$1.85 billion), +5.5% YoY, −1.2% QoQ. EPS NT$1.29.
• Gross margin: 29.2%. Operating margin mid-teens.
• Segment mix (process node): 22/28nm: 34% of revenue (22nm alone hit record 14%). 40nm + 65nm + 90nm: balance of mature-node mix. Wafer shipments +2.7% QoQ; utilisation 79%.
• End-market mix: Computer + comm approximately 50%; consumer + display drivers + MCU heavy in the 22nm uptick.
• Forward P/E: approximately 15–18× (consistent with cycle-trough multiple for foundry ADR).
• EV/Sales: approximately 3×.
• Customer concentration: Top 10 approximately 70%; broad-based.
• Moat: Mature-node cost discipline; 22nm volume ramp at over 50 tape-outs by YE26; H2 2026 price hike telegraphed.
• Catalyst: Q2 wafer shipments +HSD, ASP +LSD (mix); margin approximately 30%; utilisation low-80s. 22nm displaying as growth lever.
• Bear case: Squeezed between SMIC (China subsidised capacity) and TSMC (technology lead). Mature-node profitability will erode as China supply lights up.
• Sources: UMC Q1 2026 6-K (SEC) · UMC Q1 (BigGo)

Memory

MU — Micron

• Q3 FY26 guide: revenue $33.5 billion ± $750 million at approximately 81% gross margin — peak-cycle levels.
• HBM3E: sold out through most of 2026.
• FY26 capex: raised to above $25 billion.
• HBM pricing: SK Hynix + Samsung telegraphed approximately 20% HBM3E price hike for 2026.
• Sources: Micron Q3 $33.5B 81% GM (Seeking Alpha) · Micron Q2 FY26 8-K (SEC) · TrendForce HBM3E 20% price hike 2026

WDC — Western Digital (post-split, HDD-pure)

Following the February 2025 SanDisk separation, WDC is the HDD-only entity. NAND business is now standalone SNDK. The thesis is AI nearline HDDs.

• Latest quarter: Q3 FY26 (CQ1 2026), reported April 2026.
• Revenue: $3.34 billion (+45% YoY).
• Gross margin: 50.2% GAAP / 50.5% non-GAAP — all-time high.
• Free cash flow: $978 million (29% FCF margin).
• Segment mix: Cloud (data center nearline HDD) $3.0 billion (89% of revenue, +48% YoY) · Consumer + Client + Other 11%.
• Capacity metrics: 222 exabytes shipped (+34% YoY); 4.1M ePMR drives at up to 32TB.
• Forward P/E: approximately 27–47× (wide range across sources; consensus settling 30–35×).
• EV/Sales: approximately 3×.
• Customer concentration: Top-3 hyperscalers approximately 50%+ of cloud revenue.
• Moat: HDD duopoly with Seagate post-Toshiba consolidation; AI nearline volumes growing 30%+ on data-lake builds; HAMR roadmap.
• Catalyst: Q4 FY26 revenue +36–44% YoY; GM 51–52%; HAMR transition 2026–27; 40TB drives in qualification.
• Bear case: HDDs are still HDDs — flash continues to encroach on hot-tier; one bad demand quarter from a single hyperscaler can swing the print. Stock up above 100% YTD 2026, valuation crowded.
• Sources: WDC Q3 FY26 8-K (SEC) · WDC Q3 (BigGo) · WDC forward P/E (GuruFocus)

STX — Seagate

• Latest quarter: Q3 FY26 (CQ1 2026), reported April 2026.
• Revenue: $3.11 billion (+44% YoY). Adjusted EPS $4.10 (+116% YoY).
• Non-GAAP gross margin: 47% — all-time high.
• Operating margin: approximately 30% non-GAAP.
• Segment mix: Data center 88% of exabytes, 80% of revenue. Edge + consumer 20%.
• Forward P/E: approximately 14–17× (cheaper than WDC).
• EV/Sales: approximately 3×.
• Customer concentration: Top-3 hyperscalers approximately 50%+.
• Moat: HDD duopoly; HAMR technology lead — Mozaic 3 shipping at all major CSPs; Mozaic 4 began shipping March 2026 and qualified at top-2 CSPs.
• Catalyst: Mozaic 4 ramp Q4 FY26; raised annual revenue growth target to "at least 20%"; visibility on nearline contracts extends through 2027.
• Bear case: Same as WDC — HDD is the wrong technology long-term; HAMR delivers margin once, then commoditises.
• Sources: Seagate Q3 FY26 (Investing.com) · STX TipRanks earnings · Seagate FY26 outlook (EBC)

Networking & optics

ANET — Arista Networks

• Latest quarter: Q1 2026, reported May 2026.
• Revenue: $2.71 billion (+35.1% YoY).
• Non-GAAP gross margin: approximately 65% (typical).
• Operating margin: 47.8% (operating income $1.29 billion).
• Segment mix: Cloud / hyperscaler switching approximately 60%+ — 7800R-series, 7060X, 7280R3 800G · Enterprise campus + WAN approximately 30% · Software + services approximately 10%.
• Forward P/E: approximately 43–48×.
• EV/Sales: approximately 14× FY26e.
• Customer concentration: Microsoft + Meta historically approximately 35–40% combined. Top 5 approximately 55%.
• Moat: EOS network OS — single binary, common code base across all platforms; chosen by hyperscalers as the merchant alternative to NVIDIA Spectrum-X for back-end AI fabrics.
• Catalyst: FY26 revenue growth target raised to +27.7% (approximately $11.5 billion); AI target raised to $3.5 billion; 1.6T production scale planned 2027; XPO optics form factor with 100+ vendor endorsements.
• Bear case: Spectrum-X (NVIDIA proprietary) + emerging custom-silicon scale-up (NVLink Fusion, AMD UALink) compress merchant scale-out share; Meta/Microsoft hardware mix shift could be sudden.
• Sources: Arista Q1 2026 (Seeking Alpha) · Arista Q1 highlights (TipRanks) · ANET forward P/E (GuruFocus)

