Doubled CPU TAM, Meta Upsized to 6 GW, Q2 Guide Blows Past Street — Stock Faded on Positioning, Not Fundamentals: Maintaining Outperform, FV Raised to $325–400

Updates and supersedes the AMD Q1 2026 Flashcap published earlier today. The flashcap was written from the press release before the call had taken place; this recap incorporates the full transcript, post-call market reaction, and Street response.

Results vs. Consensus

Quality of Beat

• Revenue: Broad-based, not concentrated. Server CPU revenue +50%+ YoY (4th consecutive record quarter), with cloud and enterprise each growing >50%. Client +26% YoY (Ryzen share gains continuing). Gaming +11% on demand for Radeon 9000 series. Embedded +6% with double-digit design win momentum. Data Center AI was modestly down sequentially — but Lisa Su was explicit on the call that this was a Q4 China revenue compare artifact; Q1 China revenue was "not material" per Jean Hu. Strip the optical sequential noise out and DC AI grew handily ex-China.
• Margins: Operationally clean. The 170bps YoY gross margin expansion is doing real work despite the Instinct mix headwind starting to land. Q2 guide to 56% bakes in continued server CPU growth (which is ASP-and-mix accretive), Embedded’s accretive contribution, and client-business mix-up — offsetting MI450’s near-corporate-average dilution that begins in Q3 and ramps in Q4. Long-term GM range affirmed at 55–58% per Jean Hu.
• EPS: Beat is operational, not below-the-line. The +43% YoY EPS growth on +38% revenue and +170bps GM reflects classic operating leverage, even with OpEx running hot at +42% YoY. Tax, interest and other was a $275M net expense (in line). No share-count or below-the-line gimmickry.
• FCF: The most underappreciated print in the release. $2.6B FCF at 25% of revenue, more than triple the year-ago quarter. AMD is now generating cash at a rate that materially expands optionality on capital return, M&A, and (most importantly) on customer prepayments and supply commitments needed to lock in the 2027/2028 GPU pipeline. AMD repurchased 1.1M shares ($221M returned) in Q1 with $9.2B authorization remaining; cash and short-term investments ended the quarter at $12.3B.
• Q2 Guide: Not a maintenance guide. The $11.2B midpoint is +6.7% above the ~$10.5B Street, with +100bps of GM upside. This is management explicitly resetting the FY26 trajectory: implied 2H ramp from a $11.2B Q2 base lands FY26 in the $46–48B range vs. prior Street near $40–42B. The OpEx side is heavier than expected ($3.3B vs. ~$3.0B Street), reflecting AMD investing aggressively into the demand acceleration.

Segment Performance

Data Center

The headline number ($5.8B / +57% YoY / record / +7% sequentially) buries two distinct stories that matter more than the topline. Story one: server CPU is no longer cyclical — it is structural. Lisa Su reset the long-term server CPU TAM — from approximately $60B by 2030 (at the November Financial Analyst Day) to $120B+ by 2030, implying >35% CAGR vs. the prior 18%. The mechanism is agentic AI: as inferencing scales, agents proliferate; agents need CPUs for orchestration, data movement, and parallel execution; the historical CPU-to-GPU ratio (1:4 to 1:8 in pure host-node deployments) is moving toward 1:1+ in agentic environments. Q1 evidence corroborated the demand picture in real time: Turin ramped further; cloud and enterprise each grew >50% YoY; EPYC-powered cloud instances reached >1,600 (+~50% YoY); enterprise demand "accelerated with record revenue and record sell-through."

"Inferencing and agentic AI are increasing the need for server CPU compute, as these workloads require additional CPU processing for orchestration, data movement, and parallel execution, in addition to serving as the head nodes for GPUs and accelerators... Based on the demand signals we are seeing today and the structural increase in CPU compute requirements driven by agentic AI, we now expect the server CPU TAM to grow at greater than 35% annually, reaching over $120 billion by 2030." — Lisa Su, CEO

Story two: data center AI is on track despite the optical Q1 sequential softness. DC AI was modestly down Q/Q, but Jean Hu confirmed Q1 China revenue was "not material" — the seq decline was a Q4 China comparison artifact, not fundamental softening. Going forward, Q2 DC is guided up double-digits in both server and AI. Helios begins initial volume in Q3 and ramps significantly in Q4, with the customer order book "above our initial 2027 plans." Lisa Su upgraded confidence in delivering tens of billions in DC AI revenue in 2027 and exceeding the >80% long-term CAGR target.

