NVIDIA
The Capstone, Operating Acceleration vs Multiple Compression

Snapshot: Executive Summary

NVIDIA enters mid-2026 as the largest publicly traded company on the planet at ~$4.9T of market value, having printed the highest-quality quarter in modern semiconductor history three weeks earlier: Q1 FY27 revenue of $81.6B (+85% YoY), Data Center of $75.2B (+92%), networking of $14.8B (+199%), non-GAAP gross margin of 75.0%, and a Q2 guide of $91B that explicitly assumes zero China Data Center compute. Diluted GAAP EPS of $2.40 more than tripled YoY. Management raised the dividend twenty-five-fold to $0.25/quarter, authorized a fresh $80B buyback, and returned a record $20B in the quarter. Supply commitments now stand at ~$145B: visibility that effectively forward-books FY28¹.

Against that operational backdrop, the equity has done something remarkable: NTM P/E has compressed to ~22-23x: the 11th percentile of the ten-year valuation range and the cheapest forward multiple in the MAG7, while consensus FY28 revenue has migrated from ~$290B a year ago to ~$485B today. The setup is unusually clean: this is the moment where the operating story is accelerating and the multiple is decelerating in opposite directions².

We rate NVDA Buy with a 12-month price target of $215 and a blended fair value of $207, reflecting a balanced 25/50/25 bull/base/bear weighting that respects the legitimate custom-silicon ceiling debate without conceding the bundling and cost-per-token moat that has now been demonstrated across compute, networking, and inference software. Library distribution becomes 6 Buy / 3 Hold / 1 Balanced / 1 Sell across 11 dashboards, NVDA is the capstone every other AI silicon name (AVGO, MRVL, MU, SNDK, SMCI, MBLY) reads through to.

Tactical: NVDA at $200 sits 7.5% below our $215 12-month target with the upside biased toward the Q2 FY27 print in August 2026 and the Vera Rubin first commercial shipments in Q3 FY27 (Aug-Oct 2026). The 25% bear weight respects custom-XPU acceleration; the $80B buyback authorization provides a structural bid against any Rubin-transition air pocket. Rating Buy.

Investment Thesis

Rating: Buy. Probability weights: 25% Bull / 50% Base / 25% Bear → blended fair value ~$207. PT set modestly above blended FV at $215 to reflect (a) FY27 consensus ~$391B is set to beat given H1 visibility of ~$172.6B already locked, (b) the $80B buyback plus 25x dividend hike provide downside support, and (c) NTM P/E at the 11th percentile of the ten-year range is asymmetric on the multiple itself. The bull weight of 25% (rather than the rubin-ramp lens's 35%) and the bear weight of 25% (rather than the recovery lens's 15%) reflect the discipline that the valuation-cap lens introduces, the bear case is genuinely defensible at ~$5T market cap and peak gross margin.

Business Overview

NVIDIA reorganized its reporting structure beginning Q1 FY27, collapsing the prior five-segment market-platform breakdown (Data Center, Gaming, Pro Viz, Auto, OEM) into two market platforms, Data Center and Edge Computing: while continuing to report Compute & Networking and Graphics at the GAAP segment level. Q1 FY27 Compute & Networking was $74.55B (+88% YoY); Graphics was $7.07B (+58%)¹. Edge Computing now bundles what used to be Gaming, Pro Viz, Auto, and the emerging robotics/AI-RAN/agentic-device categories, a recognition that the device-side of NVIDIA's TAM (DRIVE Hyperion, Isaac GR00T, Cosmos, IGX Thor) is being treated as a single platform attached to CUDA-X.

Revenue history: the supercycle V

Segment mix: Q1 FY27 two-platform view

Customer mix: the new dual-bucket disclosure

Inside Data Center, management for the first time disclosed customer mix in two buckets: Hyperscale at ~$37.9B (~50% of DC, +115% YoY) and ACIE, AI Clouds, Industrial, Enterprise, Sovereign, at ~$37.4B (+74%)¹. The top four hyperscalers account for ~40% of DC. The ACIE bucket includes sovereign AI deployments active in 40 countries (G42 UAE, HUMAIN/Saudi PIF, India), dedicated AI cloud customers (CoreWeave, Lambda, Crusoe), and the rapidly-growing enterprise install base. ACIE growth at +74% is the diversification story; hyperscale at +115% is the supercycle story.