CIEN — Ciena

• Latest quarter: Q1 FY26 (reported March 2026); Q2 reporting June 4, 2026.
• Revenue (Q1): $1.43 billion record. Q2 consensus $1.50 billion, EPS $1.46 (+247% YoY).
• Non-GAAP gross margin: approximately 43% (typical recent).
• Segment mix: Networking Platforms 80.5% — Waveserver + RLS both grew above 80% YoY · Software + services 19.5%.
• Forward P/E: approximately 30–35× (directional).
• EV/Sales: approximately 3.5×.
• Customer concentration: Verizon + AT&T + Microsoft + Google + Meta cluster approximately 50%.
• Moat: WaveLogic 6e coherent DSP — leadership at 800ZR+ / 1.6T over distance; embedded in DCI corridors hyperscalers depend on.
• Catalyst: Q2 print June 4 2026; DCI + AI training cross-DC connectivity scaling; cloud/CSP capex now visible.
• Bear case: Carrier capex remains lumpy; cloud orders concentrated in a handful of contracts; competition from Cisco/Acacia and Infinera-Nokia consolidation.
• Sources: Ciena Q1 FY26 8-K (SEC) · Ciena Q2 preview (Alphastreet)

COHR — Coherent

• Latest quarter: Q3 FY26 (CQ1 2026), reported May 2026.
• Revenue: $1.81 billion (+21% YoY, +27% pro forma).
• Gross margin: 39.6% non-GAAP / 37.7% GAAP.
• Non-GAAP EPS: $1.41 (+55% YoY).
• Segment mix: Datacenter + Communications $1.4 billion (77%), up from $1.0 billion · Industrial $444 million (23%), declined (auto-laser cyclical).
• Forward P/E: approximately 28–35× (directional from EPS run-rate).
• EV/Sales: approximately 3.5×.
• Customer concentration: NVIDIA disclosed multi-year supply agreement + $2 billion equity investment — now structural anchor customer.
• Moat: Vertically integrated photonics — InP epi, EUV laser source (sole supplier to ASML for High-NA), 800G/1.6T transceivers, optical circuit switches. The breadth is unmatched.
• Catalyst: Q4 FY26 guide $1.91–2.05 billion; NVDA multi-year supply deal scales; OCS volume.
• Bear case: Industrial laser business shrinking; transceiver pricing eventually compresses; Lumentum direct competitor with same NVDA investment.
• Sources: Coherent Q3 FY26 8-K (SEC) · Coherent Q3 highlights (Futurum) · Coherent NVIDIA partnership (Simply Wall St)

LITE — Lumentum

• Latest quarter: Q3 FY26 (CQ1 2026), reported May 2026.
• Revenue: $808.4 million record (+90% YoY).
• Gross margin: approximately 37% non-GAAP (typical).
• Non-GAAP EPS: approximately $2.50 (implied from $225.7 million non-GAAP NI).
• GAAP EPS: $1.50.
• Segment mix: Cloud + AI (transceivers, OCS, laser chips) above 70%, rapid growth · Telecom + industrial balance.
• Forward P/E: approximately 30–40× (post-rerate).
• EV/Sales: approximately 5×.
• Customer concentration: NVIDIA $2 billion equity investment + multi-year supply agreement signed March 2026 — anchor customer. Top 5 historically approximately 60%.
• Moat: Indium-phosphide laser chip integration + datacom transceivers + OCS — single supplier across the optical stack; CPO (co-packaged optics) order for H1 CY27 multi-hundred-million.
• Catalyst: Q4 FY26 guide $960 million – $1.01 billion (+85% YoY); OCS backlog above $400 million; CPO order H1 CY27.
• Bear case: NVIDIA dependency now extreme; transceiver pricing power has expiry; CPO transition could disintermediate transceivers entirely.
• Sources: Lumentum Q3 FY26 8-K (SEC) · Lumentum Q3 (Investing.com)

Equipment

ASML

• Q1 2026 net sales: €8.8 billion at 53.0% gross margin.
• YE 2025 order book: €38.8 billion, of which EUV approximately €7.4 billion.
• 2026 sales guide: €36–40 billion.
• High-NA EUV ASP: approximately $380 million; 10–20 booked, plans 20/year by 2028.
• TSMC High-NA deployment: deferred to 2029 (strategic, not capacity-driven).
• EUV competition: None — Nikon (DUV only) and Canon (nanoimprint niche) publicly walked away from EUV development years ago.
• Sources: ASML Q1 2026 IR Presentation PDF · ASML Q1 2026 press release · ASML 6-K (SEC) · High-NA EUV $380M (TechPowerUp)

AMAT — Applied Materials

• Q1 FY26: revenue record at 49.1% non-GAAP gross margin.
• CY26 semi equipment revenue guide: above +20% (CEO public guidance).
• Forward P/E: approximately 39×.
• Sources: AMAT Q1 FY26 8-K (SEC)

LRCX — Lam Research

• Q3 FY26 (March quarter): $5.84 billion at 49.9% gross margin — record.
• Forward P/E: approximately 40×.
• Drivers: HBM + advanced packaging momentum.
• Sources: LRCX Q3 FY26 8-K (SEC)

KLAC — KLA Corporation

• Process control and metrology — near-duopoly with Onto Innovation.
• Historical gross margin: 60%+.
• Customer concentration: TSMC approximately 38% of revenue.
• Forward P/E: approximately 38×.
• Sources: KLAC Q2 FY26 8-K (SEC)