"As we approach production, demand for MI450 series GPUs continues to strengthen, with lead customer forecasts now exceeding our initial plans and a growing number of new customers engaging on large-scale deployments, including additional multi-gigawatt opportunities. With this expanded visibility, we have strong and increasing confidence in our ability to deliver tens of billions of dollars in annual data center AI revenue in 2027 and to exceed our long-term growth target of greater than 80% in the coming years." — Lisa Su

Assessment: The doubled CPU TAM is the single most important data point in the print. Server CPU is AMD’s most profitable, most defensible business; an upward TAM revision of this magnitude re-rates the franchise independent of any Instinct outcome. Layer Helios on top — with customer forecasts now above plan and visibility into specific 2027 data center deployments — and the FY26 DC trajectory we modeled at the Q4 print starts to look conservative.

Client and Gaming

Client — $2.9B (+26% YoY) on continued Ryzen share gains. Commercial PC was a key highlight: sell-through of Ryzen Pro PCs grew >50% YoY as Dell, HP, and Lenovo broadened AMD offerings; new enterprise wins across financial services, healthcare, and aerospace. The Ryzen AI 400 series and Ryzen AI Pro 400 series desktop CPUs extend the AI-PC offering across consumer and commercial. Gaming — $720M (+11% YoY) driven by Radeon 9000 series GPU demand; semi-custom continues its expected console-cycle decline.

The Q2H/Q4 setup matters: Lisa Su flagged that second-half PC and gaming demand will be impacted by higher memory and component costs, and Jean Hu sized the gaming impact at "more than 20% decline in 2H vs. 1H." This is a real-world cost pass-through risk that doesn’t show up in Q2 but is on management’s mind. The client business is still expected to grow Y/Y for FY26.

"We are planning for second half PC shipments to be lower due to higher memory and component costs. Against this backdrop, we still expect our client revenue to grow year-over-year and outperform the market, driven by the strength of our Ryzen portfolio and expanding commercial adoption." — Lisa Su

Assessment: Client is doing exactly what we want: gaining share, mixing up, and management is appropriately cautious about a 2H consumer-PC headwind from memory inflation. Gaming’s 2H decline guide is modestly disappointing on the margin but does not derail the AI thesis. Net read: the segment is a contributor, not the swing factor.

Embedded

$873M (+6% YoY), with double-digit design win momentum and "billions of dollars in new wins." Strength in test/measurement/emulation, aerospace and defense, and communications, plus increased adoption of embedded x86. The narrative is consistent with prior quarters: Embedded is the slow-burn margin tailwind for the corporate gross margin algorithm. Operating margin held at 39% — structurally accretive to corporate.

Assessment: Quietly compounding. Not the story of the quarter, but the ongoing accretion to GM matters when MI450 starts diluting in Q3/Q4.

Key Topics & Management Commentary

Overall Management Tone: Confident-bordering-on-emphatic. Lisa Su led with "clear inflection in our growth trajectory and a structural shift in our business," then moved through TAM expansion, Meta upsize, and the $20+ EPS framework as the framing. Jean Hu was unusually direct on gross margin drivers and on the OpEx-overrun question. The contrast with the Q4 print — where management spent significant airtime defending the FY26 DC trajectory against bear-case framings — is striking. Q1 2026 was offense, not defense.

The Doubled Server CPU TAM

This was the topic that framed every other line item. The TAM revision — $60B to $120B+ by 2030 — doesn’t just lift the addressable market; it changes the ratio analysis. CPUs that were ≤15% of an "AI infrastructure dollar" are headed for 30%+, driven by agentic workloads consuming far more general-purpose compute per inference than the foundational-model era.

Lisa Su backed the call with bottoms-up customer forecasts and workload analysis (not top-down extrapolation), and decomposed the CPU TAM into three buckets: (1) general purpose compute (low-double-digit growth); (2) AI head nodes connecting to accelerators (smaller, faster-growing); (3) agentic AI compute (largest growth contributor). She explicitly affirmed AMD’s >50% share aspiration in the larger market.

Assessment: Lisa Su’s TAM history is largely vindicated by track record — the $60B 2030 number from late 2025 already looked aggressive at the time and was clearly conservative within months. Treating $120B as the new baseline is correct. The risk is to the upside, not the downside.