Geographic mix: the China collapse and US concentration

CUDA-X software moat: the durable franchise asset

The CUDA software moat, now reinforced by Dynamo 1.0 as the production inference operating system (with TensorRT-LLM upstreaming into vLLM/SGLang/llm-d), remains the most durable franchise asset and is the reason the bundle thesis is genuinely defensible. Platform breadth flagged in Q1 commentary spans: Omniverse (NuRec released, partnerships with global industrial software leaders); DRIVE (Hyperion partnerships with Hyundai/Kia, Uber autonomous fleet, BYD/Geely/Isuzu/Nissan building L4-ready vehicles); Isaac (Cosmos and GR00T N models, NVIDIA IGX Thor general availability); BioNeMo (multiple life-sciences leaders, Nemotron Coalition launched).

Platform Cadence: Hopper to Rubin and the Per-Token Economics

GTC 2026 in San Jose was the pivot from "GPU vendor" to "AI-factory vendor." Jensen unveiled the full Vera Rubin platform, Rubin GPU (336B transistors, 288GB HBM4, ~50 PFLOPS FP4 per chip), Vera CPU (~$200B TAM, ~$20B standalone revenue framing), and NVLink 6 at 3.6 TB/s per-GPU bidirectional. The standard Vera Rubin NVL72 rack pushes 260 TB/s aggregate scale-up bandwidth; the Rubin Ultra NVL576 "Kyber" configuration scheduled for H2 2027 delivers 15 exaflops FP4 across 8 racks of 72 GPUs, with 1TB HBM4e per package and ~600kW per rack³.

Annual platform cadence: and what each generation claims

The cost-per-token pivot

The narrative shift that matters is the explicit customer KPI change to cost-per-token. Blackwell B200 reached ~$0.02 per million tokens on gpt-oss inference per NVIDIA's InferenceMAX benchmark; Rubin's claimed 10x further reduction targets ~$0.002-$0.005 per million tokens for MoE inference at NVL72 scale³. Dynamo 1.0 delivers up to 7x inference improvement on Blackwell and is being co-opted into the open inference stack rather than fighting it. The strategic implication is that each annual platform step structurally devalues installed-base GPUs faster than depreciation schedules, a feature for NVIDIA (forced refresh) and a risk for neocloud financing structures.

HBM4 supply: the gating constraint, now resolved

HBM4 capacity is the single biggest unit-economics lever for Rubin. As of June 5, 2026, all three suppliers are qualified per Jensen's Korea trip statement: SK Hynix leads with ~60-70% allocation, Samsung is in mass-production ramp at ~25-30%, Micron qualification is close to ramp³. Three-supplier qualification creates pricing leverage and removes single-source risk, critical for the FY28 unit-economics modeling.

Hopper rental pricing: the bear case that never arrived

Hopper rental pricing actually rose 15-20% through early 2026 as media generation and multi-agent workflows soaked up marginal capacity, deferring the "price decline" narrative the bear camp had expected. The H100 SXM5 8-GPU node spot price hit $33/hour in February 2026, well above the $25/hour low set in mid-2025. The implication: even on the older platform tier, demand is sopping up supply faster than the bear case modeled.

Six-Front Defense

NVIDIA faces six simultaneous competitive fronts and is winning, drawing, or losing each one with measurable evidence. The synthesis isn't whether NVDA has competition, it does, on every front, but whether the bundling response is sufficient to keep dollars expanding even as share drifts.

Front 1: China H20/H200 license saga

Status: Draw (contained). Zero Data Center compute revenue from China assumed in the Q2 FY27 guide; the Q1 FY26 $4.5B H20 inventory charge has not been followed by a comparable second charge; the 15% (H20) and 25% (H200) revenue-share remitted to the US Treasury caps the upside even if licenses flow⁵. Through early 2026 H200 China bookings remain effectively zero as Beijing steers Alibaba/Tencent/ByteDance/Baidu toward Huawei Ascend 910C and Cambricon. The Trump-era "AI OVERWATCH Act" remains the regulatory tail risk. Upside case: bilateral deal-driven framework continues to flex with administration priorities; H200 pipeline genuinely re-opens contributing $15-20B FY28.

