> 📌 Previous Analysis: [AMD Base Target Reached: Executing Partial Profit-Taking at $467 (May 2026 Reanalysis)](https://mybestinvesting.co.kr/?p=1671)
When we last covered Advanced Micro Devices (NASDAQ: AMD) in May 2026, the stock had just tagged our base-case target and we told readers to execute partial profit-taking at $467. Four months later, AMD trades at $613.14, briefly touched the $1 trillion market-capitalization club on September 21–22, 2026, and rallied roughly 10% in a single session. Our original base target of $520 has been cleared; our bull target of $650 is within touching distance. The question that matters for anyone holding a position today is no longer “is the thesis right?” — it clearly was — but “how much reward is left, and at what risk?”
This reanalysis exists because the facts changed materially. Two catalysts we could only sketch in outline back in May are now concrete: AMD announced a multi-year partnership with OpenAI to deploy 6 gigawatts of Instinct GPUs, and it formally launched the MI400 series and Helios rack-scale systems at its Advancing AI 2026 event. Both events reshape the data-center growth curve — and both are now at least partially priced in. The stock has re-rated from a contrarian value-with-a-catalyst story into a consensus Strong Buy, which is precisely when disciplined investors slow down and re-underwrite.
Three things this article will establish. First, the operating results have validated the thesis: Q2 FY2026 revenue reached $11.54 billion, up roughly 50% year over year, with data center and AI accelerators driving the surge. Second, valuation has caught up: at a trailing P/E of 157 and a forward P/E of 38.6 on consensus next-year EPS of $15.88, AMD is no longer cheap on any conventional measure, and the average analyst price target of roughly $616–633 implies only low-single-digit upside from here. Third, the risk-reward has inverted — the asymmetry that made this a high-conviction buy at $154 (the 52-week low) and even at $467 has largely closed, which changes the appropriate action from “accumulate” to “harvest and hold a core.”
Below we walk through the full company and industry picture for first-time readers (Sections 1–7), then turn to what has changed since our prior coverage, the current scorecard, revised targets, and an updated exit plan for existing holders (Sections 8–11).
—
1. Company Overview
Advanced Micro Devices designs high-performance computing silicon and sells it into four end markets. The company does not own fabrication plants; it is a fabless designer that outsources manufacturing primarily to TSMC, which lets it concentrate capital on architecture, packaging, and software rather than multi-billion-dollar fabs.
Revenue is organized into four reportable segments:
Segment What it sells Primary demand driver Data Center EPYC server CPUs, Instinct AI accelerators (GPUs), Pensando DPUs Cloud AI training/inference, hyperscaler capex Client Ryzen desktop and laptop CPUs PC refresh cycle, AI PC upgrade Gaming Semi-custom SoCs (consoles), Radeon GPUs Console cycle, discrete GPU demand Embedded Xilinx FPGAs, adaptive SoCs (from the 2022 acquisition) Industrial, automotive, networking, aerospace
The center of gravity has shifted decisively toward Data Center. In the most recent quarter, group revenue was $11.54 billion, and the data-center franchise — EPYC server CPUs plus Instinct AI GPUs — is now the largest and fastest-growing contributor. On a trailing-twelve-month basis AMD generated $41.30 billion in sales and $6.43 billion in net income, with a gross margin of 50.4%, an operating margin of 15.7%, and a net margin of 15.6%.
AMD’s market position is that of the clear number two in both of its strategic battlegrounds. In x86 server CPUs it has taken share steadily from Intel for several years on the strength of the EPYC line’s core density and performance-per-watt advantage. In AI accelerators it is the only credible alternative to NVIDIA, though a distant second — NVIDIA still controls an estimated 80–90% of the AI accelerator market. AMD’s strategy is not to displace NVIDIA outright but to capture the portion of hyperscaler and sovereign-AI demand that wants a second source, competitive price-performance, and an open software stack (ROCm) as an alternative to CUDA.
On ownership and governance, AMD is overwhelmingly institutionally held, with the major index and active managers dominating the register; insider ownership is small, as is typical for a large-cap technology company of this vintage. CEO Lisa Su, in the role since 2014, remains the single most important governance variable — her credibility with hyperscaler customers and the Street is a genuine, if intangible, asset, and a reason management-continuity risk features in our risk section.