ONTO — Onto Innovation

• Latest quarter: Q1 2026, reported May 2026.
• Revenue: $291.9 million (+9.5% YoY). EPS $1.42.
• Gross margin: approximately 55% non-GAAP (typical).
• Segment mix (system / product): Advanced packaging metrology (Dragonfly G5, 3Di): above 50% growth in 2026 expected — HBM4 ramp · Specialty device + Lithography: balance.
• Forward P/E: approximately 30–35× (directional).
• EV/Sales: approximately 7×.
• Customer concentration: Top customer (likely SK Hynix or Samsung HBM) signed above $240 million VPA for Dragonfly + 3D bump metrology through 2027. Top 3 approximately 50%.
• Moat: Process-control duopoly with KLAC at the bleeding edge; Dragonfly G5 chosen by HBM manufacturer for HBM4; 2.5D/3D advanced packaging metrology unique IP.
• Catalyst: Raised 2026 revenue outlook to above $1.3 billion; OPM target above 30% by YE; Dragonfly G5 + 3Di shipments ramp Q2.
• Bear case: Capex cycle peak risk; consolidation under KLAC umbrella conceivable; smaller scale = less optionality on China headwinds.
• Sources: Onto Q1 2026 8-K (SEC) · Onto Q1 highlights (Yahoo)

ACMR — ACM Research

Listing / regulation risk flag. ACMR is dual-structured — US-listed parent owns the ACM Shanghai subsidiary, the operating business. Heavy China customer exposure (SMIC, YMTC, CXMT). Subject to potential US-China investment / entity-list overhangs.

• Latest quarter: Q1 2026, reported May 2026.
• Revenue: approximately $240 million Q1 (implied from +34% growth and FY guide $1.08–1.175 billion).
• Gross margin: 46.5% — above the 42–48% long-term mid-point.
• Segment mix: Cleaning (mostly SAPS / single-wafer): expected to "normalise" to approximately 65% of FY mix as SPM ramps · Advanced Packaging (ex-ECP): +62% in Q1 · ECP (Cu plating).
• Forward P/E: approximately 12–18× (China-discount multiple).
• EV/Sales: approximately 2×.
• Customer concentration: Heavy China (above 80% of revenue historical) — SMIC, YMTC, CXMT, Huahong. Expanding outside China (above 20 tools, approximately 10 customers in 5 countries by YE).
• Moat: Wet-process / cleaning IP at sub-20nm and HBM advanced packaging; the China-foundry-customer SPM displacement opportunity.
• Catalyst: FY26 reaffirmed $1.08–1.175 billion (+20–30%); ex-China tool installs ramp; SPM tool revenue mix.
• Bear case: US/China policy risk; reaffirmed (not raised) full-year guide signals back-half loaded; sentiment trades like a China stock, not an equipment stock.
• Sources: ACMR Q1 2026 transcript (Motley Fool) · ACMR Q1 preliminary release (Stocktitan)

EDA / IP

CDNS — Cadence

• Q1 2026: revenue $1.474 billion (+19% YoY).
• Backlog: $8.0 billion record.
• RPO (12-month): $4.0 billion.
• Non-GAAP operating margin: 45%.
• 2026 revenue guide: raised to +17% YoY.

SNPS — Synopsys

• Latest backlog: $11.0 billion at April 30, 2026.
• Trades: slight discount to CDNS on forward multiples.
• Mix: Greater IP-licensing share than CDNS.

ARM — Arm Holdings

• FY26 royalty revenue: $2.61 billion (+21% YoY) — datacenter royalty more than doubled.
• Q4 FY26 royalty revenue: $671 million (+11% YoY).
• FY26 total revenue: $4.92 billion record.
• Licensing revenue: $2.31 billion (+25%).
• Multiple: approximately 38× EV/Revenue, approximately 100× price-to-royalty-revenue.
• Sources: ARM Q4 FY26 6-K (SEC) · ARM FY26 20-F (SEC)

Data center infrastructure

VRT — Vertiv

• Q1 2026 net sales: $2.65 billion (+30% YoY, 23% organic).
• Backlog: $15 billion (approximately 80% YoY growth).
• Book-to-bill: approximately 2.9×.
• Americas revenue: $1.81 billion (~70% of total) growing 44% organic.
• Adjusted operating margin: 20.8% (+430 bps).
• 2026 guide raised: organic sales +29–31%, adjusted EPS $6.30–6.40, net sales $13.5–14.0 billion.
• Forward P/E: approximately 54×.
• Sources: VRT Q1 2026 8-K (SEC) · VRT 2026 guidance (techjacksolutions)

EQIX — Equinix

• Q1 2026: revenue $2.444 billion (+10% YoY). EPS $4.20.
• Adjusted EBITDA margin: 51%.
• Bookings: $378 million record annualised gross bookings.
• Segment mix: Americas approximately 45% · EMEA approximately 35% · APAC approximately 20% · Interconnect + Fabric approximately 15% blended — the high-margin moat layer.
• Forward P/E: approximately 51–61×.
• AFFO multiple: approximately 25–28× forward (REIT-relevant).
• EV/Sales: approximately 10×.
• Customer concentration: Top 10 customers approximately 17% of revenue; no single above 5%. Very diversified.
• Moat: Network neutrality + Equinix Fabric — 110+ network nodes; ecosystem effects compound.
• Catalyst: FY26 guide $10.12–10.22 billion — first $10 billion+ year; 8 of top 10 AI model providers + 4 of top 5 neo-clouds expanding with EQIX; Fabric connections for major deals tripled YoY.
• Bear case: Hyperscaler self-build for above 100MW campuses; cap-rate risk if rates rise; EQIX's interconnect moat doesn't help in pure compute build leases.
• Sources: Equinix Q1 2026 8-K (Stocktitan) · EQIX forward P/E (GuruFocus)

DLR — Digital Realty

• Q1 2026: revenue $1.64 billion (+16% YoY). Core FFO $2.04/sh (+15.3%).
• Net income: $169.1 million, EPS $0.46.
• Bookings: $707 million record (100% share) including largest-ever single MW lease.
• Segment mix: North America approximately 70% · EMEA approximately 22% · APAC approximately 8% · Above 1MW hyperscale leases: bulk of contracted backlog · Below 1MW + interconnect: highest-margin retail layer (21% AI-oriented).
• Forward P/E: approximately 70×+ (REIT thesis is FFO/AFFO based; FFO multiple approximately 24× forward).
• EV/Sales: approximately 13×.
• Customer concentration: Top 20 customers approximately 50%+; Microsoft + Oracle + Meta among publicly disclosed.
• Moat: Global scale + hyperscale-grade campuses pre-leased; PlatformDIGITAL ecosystem; above 300 sites in 50+ metros.
• Catalyst: FY26 revenue guide $6.65–6.75 billion; AI deal pipeline robust; HD Hyperscale joint venture monetisation.
• Bear case: Higher hyperscale exposure than EQIX = more concentration risk; cap-rate sensitive; large new builds dilute FFO/sh near-term.
• Sources: DLR Q1 2026 10-Q (Stocktitan) · DLR Q1 transcript (Motley Fool)

