Latin America Holds the Minerals the World Wants 🌎⛏️

The region is already critical to global supply

🇧🇷Brazil dominates niobium

🇨🇱Chile leads in rhenium and lithium

🇵🇪 Peru is a major silver producer

🔸Chile and Peru are central to copper

🔸Brazil holds major graphite and rare earth resources

🇧🇴Bolivia stands out in antimony and tin

This matters because the next industrial cycle needs more

🔹 Copper for grids
🔹 Lithium for batteries
🔹 Silver for solar
🔹 Rare earths for motors and defence
🔹 Graphite for energy storage
🔹 Niobium for high-performance steel

Latin America is not just a commodity region

It is becoming one of the world’s most important strategic resource hubs

“US-Iran deal is expected to be signed”

Israel:

The USA is starting to lose the AI race in the only metric that matters now:

speed.

America has the talent, labs, capital, chips, and ambition to win. But it keeps wrapping the most important technology in human history in fear, paperwork, committees, and political theater.

Meanwhile, rivals are optimizing, deploying, copying, scaling, and moving.

You don’t win the AI race by asking permission from people who are terrified of the future.

Free the labs. Build the compute. Deploy the models. Let builders build.

Full acceleration is the only sane strategy.

🌐 THE FUTURE ECONOMY -> mapped
Forget the headlines. Here's the infrastructure the next decade actually runs on, and the names building each layer 🧵👇

💡 PHOTONICS - light replaces copper
$CRDO · $SIVE · $AAOI
The interconnect bottleneck of AI. Every GPU cluster moving from electrons to light.

🔷 GLASS & OPTICAL MATERIALS - the substrate light runs on
$GLW · $LPTH · $LPKF
The fiber that carries the signal, the infrared optics that shape it, the glass cores going into advanced packaging.

⚙️ AI INFRASTRUCTURE - the compute buildout
$IREN · $NBIS · $DGXX
Power, data centers, neoclouds. The physical backbone of the AI economy.

📡 CONNECTIVITY & RF - moving the signal
$AMPG
Re-shored 5G and the RF signal chain. The certified domestic alternative for 5G networks, SATCOM, and defense.

🤖 PHYSICAL AI - intelligence that acts
$OUST · $AMBA · $KRKNF
The eyes, the brain, and the body. AI moving into the real world.

🛸 AUTONOMY & DRONES - machines that operate themselves
$ONDS · $AUR · $MRLN
Aerial drones, driverless freight, autonomous flight. The re-shored defense and logistics layer.

⚡ ENERGY & STORAGE - powering the buildout
$TE · $EOSE · $SPKL
Domestic solar, long-duration grid storage, and short-duration data-center backup. The full duration stack.

🪨 CRITICAL MINERALS - the hard constraints
$MP · $UAMY · $ASPI
Rare earths, antimony, nuclear fuel. The materials China controls that the West must re-shore.

🚀 SPACE - the orbital economy
$ASTS · $RKLB · $PL
Direct-to-cell satellites and the launch backbone.
The thesis in one line: own the chokepoints and physical inputs the digital economy can't function without.

The screens get the attention. The infrastructure gets the returns.

🚨 The Top 20 future-focused ASYMMETRIC setups dominating the AI + Space + Nuclear + Autonomy chokepoint trade. 🔥

These aren't yesterday's winners.✅
These are tomorrow's compound machines.✅

Here's the full institutional stack:

🥇 1/ $SIVE / $SIVEF - The Photonics Chokepoint
Sole-source CW DFB lasers for ALL leading CPO winners (Ayar, Celestial, Lightmatter, $POET).
→ U.S.-Sweden TPD signed May 22
→ MSCI Sweden Small Cap inclusion May 29
→ Customers ramping production simultaneously 2026-2028
Architecturally unique. Multi-customer locked. Pricing power compounds.

🥈 2/ $IREN - AI Power + Compute Vertical Integration
$NVDA holds $2.1B warrants at $70 strike (above current price).
→ $9.7B Microsoft + $3.4B NVIDIA + $5.8B Dell signed
→ 5GW global power secured (NA + Europe + APAC)
→ Owns power AND data centers (vertical moat)
Market still calls it a "Bitcoin miner." Repricing imminent.