—
2. Industry Analysis
2-1. Market Size & Growth Trajectory
AMD sits at the intersection of two large and structurally growing markets: data-center AI accelerators and server CPUs. The AI accelerator opportunity is the one that has re-rated the entire semiconductor complex. Management has repeatedly pointed to a data-center AI accelerator total addressable market on the order of hundreds of billions of dollars by the back half of the decade, expanding at a compound annual growth rate that dwarfs the broader semiconductor market’s historical mid-single-digit pace.
The relevant frame is that we are still in the acceleration phase, not maturity. The signal for that is capital behavior: hyperscalers (Microsoft, Amazon, Google, Meta, Oracle) and a new tier of AI-native buyers (OpenAI and other frontier labs) are committing to multi-year, multi-gigawatt infrastructure build-outs. AMD’s own 6-gigawatt OpenAI agreement — with the first gigawatt of MI450 GPUs slated for the second half of 2026 — is a concrete data point that demand visibility now extends years forward rather than quarter to quarter. When customers pre-commit power and capacity at that scale, the industry is not late-cycle.
At the same time, investors should be clear-eyed that a market growing this fast on this much capex is inherently more volatile than a mature one. The same forward commitments that make the growth real also make the eventual digestion phase — whenever it comes — potentially sharp. Cyclicality has not been repealed; it has been deferred and amplified.
2-2. Structural Growth Drivers
Driver 1 — Sovereign and enterprise AI build-outs create a genuine second-source market. The most important structural change of the past year is that buyers now actively want an alternative to a single accelerator supplier. Concentration risk, pricing leverage, and supply assurance all push large buyers toward dual-sourcing. AMD is the only vendor positioned to absorb that demand at scale. The Cisco/AMD/HUMAIN platform going live in Saudi Arabia — running MI355X GPUs today with plans to scale toward 250 megawatts using next-generation MI400-class parts by 2027 — is a template for sovereign-AI deployments that were barely on the map a year ago. Each such deployment is measured in gigawatts and years, not units and quarters, which converts AMD’s opportunity from episodic product cycles into a recurring infrastructure annuity.
Driver 2 — The MI400/MI450 generation narrows the technical gap at exactly the right moment. Built on the CDNA 5 architecture and TSMC’s 2nm process node, the MI400 series represents a generational leap intended to compete directly with NVIDIA’s rack-scale systems rather than merely on a chip-to-chip basis. The accompanying Helios rack-scale system — reported to be in production and shipping within months of its Advancing AI 2026 unveiling — matters because the competitive battleground has moved from individual GPUs to integrated racks (compute + networking + software). By fielding a rack-scale answer, AMD becomes viable for the largest training clusters, not just inference and mid-scale workloads. This is the difference between competing for scraps and competing for flagship deployments.
Driver 3 — Server CPU share gains provide a profitable, less-hyped ballast. Lost in the AI-GPU excitement is that AMD continues to take x86 server share with each EPYC generation, and the launch of 6th Gen EPYC processors extends that runway. Server CPUs carry attractive margins and fund the more capital-intensive GPU roadmap. Critically, AI build-outs are not CPU-neutral — every GPU cluster needs host CPUs, and AMD’s ability to sell the CPU alongside the accelerator (and the DPU) is a compounding, platform-level advantage that a pure GPU vendor lacks. This driver is more of a steady multi-year grind than a step-change, but it is the most durable of the three.