ETN — Eaton

• Q1 2026: revenue $7.45 billion record (+17% YoY, +10% organic). Adjusted EPS $2.81.
• Gross margin: approximately 37% (typical recent).
• Segment mix: Electrical Americas: largest segment, +42% in rolling 12-month orders; AI data center orders +240% · Electrical Global: +13% orders · Aerospace: +13% orders · eMobility (to be spun off Q1 2027).
• Forward P/E: approximately 28–32× (directional).
• EV/Sales: approximately 4.5×.
• Customer concentration: Highly diversified across industrial, utility, data center, aerospace customers; none material.
• Moat: Electrical distribution + power management at scale; AI data center "grid-to-chip" portfolio expanded via Boyd Thermal + Ultra PCS acquisitions; the only credible Schneider/ABB-class US-domiciled name.
• Catalyst: FY26 organic growth raised +200bp to approximately 10% midpoint; EPS guide raised to $13.05–13.50; data center backlog 228 GW (~12 years of 2025 build rates); Mobility spin Q1 2027.
• Bear case: Multiple already prices a long AI cycle; aerospace cycle dependent on commercial OEM; integration risk on $11 billion Q1 acquisitions.
• Sources: Eaton Q1 (BigGo) · Eaton Q1 backlog (Alphastreet)

PWR — Quanta Services

• Q1 2026: revenue $7.9 billion. Adjusted EPS $2.68. Adjusted EBITDA $686 million.
• Backlog: $48.5 billion record (vs $44.0 billion at YE25).
• Gross margin: approximately 16% (construction services norm).
• Segment mix (end market): Electric power infrastructure (transmission + distribution) — largest, fastest-growing · Renewables / utility scale · Gas generation · Communications · Underground.
• Forward P/E: approximately 32–38× (rich; directional).
• EV/Sales: approximately 1.5–2×.
• Customer concentration: Utilities + IPP customers; top 10 approximately 40% of revenue.
• Moat: Largest US specialty contractor for electric utility build-out; storm-response capability; nationwide skilled-labour base; multi-year MSA contracts.
• Catalyst: Raised FY26 revenue $34.7–35.2 billion; EPS $13.55–14.25; AI data center site-prep / substation pipeline scaling.
• Bear case: Construction-services margins limited; project execution risk; ROIC structurally lower than chip names; cyclical to utility capex.
• Sources: Quanta Q1 2026 8-K (SEC) · Quanta Q1 (BigGo)

AI power

VST — Vistra

• Q1 2026: revenue $5.64 billion (+43.4% YoY). Net income $1.03 billion (turnaround from −$268 million Q1 2025).
• Adjusted EBITDA: $1.494 billion (+20% YoY, +85% vs Q1 2024).
• Segment mix (generation type): Nuclear (Comanche Peak + after Energy Harbor) — largest baseload contribution · Natural gas (now approximately 25GW post-Cogentrix close) — bulk of capacity · Renewables + storage — growing · Retail (TXU Energy brand).
• Forward P/E: approximately 18×.
• EV/Sales: approximately 2.5×.
• Customer concentration: Wholesale + retail mix; new long-term PPAs with Meta (approximately 2,600MW at PJM nuclear) and AWS — anchor hyperscaler exposure.
• Moat: Largest US competitive power generator post-Cogentrix; deregulated market positioning (ERCOT + PJM) lets PPAs price in scarcity premium.
• Catalyst: Cogentrix close (5,500MW gas); Meta + AWS PPAs guided to add to base case; FY26 EBITDA $6.8–7.6 billion reaffirmed (excludes new deals).
• Bear case: Power demand case "not yet priced in" per bulls but stock has already moved sharply; capacity-market pricing peaks; Q4 2025 deal heat starts to fade if PPA signings pause.
• Sources: Vistra Q1 2026 transcript (Motley Fool) · Vistra Q1 (TIKR) · VST forward P/E (GuruFocus)

CEG — Constellation Energy

• Q1 2026: revenue $11.12 billion (+64% YoY). Adjusted EPS $2.74 (vs $2.14 PY).
• GAAP EPS: $4.49 (vs $0.38 PY).
• Segment mix (post-Calpine): Nuclear — largest US fleet (~22GW pre-Calpine) · Gas + dual-fuel — Calpine's 23GW added in 2026 — TX, CA, PJM exposure · Renewables — smaller.
• Forward P/E: approximately 24–25×.
• EV/Sales: approximately 3×.
• Customer concentration: Long-term PPAs with Microsoft (Three Mile Island Unit 1 restart 2028), Meta, and others — above 5,650MW of long-term clean energy hyperscaler contracts.
• Moat: Largest US nuclear fleet; only credible 24/7 zero-carbon baseload at GW scale that can be signed by hyperscalers; restart optionality (Crane / TMI).
• Catalyst: 20% EPS growth target FY26; Calpine integration synergies; Hyperscaler PPA pipeline continues; restart timeline visibility.
• Bear case: Nuclear capex inflation; PPA contracted volumes already material — incremental signings get harder; Calpine integration brings gas/dispatch risk into the "clean" thesis.
• Sources: CEG Q1 2026 8-K (SEC) · CEG forward P/E (GuruFocus)