🥉 3/ $EOSE - Long-Duration Storage Bottleneck
Non-flammable zinc chemistry lithium can't replicate at hyperscale data centers.
→ Cerberus 32% holder + extended lock-up YE 2026
→ $24.3B pipeline (+56% YoY), $644.6M backlog
→ Frontier Power USA structure + $1.5B Ariel Green insurance
The ONLY genuine 🟢 in the entire 800V power trade.

⚛️ 4/ $IONQ - Quantum Institutional Anchor
Q1 revenue +755% YoY. RPO $470M (+554% YoY).
→ Missile Defense Agency SHIELD IDIQ contract
→ AWS + Azure + Google Cloud directly integrated
→ SkyWater acquisition = vertical chip integration
Defense + commercial + cloud all anchored simultaneously.

🚀 5/ $RKLB - Space Launch + Infrastructure Compound Machine
Backlog $2.2B (doubled YoY). Q1 revenue $200M record.
→ DoD $190M HASTE 20 hypersonic flights
→ Confidential customer: 5 Neutron + 3 Electron through 2029
→ Motiv Space Systems acquisition (Mars-proven robotics)
Trump Golden Dome tailwind. Vertical aerospace platform.

📡 6/ $ASTS - Space-Cellular Direct-to-Device Chokepoint
~3,900 patents + 1,100 MHz tunable MNO spectrum = irreplaceable IP.
→ 60 MNO partners covering 3B+ subscribers
→ $1.2B contracted revenue commitments
→ FCC commercial authorization received
Direct-to-device from space. Spectrum + IP combination impossible to replicate.

🔬 7/ $AAOI - Optical Transceiver Hyperscaler Anchor
Q1 revenue +51% YoY. Vertical integration laser→transceiver.
→ $200M single 1.6T transceiver order
→ $324M+ backlog with hyperscaler anchors
→ Mediacom DOCSIS 4.0 multi-year partnership
U.S. manufacturing scale (CHIPS Act aligned). 1.6T ramp H1 2027.

⚗️ 8/ $ASPI - Multi-Vertical Isotope Picks-and-Shovels
Proprietary Aerodynamic Separation Process tech across multiple verticals.
→ Fermi America 11 GW HyperGrid Texas
→ European reactor MOU May 11, 2026
→ HALEU + medicine + helium + LNG + quantum isotopes
Sells enrichment tech to whoever wins. Doesn't need to pick reactor winners.

☢️ 9/ $IMSR (Terrestrial Energy) - Nuclear "No HALEU" Advantage
Standard LEU fuel = supply chain MOAT vs every Generation IV competitor.
→ Riot Platforms MOU for data center co-located nuclear
→ 7.8 GW aggregate pipeline (~10 IMSR Plant projects)
→ NRC PIE Topical Report approved May 12
The reactor that's already solved fuel supply other reactors can't.

🛡️ 10/ $ONDS - Defense Drone Inflection
Q1 revenue +1,065% YoY. FY26 guide raised to $390M+.
→ $457M backlog stacking
→ Defense rail + autonomous drones dual exposure
→ DoD-aligned positioning at scale
Real revenue inflection at sub-$2B mcap.

🧠 11/ $PENG - AI Inference Memory Bottleneck
AI-driven biz >60% of H1 FY26 revenue (+50% YoY).
→ Integrated memory $308M (+53% YoY)
→ MemoryAI CXL-based KV cache servers deploying
→ Nokia exec David Heard joins board May 18
Memory is the inference bottleneck. PENG owns the layer.

🤖 Everyone's crowding into the AI software trade.

The asymmetric money is in the PHYSICAL layer - the chips, eyes, magnets, and motors that let AI act in the real world.

Robots can't run on vibes. Here's the stack that builds them - asymmetric setups up top, anchors at the base.