2-3. Competitive Landscape
Company Role in AI/data center Approx. data-center scale Moat characteristic NVIDIA Dominant AI accelerator leader ~$75B quarterly data-center revenue CUDA software lock-in, rack-scale scale, ~75% gross margin AMD Clear #2 accelerator + #2 server CPU Data center a growing majority of $11.5B quarterly group revenue Open ROCm stack, CPU+GPU+DPU platform, second-source demand Intel Legacy server CPU incumbent, AI laggard Declining server share Manufacturing/foundry ambitions, x86 installed base Custom silicon (hyperscaler ASICs) In-house accelerators (TPU, Trainium, Maia) Captive to each cloud Cost/workload optimization for owner’s stack
AMD’s competitive position is best understood in relation to NVIDIA. The gap remains vast: NVIDIA’s data-center revenue is roughly an order of magnitude larger, its gross margin near 75% versus AMD’s 50%, and its CUDA ecosystem is the default that a generation of AI engineers learned on. AMD is not going to win the market. But it does not need to. Even capturing a low-double-digit share of a rapidly expanding accelerator TAM, plus continued server-CPU gains, supports the growth already embedded in consensus estimates. Against Intel, AMD is simply winning; against captive hyperscaler ASICs, AMD’s advantage is that merchant silicon serves buyers who lack the scale or desire to design their own — and even the ASIC builders buy merchant GPUs for flexibility. The one competitor that can hurt AMD is the one it will never beat outright, which is why the investment case rests on “profitable number two,” not “future number one.”
—
3. Economic Moat Analysis
Moat Type 1: Platform breadth and switching costs in the data center
AMD’s most underappreciated moat is that it can sell an entire data-center compute platform — EPYC CPUs, Instinct GPUs, and Pensando DPUs — under one roof, validated to work together. For a hyperscaler or sovereign buyer standardizing on infrastructure across gigawatts of capacity, that integration reduces engineering risk and creates switching costs once a fleet is deployed and its software tuned. The evidence is in the deal structure of recent wins: the OpenAI agreement and the HUMAIN/Cisco platform are not one-off chip orders but multi-year, multi-gigawatt commitments with roadmaps attached. Buyers do not sign those unless they intend to standardize, and standardization is stickiness. The 6th Gen EPYC + MI400 + Helios stack deepens this by making AMD a rack-scale vendor rather than a component supplier.
Moat Type 2: Cost advantage from a fabless model plus chiplet architecture
AMD pioneered the chiplet approach to CPU design, disaggregating a processor into smaller dies that yield better and can be mixed and matched across product lines. Combined with its fabless model — leaning on TSMC’s leading-edge nodes without carrying fab capital — this gives AMD a structural cost and time-to-market advantage over an integrated competitor like Intel that must fund its own manufacturing. The proof shows up in the financials: a 50.4% gross margin on a business that is still ramping its highest-cost GPU products, and a balance sheet carrying almost no leverage (debt-to-equity of 0.06), which means the company funds its roadmap from operations rather than debt. That financial flexibility is itself a competitive weapon in a capital-intensive race.
Moat Durability Assessment
Will these moats hold over the next 5–10 years? The platform/switching-cost moat is strengthening as deployments scale — each gigawatt deployed makes the installed base harder to displace. The cost-advantage moat is durable as long as TSMC leadership holds and AMD’s chiplet execution stays ahead of Intel’s turnaround. The genuine threat to durability is on the software side: NVIDIA’s CUDA remains the industry default, and AMD’s ROCm, while improving, has not yet reached parity in developer mindshare. If AI frameworks continue to abstract away the underlying hardware (as PyTorch and inference runtimes increasingly do), AMD’s moat strengthens; if CUDA lock-in reasserts itself at the frontier, AMD’s ceiling lowers. Our assessment: the moats are real and widening on the hardware and platform axes, but the software axis is the swing factor that separates the base case from the bull case. This is a moat that is deepening, not one that is secure.
—
4. Financial Analysis
AMD’s financial trajectory has inflected sharply upward, driven almost entirely by data center. The trailing-twelve-month figures tell the story: $41.30 billion in revenue, $6.43 billion in net income, and a return on equity of 10.2% — the latter still modest because the balance sheet carries a large equity base (including goodwill from the Xilinx acquisition) relative to earnings that are only now scaling.
The growth rates are the headline. Quarterly revenue grew roughly 50% year over year (Sales Q/Q growth of 50.1% on the most recent comparison), EPS this year is tracking up 83%, and the most recent quarter’s EPS grew 159% year over year — the operating leverage of a business whose highest-margin AI products are finally shipping in volume. Q2 FY2026 group revenue of $11.54 billion confirms the ramp is real and accelerating, not a one-quarter spike.