GEV — GE Vernova

• Q1 2026: revenue $9.3 billion (+16%, +7% organic). Net income $4.7 billion (50.9% margin).
• Orders: $18.3 billion (+71% organic). Backlog $163 billion.
• Segment mix: Power (Gas / Steam / Nuclear / Hydro): largest; Gas Power slot reservations 100GW (target 110GW by YE26) · Electrification (grid + power conversion): data center equipment orders $2.4 billion in Q1 alone — more than all of 2025 · Wind (offshore + onshore): legacy drag, improving.
• Forward P/E: approximately 58–69×.
• EV/Sales: approximately 3.5×.
• Customer concentration: Diversified utility + IPP + hyperscaler customers; no single above 10%.
• Moat: HA-class gas turbine — 64%+ combined-cycle efficiency, only credible US alternative to Siemens Energy for above-GW campus power; "$200 billion backlog by 2027" — a year ahead of prior plan.
• Catalyst: FY26 revenue $44.5–45.5 billion; FCF $6.5–7.5 billion; gas slot reservations at 110GW YE26; data center equipment orders compounding.
• Bear case: Multiple is the most extreme in the AI power triumvirate; wind drag is real if offshore continues to slip; gas turbine supply tightness is a 2027-28 story — until then revenue is order-recognised vs delivery-recognised.
• Sources: GEV Q1 2026 8-K (SEC) · GEV forward P/E (GuruFocus)

---

Part 2 — Macro infrastructure

Hyperscaler 2026 / 2027 capex guides

Per-company FY2026 capex guides (latest disclosed, as of June 2026):

Company	FY2026 guide	YoY growth	Notes
Microsoft	approximately $190B	approximately +130%	Reported during F1Q26 (CY Q3 2025)
Alphabet / Google	$180–190B	above +100%	Raised twice in 2025; Q1 2026 reaffirmed
Meta	$125–145B	approximately +85%	Raised Q1 2026 from $100–115B initial
Amazon (AWS)	above $230B	above +50%	Largest single-company AI infra commit
Oracle	approximately $50B	approximately +100%	Stargate JV + own data center build
Big 5 total	$660–690B	approximately +75%	75% (~$450B) AI-specific
Top 9 global (adds ByteDance, Alibaba, Tencent, Baidu)	approximately $830B	—	Per Evertiq aggregation

2027 guides — what's been said publicly:

• Most hyperscalers have not given hard 2027 guides; commentary points to "continued meaningful growth from 2026 base."
• Consensus aggregates put 2027 at $1T+ for the Big 5 combined.
• Microsoft has flagged "supply-constrained" Azure backlog of approximately $80B as the demand signal for continued FY27 capex growth.
• Meta in Q1 2026 noted "expect capital expenditures to continue to grow meaningfully in 2027" — directional only.

Sources: Evertiq hyperscaler $830B 2026 · Futurum $690B 2026 · Yahoo $700B 2026 plans · CNBC Big Tech capex $1T 2027 · CreditSights hyperscaler capex 2026

Hyperscaler capex by component

Cleanest segmentation from Morgan Stanley / Bernstein / Bank of America research published Q4 2025 / Q1 2026:

Component	% of AI capex	$ in 2026 (on ~$450B AI portion)	Major beneficiaries
GPU / merchant accelerator	approximately 35–40%	~$160–180B	NVIDIA (dominant), AMD MI300/MI400
Custom ASIC / silicon co-design	approximately 10–15%	~$45–65B	Broadcom (Google TPU, Meta MTIA), Marvell, Alchip/GUC
Networking (switch, optical, NIC)	approximately 10–12%	~$45–55B	Arista, Cisco, Broadcom Tomahawk/Jericho, Coherent, Lumentum, Astera, Credo, Marvell DSP
HBM + memory	approximately 12–15%	~$55–65B	SK Hynix (above 50%), Micron, Samsung
Storage (HDD + SSD)	approximately 3–5%	~$15–22B	Seagate, Western Digital, Samsung NAND, SanDisk
Power + cooling + electrical	approximately 12–15%	~$55–65B	Vertiv, Eaton, Schneider, Carrier, Trane, ABB, MPWR
Data center shell + land + general construction	approximately 10–12%	~$45–55B	Quanta, EQIX/DLR (lease), Mortenson, Turner

Notes:

• Hyperscalers' compute side (GPU + ASIC) = approximately 50% of AI capex combined; networking is the next-largest single category.
• Custom ASIC share is growing fastest — Amazon Trainium 2 has reached $10B+ run-rate per Introl/secondary research.
• Memory + storage combined approximately 17%.
• Power + cooling now material — Eaton's Q1 2026 data center order backlog at 228GW (approximately 12 years at 2025 build rates) is the cleanest single data point.

Sources: Custom Silicon Inflection 2026 (Introl) · AI Data Center Value Chain (Silicon Analysts) · Custom ASIC Market 2026 (Oplexa) · Big Tech $650B Capex Surge 2026 (Tech Insider)

HBM supply 2026 / 2027

Demand growth:

• HBM demand: +77% YoY in 2026, +68% in 2027 per TrendForce.
• 2026 mix: approximately 55% HBM4, approximately 45% HBM3E.
• 2027 mix: HBM4E projected to capture approximately 40% of total HBM.

Supply position:

• SK Hynix: 2026 HBM supply plan finalised with major clients; "slight shortage through at least 2027."
• Micron: 2026 HBM capacity sold out; pricing agreements already locked. Internally able to meet only approximately 55–60% of core-customer demand.
• Samsung: closing the gap via HBM3E + HBM4; HBM4E targeted to complete development by H1 2026.

Supply tightness extends to 2027: All three suppliers state the supply-demand convergence happens earliest H2 2027; multi-year supply agreements with hyperscalers are reducing spot-market exposure.

Pricing: Samsung + SK Hynix telegraphed approximately 20% HBM3E price hike for 2026.