━━━━━━━━━━
⚡ THE ASYMMETRIC SETUPS
(small-cap chokepoints — multiple-X potential if the theme delivers)
━━━━━━━━━━

🧠 BRAIN - Edge Compute
$CEVA - IP royalty toll, 87% margins, 2.1B devices shipped
$AIP - on-chip "highway" (NoC); 90% margins, royalties +67%, founder owns ~32%
$AMBA - edge-AI vision SoC; Hanwha LTA $800M+ over 10 yrs in robotics & automation

👁️ EYES - Perception
$OUST - digital lidar, +49% rev, NVIDIA-qualified
$AEVA - FMCW 4D lidar, +90% rev; exclusive long-range supplier for Daimler's autonomous trucks

💪 MUSCLE - Magnets (the motor chokepoint)
$USAR - first U.S. NdFeB magnet production live, scaling to 1,200 MTPA by 2027. China controls ~90%

⛏️ FEEDSTOCK - Upstream Supply
$MP - domestic rare-earth mine-to-magnet; the raw input the chokepoint runs on

🦾 BODY - Embodiment
$KRKNF - the rare profitable robotics name (adj. EBITDA+); +35% rev, $615M Covelya deal closing Q2
$RR - service humanoids, Jetson Thor + Isaac Sim trained
$SERV - autonomous delivery, Uber + NVIDIA backed
$PDYN - embodied-AI software for industrial robots + drone swarms

🛸 AUTONOMY - Air & Ground
$ONDS - autonomous drones, $457M backlog
$RCAT - tactical defense UAS, NDAA-aligned
$AUR - autonomous trucking, driverless freight live in Texas

⚡ POWER - The Nervous System
$NVTS - GaN/SiC power chips; NVIDIA partnership, pipeline doubled to $2.4B. Efficiency is the constraint as robots scale.

━━━━━━━━━━
🏛️ THE ANCHORS
(own the theme durably - not asymmetric, but they can't lose the war)
━━━━━━━━━━
$NVDA - Isaac GR00T + Jetson Thor; the OS every robot trains on
$TSLA - Optimus; the only one with the scale to brute-force mass production
$TER - industrial automation + cobots
$QCOM - Dragonwing edge platform; the fleet's connectivity layer
$ARM - the IP tax inside nearly every edge chip
$POWI - established high-voltage power; the profitable power anchor
$ISRG - surgical robotics; the proven, profitable category leader

The asymmetric upside is in the chokepoints capital can't replicate - IP, perception, magnets. The anchors are how you own the theme while you wait.

The robots are coming. Own the parts inside them.

⚠️ Counter-thesis: most of the asymmetric layer is pre-profit, and the real revenue inflection is 2027+. Patient, multi-year bet - size accordingly.

🚨10 ASYMMETRIC NAMES BEFORE THEY GO LARGE-CAP 👇

The trillion-dollar AI names ran. The asymmetry is one tier down -> same chokepoints, still small enough to move.

🟠 $CIFR - bitcoin miner turned AI data center; $3B+ contracted with AWS, Google & Fluidstack
🔵 $ONDS - drone autonomy with real Army contracts and a fast-growing backlog
🔵 $PENG - AI memory + infrastructure; guidance raised, analyst targets climbing
🔵 $OUST - lidar perception; every autonomous machine needs eyes
🔵 $KRKNF - subsea defense robotics, the Anduril of the ocean floor
🟢 $EOSE - grid-scale zinc storage; domestic, non-lithium, contracted
🟢 $AMBA - edge AI vision chips as inference moves onto the device
🟢 $AMPX - silicon-anode batteries powering the drone & UAS buildout
🟢 $DGXX - owns the full stack: power → data center → GPUs; $35M Vera Rubin order
🟡 $SIVEF $SIVE - the lasers that move data inside AI clusters; early, real tech, real risk

The giants get the headlines. The runway lives one tier down.

Which is most underpriced?

The bears keep asking "what if AI demand slows?"

Jensen just answered with $40B of his own balance sheet.

You don't fund the whole supply chain for something you think is ending. 🔥

💡Everyone's crowding the obvious AI names. The real asymmetry is one layer down 👇

⚡ $DGXX - POWER
Just locked a $1.1B Cerebras deal (up to $2.5B over renewals) for a 40MW AI campus. And it OWNS the land, the substation, and the grid interconnection - vertical integration almost no micro-cap has. Debt-free with $100M+ liquidity, self-funding the buildout. AI can't run without power, and DGXX secured 200MW+ of it across the US.

🔬 $TRT - TEST
Every AI chip has to be burned-in and stress-tested before it ships — and TRT does exactly that. A $160M micro-cap that just landed ~$7.8M in orders for a next-gen AI GPU platform. 9-month revenue nearly DOUBLED ($26M→$48M) and it swung to an operating profit. New Malaysia facility coming online to scale capacity. Nobody's looking here.

