Welcome to the first ever edition of Expected Value. Thank you for your readership!
Every week, I read headlines screaming “AI is taking X amount of jobs” and “AI is making XYZ company irrelevant”.
In this week’s edition, we learn what’s actually driving the hype, and why this is nowhere near the aggressive predictions some CEOs out there are giving.
CEOs are talking their books to keep the capital inflow which is the lifeline of many AI startups. Hate the game, not the players.
This week:
Why is the best performing AI stock this year down 23% this week (it's not the tariffs)
Why is your job safe from AI: My contrarian bet on AGI (artificial general intelligence) timeline
How happens to the electric grid when you ask ChatGPT a question
Let’s start with a 30,000 feet view on the market…
Follow the Money Dashboard
Expected Value: Weekly Dashboard
🚀 Market Move: CoreWeave Down Bad
Coreweave down 23% last week far below Indexes
Brief: CoreWeave, an AI infrastructure company, saw its stock crash 23-33% this week after reporting bigger losses than expected despite strong revenue growth. The timing aggravated this as many insiders could sell their shares for the first time since the company went public. As a smaller, newer company without the stability of being in major stock indexes, CoreWeave faces more volatile trading.
Breakdown:
Earnings Performance: Revenue jumped 207% to $1.21 billion, beating expectations, but losses of 27 cents per share were worse than the 21 cents analysts predicted
Lockup Expiration: Over 80% of company shares became available for trading on August 15, causing major selling pressure as insiders sold over $1 billion in stock
Smaller Cap Volatility: As a newer public company worth $49 billion, CoreWeave lacks the stability of established tech giants and doesn't benefit from automatic buying by S&P 500 index funds
Valuation Concerns: Trading at 15.69x forward sales, which is over 400% higher than typical companies in its sector, making it expensive even after the decline
Market Structure: Less than 15% of shares were previously available for trading compared to 95% for most S&P 500 companies, creating an unstable trading environment
Why it's important: CoreWeave needs to convert its $30.1 billion sales backlog into actual revenue fast enough to justify its high valuation, requiring sustained execution in a competitive AI infrastructure market. The company's success depends on an extremely concentrated list of major contracts. In 2024, 2 customers accounted for 77% of total revenue. Any slowdown in AI spending from Microsoft for example could disproportionately hurt CoreWeave.
Now if you’re looking for an AI stock that’s rebounding…
🎲 Degen Corner
SoundHound AI rebounding over the last 3 months
$SOUN (SoundHound AI) • Market cap: $2.1B • Stock up 18.47% compared to the previous week and showed a 35.21% rise for the month.
The bet: SoundHound AI to grow its already substantial enterprise client base across multiple industries (Fortune 500 companies) through both direct partnerships and strategic acquisitions.
Catalyst: The Amelia acquisition was particularly significant because it transformed SoundHound from primarily an automotive and restaurant-focused company into a comprehensive enterprise AI provider with ~200 marquee enterprise customers, contracts with top 15 global banks, and strong diversification.
Risk: Despite positive momentum, larger tech companies could replicate SoundHound's capabilities and bundle them into existing platforms
Not investment advice. Speculation is gambling.
💡 One Big Idea: AI Won’t Take Your Job… For Now
Sectors with Employment Downtrend
The “AI is displacing our Jobs” narrative is Backwards
Everyone's freaking out about AI taking over jobs. The narrative of AI taking over entry level jobs is being pushed. In reality, here's what they're missing:
Most jobs, especially ones with apprenticeship models, ones that require nuance and where the context requires to pick up, and years to master, are safe from LLMs.
In fact, LLMs have crippling limitations:
Incapable of Inventing New Things and Making New Connections: LLMs are primarily designed for retrieval and paraphrasing knowledge that they have been fed already
Cannot Explore Novel Ideas or solve “open problems”: Current LLMs are unable to challenge common knowledge unless prompted. Even then, they will not push the boundary of what they know and what they’ve been trained on
Lack of Common Sense and Understanding of the Physical world: They are trained on text mainly and some video and images. Even when trained on images output lacks realism
Bottom Line: Can someone Google an SOP for your job? No? You're safe. For now. If there are no publicly available SOPs (Standard Operating Procedure document) that describe each and every scenario for your role, and that LLMs can train on. Your role is safe.
In fact, a lot of roles require a human in the loop. A computer program can’t do it because there are unlimited combinations of “open” scenarios to deal with every day: i.e. how to backfill for Steve on his PTO, someone shared a document in a pdf format with a unique layout, being invited to step out of your regular role into a “stretch” assignment.
Jobs that are threatened are the ones that require little context and/or memory. Think of: Basic Virtual Assistants, Entry Level Research and Data Collection, PMO that’s there for taking notes and updating reporting on low stakes meetings.
The real signal?
When AI memory features start to become persistent and highly customizable. i.e., AI can learn and train over months, in an apprenticeship model, by observing the day-to-day.
When AI models weights can be customized by learning: “AI neuroplasticity” if I may coin the term
Last but not least, here is a number for you:
🤯 The Number
20min
Each ChatGPT answer consumes roughly the same electricity as 20 minutes of iPhone use. Small individually, but meaningful at scale.
In fact, the If ChatGPT handles 1 billion queries a day (as it did in Dec 2024), that’s enough energy to fully charge 55+ million iPhones
The Trade: Long Nuclear Power. That’s the only form of energy that’s stable and scalable to keep up. Most Nuclear Plants can scale up 20% in a few years.
Your Move
What did you think of this week’s edition?
Loved it: Alpha-packed
Good not Great: Positive EV
Sucks: Negative EV
Got feedback? Hit reply. I protect sources :)
Know someone who may be interested? Forward this.
See you next Sunday!
Playing the odds, Johan
Expected Value - Not investment advice. Do your own research, Seriously.