Sources: TrendForce HBM4E 40% 2027 · TrendForce HBM3E 20% price hike 2026

Stargate + sovereign AI committed capex programs

Stargate (OpenAI–Oracle–SoftBank):

• Announced January 2025 as $100B initial; up to $500B by 2029.
• January 2025 commitment: 10GW of AI infrastructure in the US by 2029.
• September 2025: 5 new sites announced, total approximately 7GW planned, above $400B investment over 3 years.
• April 2026 development: OpenAI pulled back from Stargate Norway (230MW Narvik site); Microsoft is taking the compute.
• February 2026 revised commitment: OpenAI told investors it would spend approximately $600B on infrastructure by 2030 — down from $1.4T long-range commitments previously outlined.

UAE — G42 / MGX / Mubadala:

• G42 received $1.5B from Microsoft April 2024.
• OpenAI strategic partnership in place.
• Cerebras anchor customer / investor positioning.

Saudi Arabia — HUMAIN (PIF):

• HUMAIN announced 500MW AMD gear (~$10B) + 500MW NVIDIA systems over 5 years.
• Beginning 2026 with 18,000-unit GB300 order from NVIDIA.
• Plans 1.9GW data center capacity by 2030, scaling to 6.6GW by 2034.

India sovereign AI:

• IndiaAI Mission approximately $1.25B (Government of India, March 2024).
• Reliance Jio + Microsoft data center JV.
• Order of magnitude smaller than US and Gulf programs.

Sources: OpenAI Stargate Project announcement · OpenAI 5 new Stargate sites · CNBC OpenAI Stargate Norway pullback · HUMAIN AMD + NVIDIA 1GW (Introl)

US data center electricity demand

Current baseline (2025): US data centers approximately 4.5% of total US electricity consumption.

EPRI 2026 base-case projection: 9–17% of US electricity by 2030. New estimates approximately 60% higher than EPRI's 2024 projections.

Lawrence Berkeley National Lab DOE study: 325–580 TWh/yr by 2030 = 6.7–12% of US electricity.

Regional impact:

• Virginia: 41–59% of state electricity by 2030 (vs. 25% today).
• Texas (ERCOT): doubling expected.
• Phoenix metro: similar regional stress.

Power demand model behind capex thesis: AI data center power demand projected at approximately 156GW by 2030 globally per Morgan Stanley / Bernstein lineage. US share approximately 40–50% of that. Implies approximately $5.2T cumulative data center investment globally through 2030.

Microsoft signal: Q1 2026 Microsoft disclosed $80B+ Azure backlog that cannot be fulfilled due to power constraints — power is the binding constraint, not chips.

Sources: EPRI Powering Intelligence 2026 · DOE data center electricity report

Semiconductor industry total revenue

• 2025 actual: $791.7B, +25.6% YoY (vs $630.5B in 2024). Per SIA.
• 2026 forecasts:
  • WSTS: $975B (+26% YoY).
  • SIA: approximately $1T.
  • IDC (most aggressive): $1.29T (+52.8% YoY).
• Key driver in all forecasts: AI infrastructure capex; memory pricing inflection layered on top.

Sources: SIA 2025 global semi sales $791.7B · Tom's Hardware $1T 2026 SIA · WSTS forecast PDF

Lead times — current pipeline

ASML EUV (low-NA):

• Backlog at YE 2025: €38.8B, of which EUV approximately €7.4B.
• Lead time: practical 12–18 months for new EUV system orders.
• 2026 sales guide: €36–40B.

ASML High-NA EUV:

• TSMC deferring High-NA deployment until 2029.
• Intel + Samsung are the early adopters; commercial production volumes minimal until 2027-28.

TSMC N2 / N3 capacity: N3 fully booked through 2026; N7 has availability. N2 ramping but constrained by pilot-to-volume timeline.

CoWoS advanced packaging:

• Lead times 50+ weeks (the tightest single bottleneck).
• 2026 capacity ramping ~80K → 115–140K wafers/month.

HBM: 2026 capacity sold out at all three suppliers. 2027 supply tight.

Mature-node foundry (28nm and above):

• Normal lead times (~16 weeks).
• UMC + GFS + Vanguard + SMIC all running 75–85% utilisation Q1 2026.

---

Part 3 — Risk frameworks

Taiwan crisis wargaming

CSIS — primary published frameworks:

(a) "The First Battle of the Next War: Wargaming a Chinese Invasion of Taiwan" (2023).

• 24 wargames run.
• Finding: Taiwan endures as a democratic entity in most scenarios — but costs are catastrophic. US loses 2 carriers, approximately 700 aircraft. Taiwan economy "shattered." Chinese navy "in ruins."

(b) "Lights Out? Wargaming a Chinese Blockade of Taiwan" (2025).

• 26 wargames.
• Critical finding: in most free-play games, despite reluctance to escalate, teams respond to perceived escalations with minor escalations of their own. Two games spiralled to general war.

Semiconductor-specific impact framework:

• Taiwan produces approximately 90% of world's leading-edge semiconductors (3nm and below approximately 100%).
• In a sustained blockade or war scenario: TSMC offline → global chip supply contracts approximately 30% → SMIC + Intel + Samsung absorb approximately 12-18 months of catch-up time.
• For AI-specific: full disruption is multi-year reset because EUV-fabbed leading-edge cannot be substituted at scale.

Probability framing (most-cited 2026 ranges):

• Brookings / Council on Foreign Relations / CSIS aggregated commentary: approximately 10–25% probability of "major Taiwan event" (invasion/blockade) by 2030 is the modal analyst range.
• Higher-end analysts (some RAND, Heritage) cite 30%+. Lower-end (most major investment-bank political-risk teams) cite 5–10%.

Sources: CSIS First Battle of the Next War wargame · CSIS Lights Out Wargaming Blockade

China semiconductor export controls — current state

December 2024 — the "BIS package":

• Added controls on high-bandwidth memory (HBM) with memory bandwidth density above 2 GB/s/mm².
• Controls on 24 types of semiconductor manufacturing equipment + 3 types of software.
• 140 Entity List additions + 14 modifications spanning PRC tool manufacturers, fabs, investment companies.
• HBM controls extend to US-origin AND foreign-produced HBM subject to FDPR (Foreign Direct Product Rule).

