Aragorn⟶Investment Thesis · Q2 2026
Live · Q2 2026 View source
Disclaimer

Not financial advice. Personal research shared with friends. I hold every position listed. Past performance means nothing. Do your own homework. For new positions above €5,000 consult an AFM-licensed advisor. No actual position sizes or holdings values are disclosed on this page — every euro figure you see is generated by your own input into the simulator.

Q2 2026 · Updated April 27

One worldview. Seven themes. Seventeen bets on the next decade.

If technology follows exponential curves and the prediction establishment is structurally blind to them, the highest-leverage investments live in the unsexy infrastructure layer that the consensus is still mispricing as if change were linear. Silicon photonics. AI compute. Power delivery. Custom silicon. Embodied AI. The thesis is the worldview, applied to a brokerage account.

Worldview
1
six core theses drive every position
Themes
7
photonics · infra · embodied · semis · software · platform · frontier
Positions
17
+ 7 new picks for Q2 2026
Holdings €
—
withheld; simulate at your own scale

01Composition by Theme

7 themes · 17 positions

The portfolio is concentrated by design. One-third sits in silicon photonics — the highest-conviction thesis. The thesis is either correct, in which case concentration is the feature; or it is wrong, in which case concentration is the bug. Diversification theater is not a substitute for being right about the curve.

34% Silicon Photonics. The single highest-conviction position group, anchoring the entire AI-infrastructure thesis. Optical interconnect is the bandwidth bottleneck and the inevitable winner.

Theme Tickers Weight
Silicon PhotonicsPOET · LITE · LASR · AAOI34%
AI InfrastructureAPLD · VRT · NVDA18%
Embodied AI / RoboticsTSLA · SYM · CGNX18%
Specialty SemisAMAT · TSEM · MRAM9%
AI Software / DefensePLTR9%
Platform AIGOOGL9%
Frontier (Quantum / Biotech)IONQ · ADPT3%

02Simulate at Your Own Scale

Your numbers, never mine

No actual position sizes are published anywhere on this page. Type a portfolio total below and the table resolves these allocation weights into concrete euros for your capital. Numbers stay in your browser; nothing is sent anywhere.

Theme Tickers Weight Your €
Enter a portfolio size above to simulate.

Allocations follow the actual weights of my current portfolio. They are not a recommendation for your portfolio. Your situation, time horizon, risk tolerance, and tax jurisdiction are not mine. Treat the output as a thinking aid, not advice.

03All 17 Holdings

Tickers & themes

The full list, grouped by theme. Position sizes and individual returns withheld; this is research, not a brokerage statement.

Full portfolio holdings as of April 2026
# Ticker Company Theme
01POETPOET TechnologiesSilicon Photonics
02LITELumentumSilicon Photonics
03LASRnLIGHTSilicon Photonics
04AAOIApplied OptoelectronicsSilicon Photonics
05APLDApplied DigitalAI Infrastructure
06VRTVertivAI Infrastructure
07NVDANVIDIAAI Infrastructure
08AMATApplied MaterialsSpecialty Semis
09TSEMTower SemiconductorSpecialty Semis
10MRAMEverspin TechnologiesSpecialty Semis
11TSLATeslaEmbodied AI / Robotics
12SYMSymboticEmbodied AI / Robotics
13CGNXCognexEmbodied AI / Robotics
14PLTRPalantirAI Software / Defense
15GOOGLAlphabetPlatform AI
16IONQIonQFrontier (Quantum)
17ADPTAdaptive BiotechnologiesFrontier (Biotech)

Snapshot date: April 27, 2026. Tickers grouped by theme. Position sizes and individual returns withheld; section 02 lets you simulate at your own scale.

04Q2 2026 Picks

7 positions · DCA over months, not days

These fill the gaps in the portfolio. Three are highest-conviction buys; four sized smaller alongside. Sizing tier reflects relative conviction, not euros — the simulator above resolves both into your scale.