Meta Deal Expanded to 6 GW

The Meta partnership expanded to "up to 6 gigawatts of AMD Instinct GPUs spanning several product generations," including a custom MI450-based accelerator. Context: 6 GW is approximately 2–3x the deployment scale implied by prior public Meta-AMD framing. The "several product generations" + "custom" language matters — this is a multi-year, multi-product commitment, not a single-SKU order, locking in MI450 and follow-on roadmap economics.

"A key example is our expanded strategic partnership with Meta to deploy up to 6 gigawatts of AMD Instinct GPUs spanning several product generations. Our agreement includes a custom GPU accelerator based on our MI450 architecture, co-designed to support Meta’s next-generation AI workloads. Shipments are on track to begin in the second half of the year, leveraging our Helios rack scale architecture..." — Lisa Su

Combined with the OpenAI commitment (multi-GW, ramping into 2H 2026 per Q4 disclosure), AMD now has two named hyperscale anchor customers committed to multi-product Instinct roadmaps. This was the missing piece in the AI-GPU bear case — "AMD wins the spec but loses the deployment" — and it is now removed.

Assessment: The Meta upsizing is the single most important named customer disclosure since the original OpenAI deal. It validates that hyperscale deployment commitments to AMD are sticky and expanding, not one-off.

Helios / MI450 Customer Pipeline Above Plan

Lisa Su was unusually specific that customer forecasts for MI450/Helios are running above AMD’s prior 2027 plans — and that visibility extends "down to which data centers GPUs will be installed." Initial Helios volume in Q3, significant ramp in Q4, continuing into Q1 2027. Confidence raised on tens-of-billions in DC AI 2027 revenue and exceeding the >80% long-term CAGR.

"On supply, we have very good visibility into deployments on track for 2027 — visibility down to which data centers GPUs will be installed. There is tightness in supply chain and data center buildouts, but we’re confident in our ability to supply to levels of growth we’re discussing." — Lisa Su

Assessment: The "visibility to specific data centers" line is a tell — that’s the kind of granularity you only get when customers have made firm commitments and AMD has slot-allocated supply. This is materially de-risked vs. the Q4 framing of "ramping" demand.

Gross Margin Setup for FY26 Despite MI450 Dilution

Jean Hu walked through the gross margin algorithm for the year. Q2 guide of 56% (vs. Q1 actual 55%) is supported by server CPU growing >70% YoY in Q2, client mix-up, and Embedded accretion. The MI450 ramp in Q3 and significant ramp in Q4 brings near-corporate-average dilution — but the offsetting tailwinds (server, embedded, client mix) are sized to "really offset some of the gross margin dilution from MI450." She affirmed the 55–58% long-term range from the November Financial Analyst Day.

"On the other side, MI450 will start to ramp in Q3 and then ramp significantly in Q4. That is below corporate average. That will have different puts and it takes in Q4 in the gross margin side... we actually feel really good about the setup of the gross margin for 2026." — Jean Hu

Assessment: The credibility of this answer depends on the server CPU TAM call. If server is growing 50–70% YoY through 2H, the mix accretion is real and the MI450 dilution is manageable. We size FY26 GM at 55–57% (mid 56%), consistent with the long-term framework.

OpEx Running Hot — Investing Through the Cycle

OpEx was $3.1B in Q1, +42% YoY — clearly running ahead of the prior guide. Q2 is guided to $3.3B, another step up. This drew the only mildly skeptical analyst question on the call (Stacy Rasgon), and Jean Hu’s answer was unusually direct: AMD is investing aggressively because the market opportunities are tremendous, the investments are driving the revenue acceleration (Q2 guided +46% YoY), and the operating leverage is still positive. R&D is growing faster than SG&A on a Y/Y basis going forward.

Assessment: The OpEx-overrun pattern (guide-then-blow-through-then-guide-higher) is worth noting, but in context of the topline trajectory, the operating leverage math still works. Operating margin held at 25% — this is what investing through a TAM expansion looks like, not a cost-control failure. We are more concerned if revenue decelerates while OpEx stays elevated; today we have the opposite.

Memory Pricing as 2H Headwind for Client / Gaming

Lisa Su flagged memory pricing as a real and rising risk. AMD has secured supply for its needs, but consumer-PC and gaming demand may take a hit in 2H. Jean Hu sized the gaming impact specifically: 2H gaming revenue expected to decline more than 20% vs. 1H. Client is still expected to grow Y/Y for FY26, but with second-half PC shipments lower than initially planned.