January 2026 — BIS final rule (effective 15 Jan 2026):

• Licensing review policy for advanced computing semiconductors to China + Macau shifted from presumption-of-denial to case-by-case with strict conditions.
• A company is an authorised IC designer for only 180 days if it meets criteria — closing one of the most-used loopholes.

April 2026 update:

• After 13 April 2026, the 180-day authorised-IC-designer designation became operational — effectively forcing recertification of every China-customer designer/integrator.

What's likely next (Q3/Q4 2026 expected actions):

• Closing the Malaysia / Singapore / UAE re-export gateways.
• Lower H20 / H200-equivalent specs ceiling.
• China semis equipment shipment to China subject to deeper FDPR.

Sources: Congress.gov R48642 US Export Controls and China Semis · BIS Revised Semi License Review Policy Jan 2026 · Morgan Lewis BIS Revises AI Chip Policy

AI capex slowdown — leading indicators

Six signals organised from earliest to latest:

(a) ASML net bookings (12–18 months ahead). ASML EUV system order book represents capex decisions taken 12–18 months before they show up in fab spend. A sustained two-quarter decline in EUV bookings = earliest warning. 2026 status: YE 2025 backlog €38.8B, EUV €7.4B — no slowdown signal.

(b) TSMC monthly revenue / leading-edge ramp commentary. Sequential MoM declines at TSMC for advanced nodes = hyperscaler order intake softening. 2026 status: No slowdown — N3 fully booked, N2 ramp on schedule.

(c) Hyperscaler quarterly language shift. "Compute constrained" → "balanced supply and demand" or "ROI timeline elongating" = inflection language. 2026 status: All Big-4 hyperscalers still saying "supply constrained" (Microsoft Azure $80B backlog the clearest example).

(d) Hyperscaler capex intensity (capex / revenue). Capex above 25% of revenue = AI conviction. Below 18% = caution. 2026 status: Big-4 capex intensity 25–35% — peak conviction.

(e) NVDA / AMD data center revenue QoQ deceleration. Two consecutive QoQ declines = inflection. 2026 status: NVDA Data Center Q1 FY27 +92% YoY; AMD Data Center +57% YoY — no deceleration.

(f) ROI per AI revenue dollar. Cumulative AI revenue (OpenAI, Microsoft AI services, Google Vertex, Anthropic) vs cumulative AI capex. Jefferies: 2025 ratio approximately 10 cents of AI revenue per dollar of capex. Needs to improve through 2027.

Memory cycle reversal — historical triggers

Historical pattern (DRAM): 1995–96 PC oversupply · 2000–01 dot-com + Y2K hangover · 2008–09 GFC · 2018–19 hyperscaler digestion + crypto-mining collapse + smartphone slowdown · 2022–23 post-COVID inventory build + smartphone weak + PC weak.

Specific signals that flip up-cycle to down:

1. DRAM inventory weeks above 8 at top-3 suppliers (above approximately 6 normal). June 2026 status: 3.3 weeks — extreme tightness.
2. Capex/revenue at suppliers above 50% for two consecutive years → capacity overhang. Currently approximately 45–50% at SK Hynix; rising but not yet 2017-vintage levels.
3. HBM contract pricing rolling over. Currently HBM3E lifted 20% YoY for 2026 — opposite direction.
4. Hyperscaler memory order push-out commentary (none signalled in Q1 2026 prints).
5. PC + smartphone unit growth +5% YoY combined. Both currently flat-to-down.
6. Smartphone OEM cancellation of orders (the canonical 2018 trigger).

The 2026 cycle is structurally different per most analysts:

• Multi-year supply agreements signed for 2026/27 HBM = lower spot-market exposure → less prone to sharp reversal than 2017-vintage cycle.
• Trough inventory + peak YoY pricing + stock-price inflection — three preconditions for cycle peak — none are met as of Q1 2026.

Multiple compression history — when semis have had ~50% drawdowns

Period	SMH drawdown	SOXX drawdown	Recovery	Trigger
2000–02 dot-com	-85%	-70%	Multi-year+	Telecom + PC capex unwind; channel inventory implosion
2008–09 GFC	—	-52%	~6 years	Liquidity crisis + global demand stop
2018–19 trade war	—	-30%	~12 months	Smartphone slowdown + crypto mining collapse + early China export controls
2022 (Jan–Oct)	-33.5%	-35.1%	~14 months	Rates + memory digestion + China lockdown + post-pandemic PC unwind
2024 mid-cycle wobble	-20%	-20%	Months	AI-bubble fears + Yen-carry trade unwind

Pattern framework:

• The semi index falls harder and faster than S&P 500 in every downturn.
• Multiple compression typically does approximately half the drawdown work; earnings cut does the other half.
• Recovery time tracks recession severity more than semi-specific factors.

Single-name 2022 drawdowns (for context): NVDA −65% peak-to-trough · AMD −60% · MU −50%.

Sources: Quantflowlab Semiconductor Downturn History · Quantflowlab Semiconductor Cycle History

---

Part 4 — Portfolio construction

SOXX / SMH actual holdings and weights (June 2026)

SMH (VanEck Semiconductor ETF): 25 holdings, pure market-cap weighted within sector.

Rank	Ticker	Weight	Notes
1	NVDA	approximately 19%	Concentration that drove 2023 alpha vs SOXX
2	TSM	approximately 10%	Foundry anchor
3	AVGO	approximately 6.7%	Networking + custom ASIC
4	ASML	approximately 5%	Litho
5	AMD	approximately 5%
6	QCOM	approximately 4–5%
7	TXN	approximately 4–5%
8	MU	approximately 4%	Memory
9	LRCX	approximately 4%	Equipment
10	KLAC	approximately 3–4%	Equipment
Top 10		approximately 65–73%	Heavy

SOXX (iShares Semiconductor ETF): 30 holdings, modified market-cap weighted with single-name caps.

