A pie is a portfolio within your portfolio. You set a target allocation — e.g. 40% AXTI, 30% AAOI, 30% LITE — and T212 automatically splits every deposit or dividend across your stocks in those exact proportions. It removes emotion from investing: instead of manually deciding how much to put into each stock each time, the pie does it for you.

Press "Load Pies" to pull your pies from T212

A bottleneck is the one constraint that limits how fast everything else can grow. In AI infrastructure, the bottleneck right now is moving data fast enough between chips. Copper cables can't keep up beyond short distances at the speeds modern AI requires — so everything is switching to light (optical fibre). The companies that make the materials, components and modules for this transition are the bottleneck plays. Supply is constrained. Demand is exploding. That's the setup.

Soitec doesn't make chips — they make the engineered silicon wafers the entire semiconductor industry builds on, using patented Smart Cut™ technology. The only qualified volume supplier for TSMC, GlobalFoundries and Tower Semi. 70%+ global market share in RF-SOI (inside every 5G phone). Now ramping Photonics-SOI for AI data centres (CPO wave 2026+). 4,300+ patents. Demand growing from AI, 5G/6G and EVs while supply is capacity-constrained.

⚠️ Listed on Euronext Paris — check if your broker can access European stocks before researching further.

AI models need to store and retrieve massive amounts of data at extreme speeds. The bottleneck is memory bandwidth — how fast data can move in and out of the chip. High Bandwidth Memory (HBM) is the premium product that solves this for AI training chips (NVIDIA H100, H200, B200 all require it). Only three companies in the world make HBM at scale: Micron, SK Hynix, Samsung. Supply is tight. Prices are rising. Every new AI chip generation needs more of it.

How to read this sector: Don't react to earnings — by then it's priced in. Track DRAM/NAND spot prices (Samsung/SK Hynix announcements) and hyperscaler capex guidance. Those move the stock weeks or months before results.

AI is running out of high-quality training data from the internet. The next frontier is real-world proprietary data — satellite imagery, sensor readings, physical world observations. The bottleneck is launch capacity and ground-truth data collection. Only a handful of companies can put satellites into orbit affordably. The companies that own the infrastructure to collect and sell that data at scale have a structural moat. Defence and intelligence agencies are also the biggest buyers — government contracts provide stable, recurring revenue.

PL photographs the entire Earth surface daily. As AI needs proprietary real-world training data, PlanetScope becomes increasingly valuable. Partnership with Anthropic. The Owl satellite (late 2026) will deliver ~1m resolution vs 3-5m today — a step-change in data quality. Rule of 40 achieved one year early.

Autonomous vehicles and robots need to see and understand the physical world in real time. LiDAR (Light Detection and Ranging) uses laser pulses to map the environment in 3D. The bottleneck is the laser itself — specifically Indium Phosphide (InP) lasers with ultra-low phase noise for FMCW (4D) LiDAR. There are very few companies that can produce these at the quality and volume required. The companies that solve this become foundational suppliers across all autonomous systems — trucks, cars, robots, drones.

Sivers makes ultra-stable, high-power InP lasers with low phase noise — exactly what FMCW 4D LiDAR requires. CEO confirmed they are included in "all of the customer's projects." The technical specs and delivery timeline align with Aeva's production ramp for Daimler Truck autonomous freight (2026/27) and NVIDIA Hyperion integration. If Sivers is the laser engine behind this, they're not a component supplier — they're a foundational partner across the entire sensor portfolio.