If you want to invest in AI with clarity, you need to see the entire system end to end
$POET $NVDA $GOOG $GFS $AMAT

I mapped the entire 12-Layer stack

Every bottleneck, every physics wall, every place value accrues.

Here’s the full system 🧵 Layer 1 - The Blueprints (EDA / Chip Design Software)
$SNPS $CDNS

This is where every chip begins.

Before a wafer ever enters a fab, these tools simulate how electrons will behave at near-atomic scales.

They solve physics problems digitally long before billions are spent physically.

Why it matters - AI chips fail without accurate design. This layer shapes the limits of every processor built above it.

Investor angle: Recurring, unavoidable software revenue tied directly to each generation of silicon.
(Report posted on SS)

The speculative party is over. Time to find the winners.

This thread maps the 12 physical bottlenecks that will define capital allocation for the next decade.

Welcome to the reality of physics. $NVDA $NBIS $AMD Pillar I represents the Silicon Foundation.
$ASML $TSM

This layer is defined by extreme physics and monopolistic choke points.

It is one of the most complex manufacturing challenges in history.

Value accrues to those who master the atomic precision required for the next decade of compute.

I’m tracking two key players in the photonics space, $ALMU and $POET, both building the future of high-speed connectivity.

Today, let's do a deep dive on Aeluma, a semiconductor company creating a platform that could redefine AI, autonomous driving, and quantum computing

You won't want to miss this thread 🧵 Cracking the Semiconductor Code $ALMU

For years, chip makers had a fundamental problem.

You could have incredible performance from exotic compound semiconductor materials, or you could have the massive scale and low cost of silicon manufacturing.

You couldn't have both.

Aeluma has cracked the code, developing a proprietary way to grow high-performance compound semiconductor materials directly onto large-diameter silicon wafers.

This platform gets the elite performance of specialized materials and the cost structure of mass-market production, a true holy grail for the industry.

This breakthrough means they can leverage the multi-trillion-dollar global silicon manufacturing infrastructure to build chips that were previously too expensive for anything but niche applications.

Need new stock ideas?

Here is a list of 24 of my favorite names that I will be watching/buying during any down turn

I'll give a quick thesis on the investment and a chart with prices I'm watching

Let's dive in!

Disclaimer: None of this is financial advice or a recommendation. Do your own research

$SPY $QQQ $IWM 1/24 Innodata Inc. $INOD

Thesis: Innodata is a pure-play on the critical need for high-quality data in the Generative AI era. The performance and reliability of any Large Language Model are directly tied to the quality of the data it is trained on.

Innodata provides these essential data engineering services, positioning itself as a key partner for enterprises looking to build and deploy accurate, proprietary AI models. The company is a direct beneficiary of the massive, ongoing investment in the AI data pipeline.

Levels to Watch: The chart shows a stock that has just emerged from a year-long consolidation phase with a massive surge in both price and volume, indicating a significant change in character.

Buy Area: ~$70.60 (This represents the key breakout point from the prior trading range, which could now act as a new support shelf).

Support: ~$55.50 (The major resistance level from earlier in the year).

Resistance: ~$91.70 (The recent high).

Most data centers take years to build.

WhiteFiber $WYFI built one in months...
Out of a mattress factory.

That’s its edge: retrofits that are 2x faster and 40% cheaper.

Now it’s running a $90M GPU cloud and developing a 99MW NC site.

This thread unpacks the strategy, economics, comps, and risks 🧵

$CRWV $IREN $NBIS $WULF $SLNH Strategy $WYFI

At its heart WhiteFiber runs a two-pronged strategy to capture value across the AI infrastructure stack.

First, its Colocation business acts as a specialized landlord, providing secure, power-dense facilities for customers like AI hardware firm Cerebras Systems to deploy their own servers.

Second, its higher-margin Cloud Services (GPUaaS) segment offers direct, on-demand access to high-performance computing on its own fleet of NVIDIA GPUs.

The linchpin is its "retrofit" model. Instead of building data centers from the ground up, $WYFI acquires and rapidly upgrades existing industrial sites.

Management claims this approach is "two times faster and 40% cheaper," a critical edge in a market where speed to deployment is paramount.

This strategy allows them to target buildout costs of $7-$9 million per megawatt, a significant potential saving over traditional greenfield projects.

Why can one company make $7 Million from a megawatt of power, while another makes just $2 Million?

It's the most important question for valuing AI infrastructure stocks like $WYFI, $IREN, and $RIOT. They are not the same business.

This thread breaks down the one metric that matters: revenue per MW. Here's how to tell the landlord from the manufacturer. 🧵

#WYFI #IREN #RIOT #AI #Investing #DataCenter The Foundational Model - Pure-Play Mining $RIOT

Riot Platforms represents the large-scale, pure-play Bitcoin mining model.

Their business is a direct power-to-digital-asset conversion. They secure power infrastructure for the primary purpose of running their own mining fleet to earn Bitcoin.

Revenue Model: An industrial-scale digital asset production and energy arbitrage play.

Annual Revenue per MW: ~$1.5M - $2.5M.

This figure is directly calculable from the Bitcoin network's hash rate economics (hash price) and a company's reported power capacity and fleet efficiency.

The revenue is almost entirely dependent on the price of Bitcoin and their operational uptime.
#RIOT #Bitcoin #Mining #Energy

What if you could invest in the AI revolution but with a built-in safety net?

Smith-Midland $SMID is a key builder of vaults for AI data centers, but its less glamorous businesses, like making highway barriers, provide a powerful foundation.

This is a story high growth potential with a security blanket of diversification. 🧵

$NVDA $AMD $NBIS $CRWV Think of $SMID as a three-legged stool.

Leg 1: AI Data Centers. They build the essential precast utility vaults that power the AI boom.

Leg 2: Public Infrastructure. They make the J-J Hooks barriers and Soundwalls lining America's highways.

Leg 3: Architectural. Their innovative SlenderWall panels are used on modern high-rises.

Each leg supports the company, providing a powerful and balanced business model.

Upstart $UPST was left for dead in 2022–23 when rates spiked and funding dried up.

Fast forward to 2025: loan volumes are surging, the company is profitable again, and the Fed may soon cut rates.

Here’s why Upstart’s AI-driven lending model could thrive over the next 1–3 years 👇 2/ $UPST
Upstart is not a bank. It’s a marketplace.

Borrowers come to its site, AI algorithms assess risk in real time, and loans are funded by 100+ banks, credit unions, and institutional partners.

The pitch: AI can separate risk 5× better than FICO, approving more borrowers at better rates.

$UPST

Everyone’s eyes are on $UNH, but Elevance Health $ELV may be the most overlooked major player in managed care. Serving ~46M members and operating the largest Blue Cross/Blue Shield footprint, ELV is building an ecosystem that spans insurance, pharmacy, home health, and behavioral care.

The story here is about scale + services, not just insurance margins.

Let's dive in 🧵 2/
Where $UNH has Optum, $ELV has Carelon—a services arm generating ~$54B annually and growing at double-digit rates. Carelon spans specialty pharmacy, behavioral health, and home/community care. In Q1’25 alone, Carelon revenue surged 38% YoY.

This diversification makes ELV far more than “just another insurer”—it’s becoming a vertically integrated healthcare platform.