Rank	Ticker	Weight
1	MU	approximately 9.8%
2	AMD	approximately 9.2%
3	INTC	approximately 7.4%
4	AVGO	approximately 7.3%
5	NVDA	approximately 6.6%
6	TXN	approximately 5%
7	QCOM	approximately 5%
8	ADI	approximately 5%
9	MRVL	approximately 4%
10	KLAC	approximately 4%
Top 10		approximately 60%

Structural takeaway:

• SMH delivers concentration to NVDA + TSM + ASML — better post-2023 hindsight, more risk going forward.
• SOXX delivers diversification (no name above 10%), more US-centric, more equity in cycle laggards (INTC, MCHP).

Sources: SMH vs SOXX etf.com · SMH vs SOXX (HeyGoTrade)

Historical drawdown stats — SOXX / SMH

Max drawdown from inception: SMH −85% (peak-to-trough, dot-com era) · SOXX −70% (peak-to-trough).

Recent named drawdowns:

Period	SMH	SOXX	Time to recovery
2022 (full year)	-33.5%	-35.1%	~14 months
2018 (full year)	~-13%	~-16%	~12 months
2020 COVID flash crash	~-25%	~-27%	~5 months

Pattern: ETFs draw down 30–35% in "normal" semi recessions; single names draw down 50–70% in same recession. Recovery for ETFs typically 12–14 months for sub-recession declines; multi-year for deep recession.

Position sizing — institutional sector-concentration frameworks

Common institutional limits applied to thematic / sector ETFs in retail-adjacent portfolios:

Vanguard / Fidelity model portfolios — sector tilt guidance:

• Single GICS sector above 25% of equity sleeve = "concentrated."
• Most balanced model portfolios cap any single sector at 15–20% of equity sleeve.

CFA / FRM Level III portfolio construction:

• Sector tilt limit ±3% as standard institutional constraint.
• Single security above 2× index weight or above 5% absolute = "concentration" flag.

Plain-English heuristic widely used in advisor circles:

• Single sector ETF ≤15% of equity portfolio (i.e., SOXX/SMH ≤15%).
• Single name ≤5% of equity portfolio (i.e., NVDA alone ≤5%).
• Sector exposure across both ETF + single names ≤20% combined.

For a retail US-investing Indian portfolio (LRS-funded):

• Semis-only exposure typically capped at 15–25% of the equity sleeve.
• Within semis: top single name capped at 30–40% of the sleeve (so NVDA can be a third of the semis bucket but not more).

For an Indian resident — LRS / TCS / tax-efficient holding

LRS limit (FY 2026-27):

• USD 250,000 per individual per financial year.
• Per individual — not per family. Household of 4 adults = USD 1M aggregate capacity.
• Per financial year (April–March) — no carry forward.

TCS rate matrix on LRS (in force since 1 Oct 2023):

Purpose	Threshold per FY	Rate above threshold
Overseas investment (shares, ETFs, funds)	Rs 10,00,000	20%
Gifts / maintenance of relatives	Rs 10 lakh	20%
Other purposes	Rs 10 lakh	20%
Education by Section-80E loan	Rs 7 lakh	0.5%
Education other	Rs 7 lakh	5%
Medical	Rs 7 lakh	5%
Overseas tour package	Rs 7 lakh	5% then 20% above

TCS is a prepayment, not a tax cost. Credit recoverable on income-tax return.

Form 12BAA (Finance Act 2024, effective 1 Oct 2024): Salaried employees can declare TCS to employer; employer reduces monthly TDS proportionately.

Schedule FA — annual disclosure of foreign asset holdings:

• Required for every foreign asset held even a single day in the FY.
• Black Money Act, 2015 penalties for non-disclosure: 30% tax + 3× penalty + 3-10 year prosecution.

Capital gains treatment for an Indian resident on US-listed equity:

• Treated as unlisted for Indian tax purposes (foreign-listed).
• Long-term holding period: 24 months (vs 12 for domestic listed equity).
• LTCG rate post Budget 2024: 12.5% (down from 20% with indexation).
• STCG: slab rate.

Sample portfolio sketches — institutional precedents

Three sleeve archetypes, with weight ranges grounded in real institutional / advisor practice:

(a) Defensive (capital preservation + AI exposure with structural moats only):

Framework precedent: Long-only mutual funds with "quality at reasonable price" mandates. Concentration in monopolists / duopolists with proven cash conversion.

Typical sleeve composition guidance:

• 4–6 names total (max).
• Top weight 30–35% (in the single highest-quality name).
• No name below 10%.
• All names must have above 15-year track record of margin discipline OR above 50% market share in a moat layer.

(b) Balanced (mix of core + opportunistic):

Framework precedent: Multi-cap growth-and-income mutual funds. 60% core blue-chip + 25% mid-cap opportunistic + 15% thematic / late-cycle.

Typical sleeve composition guidance:

• 8–12 names total.
• Largest position 15–20%.
• Smallest position 3–5%.

(c) Momentum / aggressive (highest leverage to continued AI buildout):

Framework precedent: Concentrated growth hedge funds. Heavy single-thesis allocation + willingness to draw down 50% to capture multi-bagger upside.

Typical sleeve composition guidance:

• 5–8 names total.
• Largest position 25–35%.
• All names must have triple-digit YoY revenue growth or be the prime beneficiary of a multi-year AI capex line.
• Acknowledgement that 30–60% drawdowns are the cost of admission.

The three-portfolio matrix:

Sleeve	Defensive	Balanced	Momentum
Total names	4–6	8–12	5–8
Largest single name	30–35%	15–20%	25–35%
Cyclical / equipment exposure	15–20%	15–25%	0–10%
Pure-thematic AI (ALAB / CRDO / VRT class)	0%	10–15%	35–50%
Power / non-semi AI infra	0–15%	10–20%	10–25%
Cash buffer	10–20%	5–10%	0–5%
Expected drawdown in semis recession	-25 to -35%	-35 to -45%	-50 to -65%

Sources: Carta portfolio construction · AnalystPrep CFA L3 portfolio construction notes · Quantflowlab semi cycle drawdown framework · Fidelity sector-based portfolio framework