☁️ $BRUN - COMPUTE
An NVIDIA Preferred Cloud Partner — and the rare neocloud that's FCF-positive. $940M already contracted with the majority in production, scaling to $375M+ in recurring revenue. 6 data centers live, 5 more coming, 125MW+ capacity. The shovel everyone's renting to build AI.

📡 $HLIT - the QUIET CRUSHER
Everyone writes it off as boring. Then it posted +43% broadband revenue, blew past its own guidance, and grew backlog 87% to $582M. Beat EPS by 44%, threw off $30M in free cash flow, and RAISED full-year guidance. Now selling its video unit for $145M to become a pure-play broadband compounder. Profitable, cheap, and accelerating.

Power. Test. Compute. Cash.
Everyone watches the giants. I watch the layer that feeds them. 👀

Which one's the sleeper?

The AI Buildout Has Twelve Floors. Most Investors Only See a Couple.

Most people think the AI buildout is just a couple of stocks – Memory, CPU, GPU, NeoClouds, Photonics..

$SNDK, $MU, $ARM $INTC, $AMD, $NBIS, $LITE, $AAOI, $TSEM, $SIVE

The reality is there’s so much more to it..

MY PORTFOLIO EXPLAINED
I have been meticulously building a portfolio based on the thesis that the Physical AI revolution is upon us.
I have chosen each of these companies, both for what they do and for providing asymmetrical return opportunity to investors.

$SIVE Sivers Semiconductors
Sivers makes high-frequency radio frequency (RF) millimeter-wave (mmWave) transceivers and advanced optical semiconductors through its Wireless and Photonics divisions. In the AI ecosystem, the company is a critical enabler of next-generation data centers, collaborating with foundry giants like GlobalFoundries to integrate its laser arrays into silicon photonics and co-packaged optics (CPO) platforms to eliminate interconnect bandwidth bottlenecks for massive workloads. Concurrently, its mmWave technologies supply high-speed wireless connectivity for satellite communications, 5G telecom, and edge infrastructure, providing the high-throughput network backbone necessary for real-time edge AI coordination and smart cities.  

$LPK LPKF Laser & Electronics
LPKF makes high-precision, laser-based micro-material manufacturing systems, headlined by its proprietary Laser Induced Deep Etching (LIDE) technology. As conventional organic substrates struggle with heavy thermal loads and warpage from massive chiplet integration, LPKF provides the precise tools needed to drill defect-free through-glass vias (TGVs) at aspect ratios up to 1:50 with zero micro-cracks. This specialized capability directly underpins advanced hardware and high-performance computing, enabling leading chipmakers to scale up substrate sizes, integrate high-bandwidth memory (HBM), and implement in-package microfluidic cooling for the next wave of physical AI chips.  

$ACCON Acconeer
Acconeer makes ultra-low-power, miniature 60 GHz Pulsed Coherent Radar (PCR) sensors that fuse millimeter-level depth precision with an exceptionally small spatial footprint. Within Edge AI and robotics, Acconeer’s chips allow tiny, battery-constrained hardware to execute real-time motion tracking, presence detection, and material or surface classification without relying on heavy cloud computation or invasive optical cameras. By deploying deep learning interfaces that compile models directly onto its integrated microcontrollers, the company provides robots, drones, and consumer devices with the localized spatial awareness required for autonomous navigation, gesture control, and power-efficient smart presence sensing.  

$SILC Silicom
Silicom designs and manufactures high-performance networking and data infrastructure hardware, including server network interface cards (NICs), programmable FPGA-based acceleration cards, SmartNICs, and edge switches. The company anchors Edge AI and smart systems by offloading heavy encryption, data compression, and packet-routing tasks from host CPUs, allowing decentralized AI infrastructure to maintain ultra-low latency and maximum data throughput. Operating primarily via design wins with cybersecurity firms, hyperscalers, and telcos, Silicom enables the secure, low-latency processing fabrics required for distributed AI inference networks and 5G/6G open radio access network (O-RAN) rollouts.  