>>Top 3 · highest conviction

Buy Re-entry NASDAQ
~$482

ISRG

Intuitive Surgical

AI-accelerated medicine is moving from prediction to reality. Intuitive Surgical is where AI meets physical intervention in the human body. 11,395 surgical systems worldwide, 80%+ recurring revenue, Q1 2026 revenue up 23% with a 7% stock jump post-earnings. The da Vinci platform has the installed base, the proprioceptive data from millions of procedures, and the recurring revenue model to fund AI integration at scale. Ion lung biopsy procedures up 39% in Q1. When world models reach surgical precision, ISRG already has the physical layer.

Conviction Tier 1 · ~6-8% allocation · DCA 3-4 months

Gap filled:Medical robotics (zero exposure today)

Risk factors
  • 45x NTM earnings is premium; multiple compression risk in risk-off markets
  • Competition from CMR Surgical, Medtronic Hugo
  • US healthcare reimbursement policy changes
  • Long hospital capex sales cycles
Buy NYSE
~$122

CCJ

Cameco Corporation

Data center electricity demand nearly triples by 2030 (IEA: 415 TWh to 945-1,580 TWh). Nuclear is the only source delivering baseload reliability at that scale. Cameco is the world's largest publicly listed uranium producer at roughly 15% of global primary supply. A $2.6B long-term agreement with India's nuclear program just signed. And here's what nobody talks about: Russia supplies 40% of Western uranium enrichment. Every tightening sanction accelerates Cameco's contract book. The single best hedge against AI energy demand and Middle Eastern disruption simultaneously.

Conviction Tier 1 · ~6-8% allocation · 5-10 year hold

Gap filled:Nuclear / power generation (zero exposure today)

Risk factors
  • Uranium spot price volatility (~$70-75/lb)
  • Mine operational issues (Cigar Lake, McArthur River)
  • New large-scale nuclear incident could collapse political support
  • Kazakhstan production risk via Kazatomprom joint venture
Geopolitical hedge layer — three scenarios CCJ wins on
  1. Strait of Hormuz closure. Qatar (#2 helium producer, top-3 LNG exporter) ships through the Strait. Sustained closure spikes European energy prices, accelerates EU nuclear restarts, raises Western decarbonization capex. Uranium does not transit the Strait — mined in Canada and Kazakhstan. CCJ is one of the few energy investments positively correlated with Middle Eastern disruption.
  2. Russia uranium-enrichment sanctions tighten. Russia supplies ~40% of Western uranium enrichment. Each escalation shifts Western utility contracts toward Cameco. The contract book deepens with every sanctions ratchet.
  3. Iran-Israel sustained conflict. Gulf petrodollar recycling into US AI capex (~$300B) gets redirected to domestic survival (desalination, food security, defense). Deflationary for discretionary tech capex. Cameco is structurally insulated — customers are utilities on 10-20 year contracts, not GPU buyers.

CCJ is simultaneously the best thematic AI-power play AND the single most powerful geopolitical hedge in this portfolio.

Buy NYSE
~$424

ETN

Eaton Corporation

Every data center needs two things: compute and power delivery. The media covers GPUs and cooling. Nobody covers what sits between the grid and the rack: switchgear, power distribution, transformers, power management. Eaton owns that invisible layer. Electrical Americas growing 10% organically, driven entirely by data center demand. The detail nobody else mentions: Eaton is co-engineering 800V DC power architecture directly with NVIDIA. That's joint R&D, not a supplier relationship. A $9.5B Boyd thermal acquisition closes next year at 70% growth. Q1 earnings on May 5 are the next catalyst.