Assessment: Real risk, appropriately scoped to client/gaming, not data center. Doesn’t touch the AI thesis. Sized at maybe 1–2pp off FY26 client revenue and a bigger drag on gaming.

$20+ EPS Long-Term Anchor

Lisa Su’s prepared remarks included this line, volunteered (not asked):

"With the momentum we are seeing across the business and the expanding market opportunity, we see a clear path to exceed our long-term financial targets, including delivering more than $20 in EPS over the strategic timeframe." — Lisa Su

Assessment: The $20+ EPS framework is the emphatic anchor of the new long-term model. At the current ~1.66B diluted share count, $20 EPS implies ~$33B in net income. On AMD’s typical 25% non-GAAP NM range, that maps to ~$130B+ in revenue. That is 2030+ math at the upper bound. The market hasn’t fully discounted this trajectory yet — today’s $341 stock at ~17x that long-tail EPS is reasonable.

Guidance & Outlook

The Q2 guide is unambiguous: management is calling for sequential acceleration ($10.3B → $11.2B is +9% Q/Q), a 100bps GM step up, and continued aggressive OpEx investment. The midpoint $11.2B implies +46% YoY growth — an acceleration from the +38% YoY in Q1. This is rare at AMD’s scale.

Implied 2H ramp: If we assume Q3 ~$11.7–12.2B (initial Helios ramp + continued server strength + memory-pricing client trim) and Q4 ~$12.7–13.5B (significant Helios ramp), FY26 lands in the $46–48B range — vs. prior Street consensus closer to $40–42B. That is a 10–15% upside revision to the topline, with similar or better operating leverage.

Street at: Pre-print FY26 Street was clustered around $40–42B with consensus EPS in the $5–6 range. Post-print, expect consensus to migrate to $45–47B revenue and $7–9 EPS as the model rebases to the Q2 sequential cadence. Lisa Su’s ">$20 EPS over the strategic timeframe" reframes the long-tail FCF generation conversation entirely.

Guidance style: AMD has historically guided at-or-slightly-above whisper, then beat by 2–5%. The Q2 guide breaks pattern — this is a guide that is clearly above any reasonable whisper, suggesting management is highly confident in the demand visibility they are seeing. Track record says the actual Q2 print likely lands at $11.4–11.6B.

Analyst Q&A Highlights

The Doubled CPU TAM

• Joshua Buchalter, TD Cowen: Asked how the 2030 server CPU TAM doubled so quickly from the November Analyst Day figure, and whether AMD remains confident in its >50% share target. Lisa Su attributed the revision to faster-than-expected agentic AI scaling and bottoms-up customer forecast analysis showing CPU-to-GPU ratios moving from 1:4–1:8 toward 1:1+; affirmed >50% share confidence.
  Assessment: This was the framing question of the call. Su’s answer was confident and specific — not aspirational. The fact that it was the lead question signals the Street recognizes the TAM revision as the central event.

MI450 / Helios Customer Pipeline

• Joshua Buchalter, TD Cowen (follow-up): Asked whether the strengthening forecasts referred to OpenAI/Meta upsizing or to broader customer set, and whether the timeline shift was MI450 or MI500-and-beyond. Lisa Su confirmed it was both — OpenAI and Meta partnerships "going really well," and forecasts above the initial 2027 plan across the breadth of the customer set, with inference dominant. Many customers are also engaged on MI500 series.
  Assessment: The breadth claim is more important than the named-customer claim. Inference being the dominant workload is the right answer for AMD — that’s where ROCm-on-Instinct economics are most competitive.
• Thomas O’Malley, Barclays (effectively delivered by another caller): Asked about cadence through the year and supply concerns. Su confirmed Q2 DC up double digits in both server and AI; server CPU growing >70% YoY in Q2; Helios initial volume Q3, significant ramp Q4, continuing in Q1. On supply: "very good visibility into deployments on track for 2027 — down to which data centers GPUs will be installed."
  Assessment: The data-center-level visibility is a meaningful disclosure. AMD wouldn’t say this on the record unless customer commitments were firm. This is the most de-risked the 2027 plan has been since the OpenAI deal.