Metric (TTM / latest) Figure Read-through Revenue (TTM) $41.30B Data center now the growth engine Net income (TTM) $6.43B Margin inflection underway Gross margin 50.4% Healthy despite GPU ramp mix Operating margin 15.7% Room to expand as GPU scale builds EPS (TTM) $3.90 Trailing base still catching up to price ROE / ROA 10.2% / 8.1% Improving but not yet peer-leading Debt / Equity 0.06 Near-zero leverage; self-funded roadmap
The key operating metrics to watch are not the GAAP lines but the leading indicators: data-center GPU backlog and deployment cadence (the OpenAI gigawatt schedule, HUMAIN scaling), server CPU unit share, and gross-margin trajectory as the MI400 mix builds. The balance sheet is a genuine strength — near-zero net debt gives AMD the flexibility to fund a capital-intensive roadmap, pursue tuck-in acquisitions, and weather a demand air-pocket without distress. The margin-expansion story is the crux of the financial case: at 15.7% operating margin, AMD sits far below NVIDIA’s data-center-driven profitability, and the bull case rests on that gap narrowing as Instinct volumes scale and software attach improves. If operating margin marches toward the mid-20s over the next few years, the earnings power justifies today’s valuation; if it stalls in the mid-teens, the stock is expensive.
—
5. Valuation
AMD is no longer a valuation-friendly stock, and any honest analysis has to start there. At $613.14, the trailing P/E is 157 — a number that is meaningless on its own because it reflects earnings that are a year or two behind the revenue ramp. The relevant multiple is the forward P/E of 38.6, based on consensus next-year EPS of $15.88. Price-to-sales sits at 24.2 and price-to-book at 14.9. These are growth-stock multiples that require the growth to show up.
Because EPS is solidly positive and forward earnings are the right lens for a company at this stage of a ramp, we value AMD on a forward-P/E basis anchored to consensus EPS next year of $15.88, cross-checked against the analyst consensus target.
Base case — $640. Applying a forward multiple of roughly 40x to next-year EPS of $15.88 yields about $635, which we round to $640. A 40x forward multiple is defensible for a business compounding revenue at ~50% with a widening data-center moat, and it lands essentially in line with the analyst consensus target of roughly $616–633. Against the current price of $613, that is only about 4–5% upside — the market has largely closed the gap to fair value.
Bull case — $780. If the MI450/OpenAI ramp exceeds schedule, ROCm attach improves, and operating margin visibly expands toward the low-20s, the market could sustain a ~49x forward multiple on an upwardly revised EPS, supporting roughly $780 (about 27% above the current price). This is the “second-source wins bigger than expected” scenario.
Bear case — $400. If AI capex digestion arrives, GPU pricing normalizes, or MI400 execution slips, EPS could be revised toward ~$12–13 and the multiple compress to the high-20s, taking the stock to roughly $400 (about 35% downside). Note this is well above our prior $300 bear, reflecting a genuinely higher, execution-proven floor — but it is still a meaningful drawdown from here.
Reconciliation with consensus. The Street’s average target of roughly $616–633 (range $565–660, per recent Piper Sandler $600 and Stifel $635 notes) sits right on top of our base case. We agree with the direction and largely with the level — where we differ is emphasis: consensus frames ~$630 as a target to reach, whereas we frame it as fair value already substantially achieved. When our base case and the consensus target both sit within a few percent of the current price, the signal is not “buy more” but “the easy money has been made.”
—
6. Risk Factors
Risk 1 — Valuation compression / AI-capex digestion. This is the dominant near-term risk. AMD trades at 38.6x forward earnings and 24x sales because the market assumes multi-year hyperscale AI capex continues uninterrupted. If any large buyer signals a pause, if training demand plateaus, or if the market simply rotates away from AI beneficiaries, a stock priced for perfection can lose a third of its value without any change in the long-term thesis — purely on multiple compression. The forward commitments (OpenAI’s 6 GW, sovereign deals) provide a demand cushion, but they also mean a lot of good news is already in the price. At the 52-week low of $154.78, the margin of safety was enormous; near the 52-week high of $616.69, it is minimal. Buying here means underwriting continued flawless execution.