$GAPW-B Gapwaves
Gapwaves develops and fabricates specialized, millimeter-wave antennas utilizing its patented gap-waveguide technology. This low-loss, cost-efficient hardware is engineered to maximize the resolution, beam direction, and overall range of high-frequency radar systems without increasing power consumption. Gapwaves fits squarely into the autonomous systems and smart city landscapes by supplying Tier-1 automotive partners with the structural building blocks for next-generation ADAS (Advanced Driver Assistance Systems), enabling autonomous vehicles, industrial transport robots, and urban traffic monitors to build reliable, high-definition models of their surroundings under any weather condition.

$SEYE Smart Eye
Smart Eye makes computer-vision hardware and software platforms dedicated to Human Insight AI, tracking eye gaze, facial expressions, and head movements in real time. The company’s primary footprint is in the automotive sector, where its edge-based Driver Monitoring Systems (DMS) and interior cabin sensing modules identify operator distraction, drowsiness, and cognitive load to comply with global safety mandates. This technology bridges the human-machine interface across smart transportation and autonomous vehicles, while its subsidiaries (like Affectiva) extend this technology into emotion AI, which is increasingly utilized in behavioral research, aerospace simulators, and interactive robotics.

$AMBA Ambarella
Ambarella builds ultra-low-power, high-efficiency system-on-chip (SoC) vision and radar processors built on its algorithm-first CVflow artificial intelligence architecture. Its chips, like the 4nm CV7 and the high-compute N1 series, serve as the localized brain for Edge AI, industrial robotics, and smart city infrastructure by running complex convolutional and transformer neural networks simultaneously. By integrating advanced image signal processing, hardware-accelerated video encoding, and native neural network computation on a single die, Ambarella allows security cameras, aerial drones, and autonomous mobile robots (AMRs) to execute edge-based multi-sensor perception, visual odometry, and real-time SLAM (Simultaneous Localization and Mapping) entirely disconnected from the cloud.

$OUST Ouster
Ouster designs and manufactures high-resolution 3D digital LiDAR sensors alongside integrated perception suites, following its strategic combination with Stereolabs' ZED camera technology. By embedding native color and precise spatial mapping directly onto its proprietary silicon architecture, Ouster creates an end-to-end sensing platform for Physical AI. This hardware acts as the primary sensory organ for humanoid robotics, warehouse automation, autonomous mining fleets, and smart city infrastructure (via its BlueCity traffic management software), delivering low-latency, resilient environmental data that allows machines to map, track, and safely navigate complex physical spaces.

$AMZN Amazon
Within the AI and robotics verticals, Amazon develops custom infrastructure silicon—such as its proprietary Trainium and Inferentia acceleration chips—while engineering vast, automated logistics systems for its global fulfillment network. The company is a pioneer in warehouse robotics, deploying fleets of fully autonomous mobile robots like Proteus and intelligent manipulation systems like Sparrow to streamline its supply chain via machine learning and reinforcement learning. On the edge, Amazon implements localized AI and natural language processing through consumer ecosystems like Alexa, Ring, and the Astro home robot, establishing a ubiquitous presence across cloud AI architecture, industrial automation, and smart home systems.

$ELMT Elmet Group
Elmet Group fabricates precision-engineered components and high-power radio frequency (RF) or microwave equipment, operating through its Critical Materials Components and Engineered Microwave Products divisions. Specializing in the custom manufacturing of refractory metals like pure tungsten and molybdenum, Elmet supplies the foundational, high-durability hardware components required for advanced semiconductor fabrication, aerospace systems, and defense electronics. It fits into the advanced hardware and AI supply chain by delivering the critical thermal management substrates, ion implantation shielding, and robust RF distribution modules necessary to manufacture and run high-energy compute chips.

$TRT Trio-Tech International
Trio-Tech provides semiconductor back-end testing services and manufactures heavy-duty reliability equipment, including customized burn-in boards, wet process stations, and environmental testing chambers.

The company serves as a vital gatekeeper for advanced hardware reliability, validating that complex chips can withstand severe thermal and electrical stress without failing prior to commercial rollout. Driven by the massive ramp-up in next-generation computing architectures, Trio-Tech directly supports AI and smart automotive expansion by stress-testing high-performance GPUs, CPUs, and electric vehicle (EV) electronics across its key facilities in Southeast Asia.