Competitive Landscape (x86 vs. Arm)

• Ross Seymore, Deutsche Bank: Asked how AMD differentiates with Arm competition signing up the same customers and the x86 competition having more supply. Lisa Su was emphatic: AMD’s portfolio breadth is the differentiator; hyperscalers will use both x86 and Arm; demand is high.
  Assessment: A reasonable answer, though slightly evasive on the supply-disadvantage question. AMD’s TSMC-only manufacturing model means it is structurally tighter than the integrated competitor — this is real but management appropriately frames it as a supply chain partnership strength.

Gross Margin Trajectory

• Ross Seymore, Deutsche Bank (follow-up): Asked about long-term GM with Helios/Instinct ramping. Jean Hu walked through the offsetting tailwinds (server >70% Y/Y in Q2; client mix-up; Embedded accretion) vs. MI450 dilution beginning in Q3, and affirmed the 55–58% long-term range.
  Assessment: The answer was structured and credible. Management is signaling FY26 GM holds at-or-above current levels despite the MI450 mix headwind. We are underwriting 55–57% for FY26.

Server CPU Units vs. Pricing

• Timothy Arcuri, UBS: Asked how much of server CPU growth is units vs. pricing. Lisa Su was clear: Q1 was much more unit-driven than ASP-driven; Q2 and 2H growth is also predominantly unit-driven, with some inflationary cost pass-through but not aggressive pricing.
  Assessment: This is the right answer for franchise durability. Unit-driven growth is share-driven growth; ASP-driven growth is mix-driven and more vulnerable to commoditization. The unit-led story is more defensible.

OpEx Discipline

• Stacy Rasgon, Bernstein: Asked why OpEx keeps blowing past targets, calling out the guide-then-overrun-then-reguide pattern. Jean Hu’s answer was unusually direct: investments are driving revenue acceleration, market opportunities are tremendous, AMD is leaning in.
  Assessment: The honest answer is that AMD is choosing to invest ahead of revenue, and the operating leverage is still positive. This should not be conflated with a cost-control problem. The risk is asymmetric: if topline holds, the equation works; if topline trips, the leverage reverses fast.

What They’re NOT Saying

1. No specific FY26 revenue guide. AMD historically does not give annual guidance, but the implicit FY26 framing (Q1 print + Q2 guide + commentary on second-half ramp) is the closest they get. Management chose not to put a number out, even though the data clearly supports a $46–48B FY26 walk. The omission is structural, not signaling.
2. No specific Helios revenue number. "Initial volume in Q3" and "significant ramp in Q4" are not quantified. The Street is left to back into Helios contribution from total DC growth. Our estimate: Q3 Helios at ~$200–500M, Q4 at ~$1.0–1.5B based on the cadence framing — but management did not confirm.
3. Instinct-margin convergence timeline remains unspecified. Jean Hu acknowledged MI450 is below corporate average but did not commit to a timeline for convergence. Sizing this risk is hard but real.
4. No commentary on residual MI308 inventory. The Q1 2025 print included an $800M MI308 inventory writedown for the China export ban. There was no commentary on residual MI308 inventory or any China policy update on this call. Either it is a non-issue (likely) or management deferred (less likely).
5. No update on ZT Systems integration. The acquisition closed in 2025; management has historically referenced its contribution to system-level engineering. Not mentioned today — likely a sign integration is on track and not noteworthy, but the absence of mention is itself a small data point.
6. No semi-custom 2026 guide and no console-cycle next-gen update. Jean Hu noted semi-custom continues to decline as expected; Lisa Su mentioned engagements on next-generation platforms remain strong but provided no specifics on timing or design wins.
7. Q2 ETR drop to 13% is not explained. A meaningful step-down from the FY25 effective rate. Likely R&D credit timing or geographic mix, but management did not call out the driver. Worth tracking.

Market Reaction

• Pre-print setup: Stock close ~$341, +59% YTD, +255% over the trailing 12 months, +~70% from the Q4 2025 post-print low (~$200). Forward P/E mid-30s on FY26 estimates. Options market priced an implied move of ~10–12% straddle.
• After-hours move: Initial reaction +~6% on the headline beat. As the call progressed and the OpEx step-up + 2H gaming −20% color landed, the move reversed to ~−5% net. Classic positioning unwind — AMD’s last two prints (Q3 25 and Q4 25) faded similarly on otherwise clean prints.