Risk 2 — The NVIDIA/CUDA moat and competitive response. AMD’s entire second-source thesis depends on buyers wanting an alternative and on ROCm being good enough to make switching costs bearable. NVIDIA is not standing still: its scale (roughly 10x AMD’s data-center revenue), ~75% gross margins, and entrenched CUDA ecosystem give it the resources to compete on price, bundle software, and out-invest AMD on the next node. If NVIDIA’s next generation widens the performance-per-dollar gap, or if CUDA lock-in proves stickier than expected at the frontier labs, AMD’s share ceiling lowers and the bull case evaporates. This is a structural risk that no single quarter will resolve.
Risk 3 — Execution and concentration risk on the MI400 ramp. The bull case rests on the MI400/MI450 generation and the Helios rack shipping on schedule, at yield, on TSMC’s 2nm node. Any slip — process yield issues, packaging constraints (advanced packaging capacity is tight industry-wide), or software integration delays — would directly hit the 2026–2027 revenue curve that the current multiple assumes. Compounding this is customer concentration: a handful of hyperscalers and now OpenAI represent a large share of the data-center opportunity, so the loss or deferral of a single large commitment would be material. Finally, key-person risk around CEO Lisa Su is real; her relationships and credibility are woven into the largest deals, and an unexpected transition would introduce genuine uncertainty at a delicate moment in the roadmap.
—
7. Conclusion & Exit Plan
Investment rating: Hold (with partial profit-taking). The thesis worked. AMD did exactly what we underwrote — data center inflected, the MI400 generation landed, a landmark OpenAI partnership materialized, and the stock cleared our base target and approached our bull target. But a correct thesis and a good entry point are different things. At $613, with ~4–5% upside to our base case and to consensus, the risk-reward that made this a high-conviction buy has largely closed. The right posture now is to harvest gains on schedule and hold a core position for the multi-year MI400/sovereign-AI optionality, not to chase.
Entry price range (for those without a position). We would not initiate a full position at $613. A patient buyer should look to accumulate on weakness toward the $480–520 zone (a re-test of the old base target as support) and treat anything near the bear case of $400 as an aggressive-add level. Buying at the high after a 4x run from the 52-week low is the opposite of the discipline that made this trade work.
Exit conditions.
– Target achieved: trim on strength — take partial profits into the $640 base target and further into the $780 bull target (detailed staging in Section 11).
– Fundamental break: reduce materially if data-center revenue growth decelerates below ~20% year over year, if operating margin fails to expand above the mid-teens over two consecutive quarters, or if the MI450/OpenAI deployment schedule slips beyond H2 2026.
– Time-based: reassess in 6 months (by roughly March 2027) or immediately upon the next earnings report and any MI400 ramp update.
Item Detail Company Advanced Micro Devices, Inc. (AMD) Current Price $613.14 Target Price (Base) $640 Upside to Base ~4–5% Rating Hold (partial profit-taking) Key Thesis Profitable #2 in AI accelerators + server CPU share; OpenAI 6 GW + MI400 validate the multi-year ramp Main Risk Valuation compression / AI-capex digestion at 38.6x forward earnings
—
8. What Changed Since Last Analysis
When we first covered AMD in depth and then reanalyzed it in May 2026 at $467, the core of our thesis rested on five ideas. Here is where each stands today.
Idea 1 — Data-center revenue would inflect on AI accelerators. Status: strengthened, playing out better than expected. At the time we pointed to Q1 data-center revenue of $5.8 billion, up 57%. Since then, group revenue reached $11.54 billion in Q2 FY2026, up roughly 50% year over year, and the most recent quarter’s EPS grew 159%. The inflection we forecast is not just intact — it is accelerating, and the operating leverage is finally visible in the margin line.
Idea 2 — AMD’s server-CPU share gains would continue as profitable ballast. Status: intact. The launch of 6th Gen EPYC processors extends the runway we described, and server CPUs continue to fund the GPU roadmap. Nothing here has weakened; it remains the steadiest leg of the stool.