Front 2: Hyperscaler custom silicon (TPU / Trainium / MTIA / Maia)

Status: Loss (structural share shift), offset by dollars. Every hyperscaler now has a production program: TPU v7 Ironwood at ~4M units 2026 with v8 split into Sunfish (training) / Zebrafish (inference) for late 2027; Trainium 2/3 supporting Anthropic at 5GW (Project Rainier); MTIA on Broadcom multi-GW through 2029; Maia 200 live since January 2026⁴. Custom ASIC unit growth at +44.6% YoY outpaces merchant GPU at +16.1% (TrendForce); ASIC unit shipments cross merchant GPU around 2028. NVDA's defense is networking attach plus CUDA inertia for frontier training. Meta's February 2026 multi-billion NVDA commitment demonstrated that bundling extracts NVDA dollars even when the compute socket goes custom.

Front 3: AMD MI355X/MI400/MI450

Status: Draw (limited share, real second source). MI355X shipping at limited CSPs through May 2026; OpenAI 6GW strategic agreement begins 1GW MI450 in H2 2026⁴. AMD's Data Center revenue could grow 60-70% in FY27, meaningful but not share-defining given CUDA and Dynamo lock-in. The optimistic interpretation: AMD takes share from custom-XPU plans (Maia/MTIA) rather than NVDA, leaving NVDA's merchant slot intact. The realistic interpretation: AMD captures 5-7% incremental at the margin, mostly at the cost-sensitive tier where NVDA's premium positioning was always vulnerable.

Front 4: Spectrum-X vs Arista in AI networking

Status: Win (decisively). NVDA networking at $14.8B/qtr is roughly 4.2x Arista's full-year $3.5B AI target. Spectrum-X annualized run-rate crossed $10B with Microsoft Azure, Meta, CoreWeave, xAI, Oracle, and Google Cloud all named⁶. Arista is winning the disaggregated/scheduled-fabric back-end at Meta and a fourth major InfiniBand-to-Ethernet migration; NVDA is winning everything GPU-bundled. UALink and Ultra Ethernet Consortium remain 2027+ structural threats but have not materialized in shipping deployments yet. The networking line is the single most under-appreciated piece of the NVDA story, it's a $60B+ run-rate business growing faster than compute.

Front 5: Gaming/PC cyclical exposure

Status: Win (rounding error). Graphics segment at $7.07B (+58% YoY), robust but a rounding error against Data Center; Edge Computing's new bundling acknowledges the segment's strategic re-rating away from a standalone cycle. The Switch 2 ramp (Nintendo Tegra), the RTX 5090/5080 generation, and AI-enhanced gaming (DLSS 4) all contribute. Importantly: the bundle now includes Auto, Pro Viz, Robotics, and AI-RAN, a strategic re-statement that Gaming alone is no longer the operative cycle driver.

Front 6: Valuation cap (the most defensible bear)

Status: Real constraint. At ~$5T market cap and ~22x NTM P/E with non-GAAP GM at 75% peak, the multiple has limited expansion runway; this is the single most defensible bear thesis. Comparable mega-cap-tech valuations (AAPL 31x, MSFT 22-24x, AMZN low-20s) trade in the 22-31x P/E range. Multiple compression risk is structurally one-directional. Upside case: forward P/E at the 11th percentile of NVDA's 10-year range provides asymmetric re-rating opportunity if FY27 consensus migrates above $400B and FY28 above $500B. This is why we hold the bear weight at 25% rather than the recovery lens's 15%.

Financial Health & Trends

Q1 FY27 print highlights (reported May 20, 2026)

Q1 FY27 was the strongest single quarter in modern semiconductor history. Networking growing 199% YoY at a $14.8B/qtr clip is the standout, a $60B+ run-rate business that almost no model gave NVDA credit for 18 months ago.

Networking trajectory: Spectrum-X + InfiniBand inflection

Data Center quarterly trajectory: the operational acceleration

Margin trajectory: non-GAAP GM at structural peak

Three-year P&L summary

Supply commitments: the single most important visibility data point

Manufacturing/supply/capacity purchase commitments stood at ~$119B as of April 26, 2026, with ~$95B due in the remainder of FY27 and the balance running into FY31; total supply commitments framed at ~$145B on the call. That commitment level is the single most important visibility data point, it forward-books FY28 demand at a granular SKU level and is consistent with the consensus migration from $290B to $485B for the year. Inventory rose to ~$25.8B (from ~$10.3B a year prior) reflecting Blackwell Ultra ramp and Rubin pre-build, a planned step-up rather than a digestion signal.