This is the third consecutive AMD beat-and-fade — a positioning-driven phenomenon when the stock has run hard into the print, not a fundamentals reaction. The print is not the issue; the +59% YTD setup is. The fundamental upgrade case (doubled CPU TAM, Meta to 6 GW, MI450 customer pull above plan, $20+ EPS framework) is intact and arguably strengthened — the market is repricing positioning, not the underlying franchise trajectory.

Street Perspective

Debate: Is the $120B Server CPU TAM Credible?

Bull view: Lisa Su has a track record of TAM calls vindicated by reality (the $60B 2030 number from late 2025 was already conservative and is being revised in 6 months); the bottoms-up customer forecast methodology gives the new number specific underpinning rather than top-down hand-waving; Q1 actuals (Turin ramp, instances +50% YoY, ASPs lifting) corroborate the demand picture in real time.

Bear view: Doubling a TAM in two quarters is unusual and could reflect cyclical pull-forward (memory hoarding by hyperscalers ahead of GPU buildout) rather than structural demand; the agentic AI thesis is still nascent — if agent adoption disappoints, the CPU growth doesn’t materialize at the projected pace; the $120B number assumes pricing holds, which is not guaranteed if x86 supply expands.

Our take: The bull view has the better evidence. Bottoms-up customer-by-customer rollups carry more weight than top-down extrapolation, and the breadth of corroboration in Q1 actuals (instances, units, ASPs all moving together) is hard to fake. We treat $120B as the new baseline; risk is to that number is upside, not downside.

Debate: Are FY26 GMs Sustainable Once Helios Ramps?

Bull view: Jean Hu walked through the offsetting tailwinds (server >70% Y/Y, client mix-up, Embedded accretion) and explicitly said "we feel really good about gross margin setup for 2026." Long-term 55–58% range affirmed.

Bear view: Helios at "below corporate average" is a real headwind starting Q3 and accelerating Q4; the offsetting tailwinds depend on server CPU continuing to grow 50%+ YoY, which requires the agentic AI thesis to materialize as forecasted; if even one of the offsetting tailwinds underperforms, FY26 GM trips below 54%.

Our take: The bull view is more credible but the asymmetry is real. FY26 GM in the 55–57% range is the central case; risk is to the downside if server decelerates. We size FY26 GM at 56% +/- 100bps in our framework.

Debate: Beat-and-Fade — Is the Stock Run Sustainable?

Bull view: The print and Meta-6GW deal materially de-risk the FY27/28 trajectory; the doubled TAM expands the multi-year EPS power; the operating leverage on $46–48B FY26 revenue creates earnings that justify the multiple even after the run.

Bear view: +4x in 12 months and +59% YTD is a cliché peak-sentiment setup; the AI capex cycle is overdue for a digestion period; valuation at 17–20x FY28 EPS (on Street’s new numbers) is at the high end of historical AMD multiples; this is the third consecutive beat-and-fade — the stock’s buyers have been wrong-footed for three straight quarters.

Our take: The bull case has stronger near-term support (Q2 guide, Meta deal, customer visibility to 2027), but we acknowledge multiple compression risk. The beat-and-fade pattern is real and reflects extreme positioning, not fundamentals. The print justifies our updated $325–400 fair-value range; the post-fade level near $324 is at the lower bound, leaving meaningful upside without requiring multiple expansion. Holders should expect continued volatility but the operating story has not been better.

Model Update Needed

Valuation impact: Our prior fair-value framework was $235–280, set at the Q4 2025 print. We raise to $325–400:
• Low end $325 = ~28x forward FY27 EPS of ~$11–12 (conservative bridge)
• High end $400 = ~33x forward FY27 EPS of ~$11–12 (in line with Lisa Su’s $20+ long-term framework)
The new range straddles the post-fade level near $324 and gives room for the typical 12–18 month migration up the long-term EPS curve. The stock is roughly fair-valued at the post-fade level, with upside if the FY27 trajectory continues to come in above plan.

Thesis Scorecard Post-Earnings

Overall: Thesis materially strengthened. Three of four bull pillars confirmed-plus; the new fourth pillar (agentic AI / TAM expansion) is the upgrade-class data point of the cycle. Bear case is intact but appropriately scoped — no thesis-breaking surprises.

Action: Maintaining Outperform; raising fair value to $325–400 from $235–280. We do not chase — the post-fade level is at the lower end of the new range — but we would not trim a position here. Holders should reassess sizing if the stock breaks above ~$390 (reaching the upper half of fair value) or underperforms peers by >10% over a multi-month period without thesis damage.