$HLIT Harmonic
Harmonic produces virtualized broadband access networks and software-based cloud video delivery infrastructure, known for solutions like its VOS360 cloud platform and cOS applications. The company embeds AI into smart systems by leveraging Content-Aware Encoding (CAE) algorithms and dynamic ad-insertion technologies to process mass media streams at the network edge, squeezing bandwidth requirements by up to 50% without compromising visual quality. This positions Harmonic as a fundamental utility provider for edge systems and smart cities, optimizing the heavy data pipelines that distribute live television, ultra-high-definition streaming, and localized broadband connectivity at scale.

🚨 The brutal truth about retail vs institutional investors:

❌ Sees a green candle. Chases the top.
❌ Watches the -15% pullback. Panics.
❌ Sells the bottom.
❌ Watches it rip back to new highs.
❌ Re-enters higher than they sold.
❌ Repeats the cycle. Blames the stock.

Institutional money? Different sport entirely.

Here's the 5-principle framework that separates compounding from chasing:

1️⃣ THESIS FIRST, PRICE SECOND
Anchor to WHY you own it. The daily candle is noise.

2️⃣ DEFINE YOUR KILL CONDITION
"What event proves me wrong?" If it doesn't happen, hold through anything.

3️⃣ THE ACCUMULATION PHASE IS VIOLENT
Chokepoint trades don't move in straight lines. Smart money uses dips. Retail panics on them.

4️⃣ POSITION FOR THE DRAWDOWN, NOT THE DREAM
Size for the 30% pullback you WILL experience. Conviction ≠ concentration risk.

5️⃣ READ THE CATALYST STACK, NOT THE CANDLE
The 10x compounders have stacking catalysts + institutional anchors + multi-year contracted visibility. The candle is the result. The catalyst is the cause.

The institutional difference isn't stock picking.

It's the discipline to think in 18-month cycles when everyone else is staring at 18-minute candles.

What's "hot" today rarely matters in 24 months.
What's structurally positioned today often does.

Save this. Read it next time you're tempted to chase a spike.

From AGI → ASI

#Google just mapped what comes after human-level AI.

there's a brilliant new paper from Google DeepMind called "From AGI to ASI," and it skips the fight everyone else is having.

almost every AI debate today is about reaching human-level intelligence. this one asks the harder question instead.

once we have an AI as capable as a person, where does it go next?

their reasoning is uncomfortable because it's so simple.

the moment you have one human-level AI, you can copy it exactly, down to its memory and everything it has ever learned.

so you don't end up with one digital worker. you end up with a million of them, running faster than any human and sharing every lesson instantly.

a group like that could outthink any company or research lab long before any single AI becomes a lone genius.

that's the heart of it. scale alone might carry us from "as smart as a person" to "smarter than entire teams of the best experts."

the paper maps four ways this could play out, and they can all happen at once.

keep scaling what already works. invent a new approach when the current one runs dry. let AI improve AI in a loop that feeds itself. or let armies of AIs organize into something smarter than any one of them, the way a company outperforms any single employee.

then comes the honest part, which is rare for a report like this.

they list everything that could slow it all down. running out of good training data. costs and energy becoming impossible to sustain. research simply getting harder as the easy wins disappear.

but the most interesting roadblock isn't a wall. it's a question.

today's AI learns from everything humans have already figured out, and it's brilliant at remixing our ideas. what it has never done is invent a genuinely new idea from scratch.

the test they borrow from Demis Hassabis says it best. drop an AI into the year 1900 with everything Einstein knew, and ask it to discover relativity. right now it can't, and something is clearly still missing.

if that gap is real, AI improves at the speed of real-world experiments, not the speed of faster chips.

here's the shift in thinking i keep coming back to.

the popular picture is one big moment where everything changes overnight. this paper argues for something different.

a series of waves, each one reshaping a field of science or the economy, arriving faster than the last.

and the most useful work right now isn't guessing when. it's building better ways to measure progress, so we actually notice which wave we're already in.

one last detail i loved.

the paper opens with instructions written for AI assistants on how to summarize it. the authors simply assumed you'd ask a model to read it for you.

Link to the paper in the next
arxiv.org/pdf/2606.12683

OpenBB: A free alternative to the $20,000 Bloomberg Terminal

Available 100% free on GitHub: github.com/OpenBB-finance/OpenBB

Bro: "I'm going to start trading It looks easy all I have to do is press buy and sell"

Me:

" i heard people made millions of dollars from trading "

me: