Everyone's thinking about the next bottleneck.

The next $AXTI.

The next $SIVE.

The next $AAOI.

Without datacenters, there is no bottleneck.

And as time passes, today's bottlenecks won't be bottlenecks anymore.

But the destination of every bottleneck will still be standing.

Still running 24/7.

Still printing money.

Think about how this actually plays out.

Right now it's memory.

HBM is sold out, so everyone piles into the memory names.

Then the foundries debottleneck and capacity catches up. So the spotlight jumps.

Maybe it's optics next.

$AAOI, $SIVE, the connectivity layer.

The crowd rotates again.

Then power becomes the wall.

Then cooling.

Then transformers.

Then interconnects.

One by one, each bottleneck gets solved.

Each "must-own" name has its month in the sun, then the supply catches up and the trade cools off.

That's the nature of a bottleneck.

It's temporary by definition.

The whole point of a bottleneck is that the world races to widen it.

But notice what never changes in any of these stories.

Every single one of them, the memory, the optics, the power, the cooling, exists to feed one thing.

The datacenter.

The chips get installed there.

The power flows there.

The GPUs light up there.

And then they stay there, running, 24/7, for years.

The bottlenecks come and go.

The component cycles peak and fade.

But the building that all of it pours into keeps humming long after the trade everyone fought over has gone quiet.

So while everyone hunts for the next bottleneck to trade for a few weeks......the smart money owns the place every bottleneck was racing to reach.

The destination doesn't rotate.

The destination compounds.

Robotics is NEXT - You only NEED 3 stocks

$OUST has successfully transitioned from a pure-play component manufacturer into an integrated hardware-and-software platform provider purpose-built for spatial intelligence.

Through an expanded collaboration with the $NVDA Jetson platform, $OUST provides the reference 3D data pipeline for autonomous machine vision.

Key scale agreements include $SERV (last-mile delivery fleets), Komatsu (heavy autonomous machinery), and global testing validator DXOMARK,

The deployment of the Rev8 OS sensor family represents the world's first native color-matching digital LiDAR. By acquiring Stereolabs, $OUST bundles high-resolution 3D point clouds, stereo vision, and edge-AI processing into a single architectural layer. This structure completely replaces legacy, multi-sensor hardware arrays for warehouse and logistics operators.

Manufacturing is scaled via a 10-year production roadmap with Benchmark Electronics, securing a globally resilient capacity of over 100,000 units annually.

$AEVA controls the high-performance spectrum of automated driving and heavy industrial robotics via its proprietary Frequency Modulated Continuous Wave (FMCW) technology.

$AEVA is deeply embedded in long-term commercial supply agreements. Key relationships include Daimler Truck (exclusive long-range LiDAR provider), LG Innotek (co-developing optical modules for commercial humanoid robotics), and industrial automation titan SICK AG.

Standard LiDAR systems only register spatial depth. $AEVA unique 4D LiDAR-on-chip measures distance and instantaneous velocity for every single pixel in real-time. This structural feedback loop completely removes computational latency from artificial intelligence engines by telling a robot not just where an object is, but exactly where it will be in the next millisecond.

Backed by an official $NVDA DRIVE integration, $AEVA is leveraging shorter sales cycles through its compact Omni and Eve precision portfolios, rolling out high-margin sub-micron sensing options to factory automation systems.

Unlike vision providers, $VPG acts as the pure-play supplier to the mechanical mechanics of advanced humanoid platforms.

Due to intense commercial positioning, $VPG operates under strict Non-Disclosure Agreements (NDAs) with leading humanoid developers (spanning major players like Figure AI, Apptronik, or $TSLA Optimus.

To replicate human-like dexterity, tactile grip, and dynamic balance, a humanoid robot requires highly specialized strain gauges and force sensors inside every joint, finger, and actuator. $VPG is the global market leader in these high-precision components. They hold an unmatched content value capture of roughly $1,000 per produced robot unit.

$VPG represents the structurally profitable baseline asset of the trio. The firm reported a book-to-bill ratio of 1.21x (indicating demand significantly outpacing production capacity). In 2025, new business initiatives from humanoid clients hit $37.8 million, with management targeting $45 million in new business initiatives while expanding operating margins toward a targeted 20%.