Idea 3 — A credible second-source AI accelerator would find real demand. Status: dramatically strengthened — this is the biggest positive surprise. In May we could only argue that second-source demand should exist. It now demonstrably does: AMD announced a multi-year partnership with OpenAI to deploy 6 gigawatts of Instinct GPUs, beginning with 1 gigawatt of MI450 parts in the second half of 2026, complete with a warrant issued to OpenAI for up to 160 million shares tied to deployment and share-price milestones. The Cisco/AMD/HUMAIN sovereign deployment in Saudi Arabia reinforces the same point. This converted our hypothesis into signed, multi-gigawatt commitments.
Idea 4 — The next-generation roadmap would narrow the gap with NVIDIA. Status: delivered, ahead of where we expected. AMD formally launched the MI400 series (CDNA 5, TSMC 2nm) and the Helios rack-scale system at Advancing AI 2026, with Helios reported to be in production. AMD is now a rack-scale competitor, not just a chip supplier — a step-change in addressable deployments.
Idea 5 — The contrarian valuation edge would re-rate as the thesis proved out. Status: fully realized, and now a headwind. This is the idea that has “faded” — not because it was wrong, but because it succeeded completely. The re-rating we wanted has happened; analyst consensus is now Strong Buy, and the discount is gone. What was our tailwind is now our caution.
New idea that has emerged: the OpenAI equity-warrant structure introduces a novel alignment (and potential dilution) dynamic that did not exist in May — a customer with an equity stake tied to deployment. It is a powerful demand signal, but the up-to-160-million-share warrant is also a future dilution consideration to monitor.
New risk not present before: customer concentration has increased, not decreased. The very deals that de-risked demand (OpenAI, sovereign platforms) also concentrate a larger share of the opportunity in a handful of counterparties.
9. Current Assessment
The scorecard, using concrete numbers, is unambiguous on performance and cautionary on forward reward.
– Price journey: From the May 2026 reanalysis level of $467 to today’s $613.14, the stock has returned roughly +31% in about four months. Measured from the 52-week low of $154.78, it is up nearly 4x.
– Targets reached: Our prior base case of $520 has been cleared (the stock is ~18% above it), and the bull case of $650 is within ~6% — essentially in reach. The bear case of $300 is a distant memory. In target terms, the trade has substantially completed.
– Milestone: AMD briefly entered the $1 trillion market-capitalization club on September 21–22, 2026 (market cap $1,000.94 billion), a symbolic confirmation of how far sentiment has traveled.
– Time elapsed: Approximately four months since the prior reanalysis, and about four and a half months since our initial deep-dive coverage in May.
Current holding stance: maintaining a core position while taking partial profits. The thesis is intact and, on the fundamentals, stronger than when we wrote it. But the reward remaining to our base case is now low-single-digit percent, versus the roughly 35% downside to our bear case — an inverted risk-reward relative to the entry. That combination — winning thesis, stretched price — is the textbook setup for booking gains on plan and holding a reduced core for the long-dated optionality, rather than adding.
10. Revised Price Target & Valuation
We recalculate fair value on the same forward-P/E framework, anchored to consensus next-year EPS of $15.88, and compare to our prior targets.
Scenario Previous Target Revised Target Change Key Driver Base Case $520 $640 +23% Consensus EPS $15.88 × ~40x forward; in line with Street Bull Case $650 $780 +20% MI450/OpenAI ramp beats, margin expansion to low-20s, ~49x forward Bear Case $300 $400 +33% Higher execution-proven floor; AI-capex digestion cuts EPS toward $12–13
What drove the changes. Every target moved up, but for a specific reason in each case. The base case rose because the earnings base itself re-rated — next-year consensus EPS of $15.88 is far above what was modeled in May, so even a disciplined ~40x forward multiple produces a much higher fair value. The bull case rose modestly to reflect that the MI400/OpenAI catalysts we hoped for are now confirmed, raising the credible ceiling on both EPS and the sustainable multiple. The bear case rose the most in percentage terms because execution risk has genuinely fallen: the products shipped, the deals signed, and the demand is contracted, so the realistic downside floor is higher than the speculative $300 we carried when the ramp was unproven.