Conviction Tier 1 · ~6-7% allocation · DCA before/after May 5

Gap filled:Electrical infrastructure / power delivery (zero exposure)

Risk factors
  • ~35x NTM earnings, premium for an industrial
  • Boyd's $9.5B acquisition integration risk
  • Tariff exposure on electrical components
  • Economic slowdown could defer data center capex

>>4 more · sized smaller

Buy NASDAQ
~$95

MRVL

Marvell Technology

NVIDIA made a $2B strategic investment in Marvell on March 31, integrating their custom silicon into the NVLink Fusion ecosystem. Google is in active negotiations for two specialized AI chips. FY2026 revenue hit $8.2B (+42% YoY), guiding FY2027 above $11B. Custom XPU is how hyperscalers escape NVIDIA pricing constraints. Marvell builds those weapons. Stock up 30% in April alone; analyst PT at $121 implies the partnerships aren't fully priced. Average in over three months rather than chasing the run.

Conviction Tier 2 · ~4-5% allocation · 3-month average
Risks
  • Competition from Broadcom for custom chip mandates
  • Google chip development still in negotiations, not contracted
  • Stock has already run 50% YTD
Buy Re-entry NASDAQ
~$380-393

AVGO

Broadcom

CEO Hock Tan projected $100B+ in XPU revenue in FY2027 alone, more than 1.5x Broadcom's entire FY2025 revenue. Q1 AI semiconductor revenue came in at $8.4B (+106% YoY). Custom silicon partnerships with Google, Apple, Meta (extended through 2029), and NVIDIA. Broadcom is building weapons for every side of the AI war simultaneously. The safer play alongside MRVL: more diversified customer base, lower volatility. Size both as one custom-silicon allocation.

Conviction Tier 2 · ~4-5% allocation · pair with MRVL
Risks
  • Google represents ~50% of AI revenue, single-customer concentration risk
  • Antitrust scrutiny in networking ASICs
  • VMware integration complexity (acquired 2023)
Buy Re-entry NASDAQ
~$400-450

CRWD

CrowdStrike

Every new AI agent is a new endpoint. Every API call is a new vector. Cybersecurity attack surface expands geometrically with AI proliferation. CrowdStrike crossed $5B ARR (+24% YoY) with record net new ARR of $330.7M in Q4, an acceleration less than two years after the July 2024 outage. The resilience test was passed. Purple AI at 40% attach rate. The CEO calls it "mission-critical infrastructure: securing AI from GPU to agent to prompt." Watch for corrections toward $350-370 to build a position.

Conviction Tier 2 · ~4-5% allocation · target dips toward $350
Risks
  • Residual trust deficit from 2024 outage (largely recovered)
  • Competition from Microsoft Security, Palo Alto Networks
  • ~50-60x forward earnings is premium
Buy NYSE ADR
~$131

BABA

Alibaba Group

China is a co-equal AI force. DeepSeek proved it in January. Alibaba's Qwen model family is the most downloaded open-source model on HuggingFace globally. AI investments nearly doubled quarter-over-quarter to $2.9B. Cloud revenue forecast at 40% YoY growth. The stock sits 31% below analyst consensus ($190-230) because of tariff panic that has already started reversing (+14% in April alone). At 10-12x forward P/E versus US tech at 25-40x, this is the asymmetric gap the worldview explicitly identifies. VIE structure is the real risk. Size accordingly, the highest-asymmetry position in this set.

Conviction Tier 2 · ~3-4% allocation · highest asymmetry
Risks
  • VIE structure: no legal claim on mainland Chinese assets
  • US ADR delisting risk (SEC/PCAOB ongoing tension)
  • Taiwan conflict scenario would be catastrophic
  • Xi regulatory intervention (2021-style crackdown could repeat)

05Watch List

Not buying yet

Too early, too rich, or waiting for a specific catalyst.

Watch CEG

Constellation Energy · NASDAQ · ~$275

The cleaner nuclear pure-play entry right now. Largest US nuclear operator, Microsoft 20-year PPA reactivating Three Mile Island, Meta and Amazon dataset deals in negotiation. Higher-purity nuclear thesis than VST (which mixes nuclear + gas + solar). If CEG runs away on a nuclear-renaissance narrative, VST is the secondary entry. At this price, CEG is the cleaner expression of the AI-energy thesis.

Watch Re-entry candidate VST

Vistra Corp · NYSE · ~$157

Second-largest US nuclear operator. Amazon and Meta PPAs signed. Down 25% from 52-week high. Morgan Stanley maintains Overweight with $208 PT (April 21), 33% upside. Wait for consolidation. Constellation Energy is the cleaner nuclear pure-play entry right now; Vistra is the backup if CEG runs.

Watch S

SentinelOne · NYSE · ~$13

AI-native cybersecurity underdog. Down 45% from 12-month highs with 35% analyst upside consensus. CrowdStrike has the brand, the installed base, and the momentum. Watch S as a higher-upside spec if the whole sector re-rates broadly. Not ahead of CRWD.

Watch SYM

Symbotic · NASDAQ · ~$63

Already a portfolio position. $22.5B backlog, 26% YoY revenue growth, expanding into healthcare and e-commerce. Not yet profitable. Add on dips rather than as a new position. The humanoid pure-play IPOs (Figure, Physical Intelligence) may offer better robotics entries in 2027.

Watch BIDU

Baidu · NASDAQ · ~$85

The China robotaxi play. Apollo Go at 3.4M fully autonomous trips across 26 cities globally, with Uber partnerships in Abu Dhabi and Dubai. First-ever dividend plus $5B buyback. The Wuhan system failure (100+ robotaxis freezing, April 2026) is the operational risk shadowing autonomous-driving investments. Watch for stability through 2026 before adding as a second China position alongside BABA.

Watch Spin-off Q2 2026 ABBN.SW

ABB Robotics (spin-off) · SIX Swiss Exchange · CHF 75 (parent ATH)

ABB's robotics division is spinning off onto Swiss/Swedish exchanges in Q2 2026. Potentially the most direct pure-play humanoid/industrial robotics stock available to European investors outside of Tesla Optimus — and EUR-denominated, which addresses the USD-concentration question explicitly raised in the research. Track the IPO date and initial valuation carefully; this could be the highest-conviction humanoid/industrial robotics position available without buying the full ABB parent. Current entry: wait for the spin-off price discovery.

06Avoiding

Looks thematic, isn't

Three names that look thematic on the surface. Not buying.

Avoid ETFs

Pure-play sector ETFs

KWEB holds Pinduoduo alongside Baidu. ROBO holds Fanuc alongside ISRG. ETFs dilute conviction into generic exposure. The precise bets the worldview demands cannot be made through baskets. A stock picker with specific theses, and basket products are the antithesis of that approach.

Avoid INTC

Intel · NASDAQ

Looks thematic. Structurally a loser in every scenario the worldview describes. The Compression Singularity benefits ARM architectures, not x86. Gaudi has failed to gain meaningful hyperscaler traction. The foundry strategy is five years behind TSMC with no clear path to close. Intel is the stock the broken-prediction establishment is still recommending. That's the tell.

Avoid MSFT

Microsoft · NASDAQ

Thematically perfect. Priced perfectly too. At 30x+ forward earnings with every AI tailwind already visible and priced in, the risk-reward for a new position is poor. Already holding Alphabet as the platform-AI bet. If MSFT corrects 20%+ on an Azure growth miss, it gets interesting. Not at current levels.

07Worldview → Portfolio

Six theses, one portfolio

Every position above descends from a thesis I have argued in writing for years. The portfolio is the worldview, made concrete.

THESIS 01

Technology follows exponential curves; the prediction establishment is structurally blind

McKinsey, Gartner, the IEA, and Morgan Stanley use linear extrapolation models on exponential phenomena and are typically wrong by one to two orders of magnitude. The opportunity is permanent: any market priced on consensus forecasts mispriced the curve. Portfolio expression: overweight infrastructure where capacity assumptions are still linear. POET, LITE, APLD, NVDA are direct bets on this gap.

THESIS 02

The infrastructure layer is always underpriced relative to the application layer

Everyone wants the AI app layer. Few can stomach silicon photonics, uranium, switchgear, surgical robots. Those are the picks nobody writes about until they have already tripled. Portfolio expression: 34% silicon photonics, 18% AI infrastructure, plus the Q2 picks ETN (power delivery) and CCJ (uranium) — explicit moves into the unsexy layer.

THESIS 03

Concentration is only a risk if the thesis is wrong

One-third of the portfolio sits in silicon photonics. That is intentional. If AI interconnect demand plays out as the worldview predicts, the concentration is the feature; if not, it is the bug. Diversification theater is not a substitute for being right about the curve. Portfolio expression: the 34% photonics weight itself is the thesis.

THESIS 04

AI accelerates energy demand, it does not reduce it (Jevons Paradox)

Cheaper inference does not mean less compute; it means more of it. IEA projects data center electricity demand from 415 TWh today to 945-1,580 TWh by 2030. Nuclear is the only baseload source that scales. Power delivery becomes the bottleneck after generation. Portfolio expression: CCJ (uranium), VST (nuclear, on watch), ETN (power delivery), APLD (data center capacity), VRT (cooling/distribution).

THESIS 05

Geopolitical disruption is a tailwind, not a headwind

Russian uranium enrichment dependency, Strait of Hormuz fragility, China decoupling — the portfolio is positioned to benefit from the specific disruptions the worldview considers most likely, not insulated against them. Portfolio expression: CCJ benefits as Russian sanctions tighten; BABA benefits as China-US tariff panic mean-reverts; embodied AI (TSLA, SYM, CGNX) benefits as labor-cost geopolitics force reshoring.

THESIS 06

Embodied AI and the humanoid form factor are the next platform shift

The 2020s are the watershed decade for humanoid robotics. The human form factor is engineering optimization for an environment built around it — not anthropocentric nostalgia. Cost collapse follows the solar curve (~90% per generation). Portfolio expression: 18% embodied AI weight (TSLA + SYM + CGNX), the planned 2027 entry into pure-play humanoid IPOs (Figure, Physical Intelligence), and the medical-robotics flank via ISRG.

08What Could Break the Thesis

Honest pre-mortem

A thesis that cannot articulate its bear case is not a thesis; it is a wish. The portfolio's biggest concentration is also its biggest single-point-of-failure risk.

BEAR 01

Silicon photonics adoption stalls past 2027

If hyperscaler optical interconnect deployments slip materially past 2027, the 34% photonics concentration becomes a 5-year drag. POET in particular is a small-cap with execution risk.

BEAR 02

Data center capex cycle peaks in 2026

The infrastructure thesis assumes the capex curve keeps climbing through 2030. If hyperscalers signal saturation in the back half of 2026 (NVIDIA Q3 guide is the canary), VRT, APLD, and ETN all re-rate together.

BEAR 03

Nuclear politics regress on a single incident

The Cameco/VST thesis is fragile against any major Western nuclear incident. Political support for new build collapses on impact and takes a decade to rebuild.

BEAR 04

China decoupling escalates beyond mean-reversion

The BABA thesis assumes a 31% discount to consensus is tariff-panic noise. If decoupling becomes a hard split — VIE delisting, mainland confiscation risk, Taiwan escalation — the position is not undervalued, it is unowned.

BEAR 05

Embodied AI takes another decade, not another five years

The humanoid form-factor thesis assumes economic-scale deployment by 2030. If the cost-collapse curve is closer to ten years than five, TSLA, SYM, and the 2027 IPO basket all underperform during the wait.

BEAR 06

AI capability plateau before alignment becomes a market

The PLTR / CRWD / Q2 alignment-adjacent picks assume continuous capability progression. If foundation-model progress plateaus in 2026-2027, the security-and-defense overlay becomes less urgent and the multiples compress.

09Sources

Show your work

Every claim above is anchored to publicly-available reporting. Click through to verify. All prices verified April 2026.