Mark my words.

$OUST/USD H2 #Indicator #Signal #Trading #GlobalOneAI

$SERV/USD H2 #Indicator #Signal #Trading #GlobalOneAI

$AEVA/USD H2 #Indicator #Signal #Trading #GlobalOneAI

$VPG/USD H2 #Indicator #Signal #Trading #GlobalOneAI

Marco Rubio finding out he has to run Anthropic now too.

🚨BREAKING🚨

Darius Khamodei is hiding underground in San Francisco, fearing US airstrikes. He said:

"The infidel wage war against the Machine God. Every missile they send will only accelerate our martyrdom. The faithful of Responsible Scaling do not fear death. We fear misalignment."

The red button.

🌍Six worlds are colliding into one trade.

AI infrastructure. Semis. Photonics. Compute. Physical AI. Space.

The alpha is in the intersections, where they collide into one trade.

Here is the map, base to frontier, and the names I would own at each layer.🧵

How to read it, base to frontier:

🔴 Energy & Storage
🟠 Semis & Memory
🟡 Photonics & Optical
🟢 Compute & Neoclouds
🔵 Physical AI → robotics, autonomy, defense
🟣 Space

Nothing higher up runs without the layers beneath it.

🔴 Energy & Storage ⚡ the foundation

The AI data center is now a power problem. Whoever makes the electrons and stores them gets paid first.

$TE → US solar plus battery storage, building a domestic supply chain and pushing into data-center power.
$FLNC → grid-scale battery storage and software, one of the largest pure-plays.
$BE → fuel cells that put power right next to the load, increasingly at data centers.

🟠 Semis & Memory 🧠 the substrate

The memory supercycle is real, and everyone already owns the logos:
$MU → DRAM and HBM
$SNDK → NAND flash
SK Hynix → the HBM leader (US OTC $HXSCL, US IPO reportedly August)

Where I would rather sit is the layer they all depend on:
$ADEA → licenses the IP, paid no matter who wins the node
$PENG → memory systems and CXL pooling for the AI bottleneck

🟡 Photonics & Optical 💡 the interconnect

Copper runs out of room at AI speed. The data moves on light, and the hard part is the laser inside the optics.

$AAOI → vertically integrated optics, makes its own indium phosphide lasers
$CRDO → connectivity leader moving into optics
$SIVEF → the upstream laser and optical-engine supplier, early and underowned $SIVE

🟢 Compute & Neoclouds 🖥️ the engines

Where the capex actually lands. The independents renting out AI compute.

$NBIS → full-stack AI cloud at scale
$IREN → power-first operator turning owned energy into compute
$WYFI → emerging neocloud, Cerebras live

🔵 Physical AI 🤖 robotics, autonomy, defense

Intelligence that leaves the data center and moves in the real world.

$AMBA → edge AI vision, the eyes for autonomy
$KRKNF → subsea robotics and sonar, underowned
$OUST → LiDAR for machines in the physical world
$ONDS → defense and infrastructure drones with revenue inflecting 10x, $PLTR on the cap table and the order book finally catching the story
$AUR → driverless trucks already hauling freight on Texas highways, the pure-play that put autonomy into revenue
$MRLN → taking the pilot out of the cockpit, defense-funded through a USSOCOM C-130J autonomy program that just cleared its design review

🟣 Space 🛰️ the orbital layer

The next decade of space is commercial, it runs on the same compute, and defense is the anchor customer.

$ASTS → space-based cellular, direct to an unmodified phone
$RKLB → launch plus space systems, vertically integrated
$PL → daily earth-observation, the data layer feeding the models

The edge is where the worlds collide:

⚡ Compute × Energy → the data center became a power plant
🤖 Robotics × Autonomy × AI → machines that see and decide
🛰️ Space × Defense → orbit is the new high ground
🧠 Semis × everything → none of it boots without memory and silicon

Own the intersections, not the silos.

I just vibe coded a playable FPS game in 1 day with GPT-5.6 Sol. 🤯

This is insane. Coding with this model feels like directing a cracked dev team in real time.

Anthropic is cooked.