Capital Allocation

Capital allocation is now unambiguously shareholder-friendly. Q1 FY27 returned a record $20B to shareholders. The board authorized a fresh $80B buyback on top of remaining prior authorizations. The quarterly dividend was raised 25-fold from $0.01 to $0.25 effective June 26, 2026, a signal that management views forward FCF as not just durable but excess to capex needs¹.

The $80B buyback signal

Net debt is approximately zero ($11.4B debt against $10.6B cash, with $50B+ of quarterly FCF rapidly accumulating); the balance sheet supports a step-change in capital return without any leverage strain. The $0.25 dividend remains a token yield (~0.5%) but the directional change from $0.01 is what matters: it brings NVDA into the dividend-grower cohort that anchors mega-cap capital allocation expectations. We model continued $15-20B per quarter in buybacks through FY27, reducing diluted share count from ~24.30B to ~23.8B by FY28, modestly accretive to EPS at the margin and a stabilizer for the multiple.

R&D intensity: the moat investment

NVIDIA spends ~$15B annually on R&D (~4% of FY27E revenue), modest as a percentage given the revenue scale but the absolute dollar amount dwarfs any individual competitor's full R&D budget. The R&D is concentrated on (a) next-generation silicon (Vera Rubin, Rubin Ultra, Feynman), (b) software stack (CUDA-X, Dynamo, TensorRT, NIM, NeMo), (c) networking IP (Spectrum-X, NVLink, InfiniBand), and (d) emerging platform extensions (Omniverse, Isaac, BioNeMo, DRIVE). The compounding R&D advantage relative to AMD ($5B) and the individual hyperscaler custom-silicon teams (~$1-3B each) is the most under-discussed moat asset.

Valuation Overview

The valuation paradox

NVDA trades at ~$4.9T market cap and ~$4.9T enterprise value (net cash position approximately zero) on ~24.3B diluted shares. NTM P/E at ~22-23x is the cheapest forward multiple in the MAG7 and the 11th percentile of the ten-year NVDA range (per Bernstein). NTM EV/Sales at ~10-11x compares to a 10-year average of 20.8x, a 2024 peak of 33.7x, and a 2023 trough of 16.8x. The disconnect is the central bull-case fact: forward P/E is down ~27% year-over-year while estimates have moved sharply higher.

The cross-check: multiple expansion vs operating leverage

On Morgan Stanley's $587B FY28 estimate, NVDA implies ~8.5x EV/Sales, comparable to AVGO's current NTM multiple but on more than 5x the revenue base and superior margins. On the bear's $405B FY28, the implied multiple at 7x EV/Sales = $130, a level that the $80B buyback authorization makes structurally supported. The base case at 22x P/E × $11.40 EPS = $251 and 10x × $485B EV/Sales = $204; we blend these at 50/50 to $228, then discount back ~6% to $215. The valuation-cap critique is real (peak GM, structurally one-directional multiple compression risk, $5T base) but is offset by the EPS growth runway and capital return cadence.

Why we're at Buy rather than Hold

At the 11th percentile of a 10-year multiple range, on the asset that has just printed the strongest quarter in its history, the asymmetry is upside. The valuation-cap discipline is what sets the PT at $215 rather than $268 (the rubin-ramp lens proposal); the rating call is Buy because Hold would require denying both (a) the 11th-percentile multiple framing and (b) the $145B supply-commitment visibility. The rubin-ramp lens at $268 and the custom-silicon-pressure lens at $190 average to $229 even before weighting, our $215 PT is the disciplined version of that consensus.

Editorial note on the recent overshoot: NVDA traded above the $215 12-mo PT in May 2026, printing an all-time-high close of $235.47 on May 14 and a 52-week high of $236.54, both six days BEFORE the Q1 FY27 print on May 20. By the June 10 snapshot, the stock had pulled back to ~$200 on positioning unwind. We retain $215 as the disciplined PT (anchored to blended FV $207) and note that the "+7.5% Base Upside" in the hero stat reflects PT/spot upside, not a forecast that the May high cannot be revisited. A Q2 FY27 beat-and-raise plus Rubin first-shipment cadence is the most credible path back into the $235-$250 zone.

EV/Sales Scenario Model

FY28E revenue × EV/Sales multiple = enterprise value, plus net cash, divided by diluted shares = forward fair value, discounted one year at 6% (mega-cap time-value) to today's PT.

Sensitivity grid: FY28E revenue × EV/Sales multiple (PT in $, Base-case assumptions)

Cells use canonical Base-case shares (23.8B), zero net cash, and 6% one-year discount. Switching scenario tab moves the highlighted cell anchor but does not change cell values, for honesty, we don't silently swap assumptions per scenario.

Quick PT calculator

Risk / Reward calculator

Ask the Thesis AI-assisted checking…

Describe an NVDA scenario in natural language; the assistant returns a structured impact analysis against this dashboard's EV/Sales model, share-of-AI-compute framework, and custom-silicon ceiling debate. Powered by Claude via a Cloudflare Worker proxy.

Upcoming Catalysts

Risk Factors (Including Upside Cases)

For a Buy rating, we list bear-case risks (dominant for sizing) with embedded upside cases for each. The framework respects that custom-silicon pressure, China regulatory volatility, and valuation cap are all real, but each has a credible offset.

• Custom-silicon acceleration. Google TPU v7 Ironwood at ~4M units 2026 with 70%+ internal share; AWS Trainium 2/3 supporting Anthropic at 5GW; Meta MTIA on Broadcom multi-GW through 2029; Microsoft Maia 200 in production Jan 2026. Custom ASIC unit shipments +44.6% YoY vs merchant GPU +16.1%; cross-over forecast 2028. NVDA share-of-AI-compute compresses faster than dollars backfill in the bear scenario. Upside case: Google supply constraints (TSMC 2nm + HBM4 allocation) force more NVDA usage at Anthropic-scale builds; Meta's Feb 2026 NVDA commitment proves bundling beats in-house silicon.
• China H20/H200 regulatory durability, bear-case quantification (see Front 1). Trade attorneys flagging the 15%/25% Treasury revenue-share as potentially unconstitutional adds tail risk that Front 1 only acknowledges qualitatively. AI OVERWATCH Act could re-tighten; Taiwan Strait escalation could trigger comprehensive bans. Bear-case impact: zero China DC compute through FY28 plus a $200-500M one-time charge if AI OVERWATCH passes.
• Valuation cap, bear-case quantification (see Front 6). NTM P/E at 22-23x and non-GAAP GM at 75% peak leaves limited multiple-expansion runway. Bear-case impact: multiple compresses to 16x P/E on $8.10 FY28 EPS = $130 ($120 floor with $80B buyback support). This is the binding constraint that anchors the 25% bear weight.
• Power-grid / interconnect constraints on FY28 deployments. 1.5GW+ data center commissionings (G42 Stargate UAE, CoreWeave Texas, Microsoft Atlanta) are increasingly constrained by transmission permits and substation capacity rather than NVDA silicon. A 6-12 month slip in 5GW+ greenfield buildouts would push $20-30B of FY28 demand into FY29. Mitigation: NVIDIA is investing directly in power infrastructure partnerships (Crusoe, Vertiv, Eaton) to de-risk commissioning timelines.
• Neocloud financing under faster depreciation. CoreWeave / Lambda / Crusoe / Nscale balance sheets are structured around 4-6 year GPU depreciation, but Rubin's 10x per-token cost reduction structurally devalues installed Hopper/Blackwell faster than scheduled. Refinancing risk if neocloud asset values are written down. Mitigation: anchor customer contracts (Microsoft, Oracle, OpenAI) provide long-term lease coverage.
• Supply-commitment accounting risk. The $145B supply commitment is a forward purchase obligation, not contracted revenue. If hyperscaler capex slips in 2H 2027, NVDA could face an inventory write-down (similar in mechanic to the Q1 FY26 H20 $4.5B charge but potentially larger in absolute terms). Mitigation: supply commitments are structured against signed customer purchase orders, not speculative builds.
• GB300 → Rubin transition execution. Hyperscalers historically pause tail orders ahead of new platforms; Q3-Q4 FY27 could produce an "air pocket" even with Dynamo software upgradeability. Upside case: annual platform cadence flattens transition risk; Vera CPU adds standalone $20B incremental.
• AMD MI400/MI450 share-take. OpenAI 6GW strategic agreement with first 1GW MI450 in H2 2026; AMD DC revenue could grow 60-70%. Upside case: AMD takes share from custom-XPU plans (Maia/MTIA) rather than NVDA, leaving NVDA's merchant slot intact.
• HBM4 supply concentration and cost. SK Hynix 60-70%, Samsung 25-30%, Micron remainder; HBM4 cost inflation is the single biggest gross margin lever. Upside case: three-supplier qualification (confirmed June 5, 2026) creates pricing leverage and removes single-source risk.
• Hyperscaler capex digestion. $725B Big-4 capex in 2026 stepping to $830-900B in 2027, possible 2H27 deceleration if AI ROI metrics underwhelm. Upside case: capex is structurally ahead of demand by 12-18 months given training scaling laws and inference token volume growth.
• Software stack erosion. Open-source inference stacks (vLLM, SGLang, llm-d) reduce CUDA stickiness over time. Upside case: NVDA is co-opting via Dynamo and TensorRT-LLM upstreaming, open-stack adoption may be net positive by anchoring NVDA at the optimization layer.
• Customer concentration. Top 4 hyperscalers ~40% of DC; Microsoft alone likely largest single customer. Upside case: ACIE/sovereign growing +74% YoY actively diversifies; 80+ partner data centers >10MW (nearly doubled YoY); sovereign AI active in 40 countries.
• Networking commoditization. Ultra Ethernet Consortium v1.0 finalized; UALink ratified with first scale-up switches late 2026. Upside case: Spectrum-X bundling beats open standards through 2027; even if UALink scales, NVDA captures InfiniBand and NVLink markets that ANET/Cisco cannot reach.

Bull vs Bear Debate

The bulls are right about FY27 and most of FY28: the supply commitments validate visibility, the Rubin generational lead is real through 2027, sovereign AI and Spectrum-X are providing genuine non-hyperscaler offsets, and the multiple has compressed to a level where a beat-and-raise cycle can re-expand it 10-20%. The bears are right about FY29 and beyond: custom ASIC unit shipments will cross merchant GPU around 2028 per TrendForce; Google has proven full-stack viability at frontier scale with Gemini on TPU; Anthropic is running ~100% on Trainium+TPU; and inference (now ~2/3 of AI compute) is where ASICs are most cost-competitive.

The crux is whether NVDA can hold blended accelerator share above ~65% in FY28. We model 70% in the base, 78% in the bull, 58-62% in the bear. The networking bundle ($14.8B/qtr already, growing faster than compute) and the software moat (Dynamo + CUDA) are what keep the bear case from going below $120. The valuation-cap proposal is the most disciplined of the three rating lenses and is the reason we settled at a base PT of $215 rather than the rubin-ramp lens's $268, but it is wrong on the rating call because at the 11th percentile of a 10-year multiple range, you do not assign a Hold to the asset that has just printed the strongest quarter in its history.

Technical Analysis

Trader's view

• Three-month range $172-$215 post Q1 FY27 print; currently mid-range at $200.
• Key support: $172 (200-DMA proxy). A close below opens the path to $158 and then $138 (52-week low).
• Key resistance: $215 (our 12-mo PT and a prior 2026 reaction high). Note that NVDA already overshot this level intra-period, printing an all-time-high close of $235.47 on May 14, 2026 (six days before the Q1 FY27 report) and a 52-week high of $236.54. The pullback to the $200 anchor reflects the post-print positioning unwind; our valuation-cap discipline keeps the published 12-mo PT at $215 even after the May overshoot, see the editorial note in the Valuation section. A weekly close above $237 unlocks the path to $250-$280 where sell-side high PTs sit (Bernstein $315, Morgan Stanley $285-$288, Wedbush $310, Rosenblatt $250+).
• The stock had run 13.7% since the Q4 FY26 print, so positioning was crowded into May 20. The modest "buy the rumor, sell the news" reaction despite a clear beat-and-raise is positioning-driven, not thesis-driven.
• Implied volatility has compressed materially from the late-2025 peaks, signaling that options markets are pricing the next leg as a function of Q2 FY27 (August 2026) and Rubin commercial-shipment cadence rather than as an existential question.
• R/R from $200 to $215 target with $172 stop = 0.54:1, weak for a tactical long. For position sizing, R/R from $200 to $320 bull target with $172 stop = 4.3:1, strong for an event-driven position around Q2 print and Rubin first shipment.
• The $80B buyback authorization provides a structural bid at lower price levels that should dampen any FY27 transition-air-pocket scare to a single-digit-percent drawdown rather than a full mean-reversion.