Comparison with consensus. Recent analyst targets cluster around $616–633 on average (range roughly $565–660), with Piper Sandler at $600 and Stifel at $635. Our base case of $640 sits at the upper edge of that consensus band — we agree with the Street on level, and our small premium reflects giving AMD credit for the confirmed OpenAI/MI400 catalysts. Where we diverge is interpretation: because our base case and consensus both sit within a few percent of the current $613, the appropriate takeaway is that AMD is trading at fair value, not below it. The valuation math no longer argues for accumulation.
11. Updated Exit Plan
Recommended stance: continue holding a core position, trim on strength. For current holders who have ridden this from lower levels, the plan is to convert paper gains into realized gains on a disciplined schedule while retaining exposure to the multi-year AI-infrastructure optionality.
Staged profit-taking.
– The prior plan already called for booking 50% into the $380/$450 levels (confirmed in the May reanalysis) and a further 25% at the $520 base — both of which have now been reached. Holders who followed the plan have already de-risked substantially.
– Take another 25% off the table around the $640 base target as the stock trades into it.
– Trim a further portion (target another 25% of the remaining position) if the bull case near $780 is reached, leaving only a long-term core.
– The objective is to finish with a modest core position carried for optionality, funded by realized gains, rather than a full-size position exposed to a stretched multiple.
Updated stop-loss / impairment triggers. The core thesis is invalidated — and the remaining position should be cut — if any of the following occur:
– Data-center revenue growth decelerates below 20% year over year for two consecutive quarters.
– Operating margin fails to expand above the mid-teens (i.e., stalls near 15–16%) over two consecutive quarters, signaling the margin-leverage story is broken.
– The MI450/OpenAI first-gigawatt deployment slips materially beyond H2 2026, or MI400 yields disappoint.
– A close below $400 (the revised bear case) would trigger a technical and fundamental re-evaluation and a reduction toward a minimal position.
Next review date: approximately March 2027 (six months out), or immediately upon the next quarterly earnings report and any concrete MI400 ramp/OpenAI deployment update — whichever comes first.
One-sentence summary: For current holders, we recommend holding a reduced core position and taking partial profits into the $640 base and $780 bull targets, because the thesis has succeeded and the stock now trades at fair value with an inverted risk-reward — the discipline that made this trade work argues for harvesting, not chasing.
—
This content is general investment information provided to an indefinite/unspecified audience by a quasi-investment advisory business registered under Korea’s Financial Investment Services and Capital Markets Act, and is not personalized 1:1 investment advice tailored to any individual investor. This analysis is for informational purposes only and is not a solicitation to invest. All investment decisions and their consequences rest solely with the investor. The estimates and assumptions in this report are as of the writing date (2026-09-22) and may not materialize depending on market conditions and geopolitical variables. Financial data used reflects sources such as company filings and analyst consensus, and the scenarios and price targets represent the author’s conservative assessment. All investments carry the risk of principal loss, and past performance or analytical track record does not guarantee future results. As of the writing date, the author currently holds a position in this stock; this article is a review of an actual position. The author’s holdings and positions may change without prior notice depending on market conditions.
함께 읽으면 좋은 글
- AMETEK Record $4.11B Backlog and the Acquisition Engine: Why Analysts Keep Raising Targets Toward $282 (2026 Analysis)
- Target Turnaround Analysis 2026: Why 3.8% Comps and the $2B Fiddelke Reset Point to Margin Inflection Beyond the Tariff Refund
- Circle Internet (CRCL) Stock Analysis: The Arc Launch, $73B in USDC, and Why Rate Cuts Cap the Near-Term Upside at ~$100
- Sea Limited Three-Engine Flywheel: Why Monee’s 62% Loan Growth and Shopee’s $1B EBITDA Target Point to 56% Upside in 2026
- Amgen MariTide Obesity Phase 3 Readout Impact on Stock 2026: Why a $415 Base Case Rests on the $100B Weight-Loss Optionality