AI power demand is colliding with a supply gap, and the video splits the trade between the infrastructure names already winning orders and the riskier “picks-and-shovels” plays still in early monetization. NVDA-era data-center buildout is framed as a 38 GW global power shortfall through 2028, pushing interest toward VST, CEG, GEV, and BE because hyperscalers are locking up 20-year electricity contracts. At the same time, REDDIT’s ad/data leverage, AMZN’s capex shock risk, and TTD/TTWO-style valuation resets show how AI can lift some businesses while compressing others. The throughline: own the energy basket, treat timing as the main risk, and cut exposure fast when the market invalidates the story.
Verdive posted 61% revenue growth, orders more than doubled, and the stock jumped 16%.
GEV reported $29 billion in orders, a $197 billion backlog, and raised guidance, yet the stock sold off.
Bloom Energy was crushed 20% in 2 days last week, then rebounded 18% in 3 days; the note also cites a 20% down move followed by 18% up within 5 days.
Morgan Stanley estimates a 38 GW AI global power shortfall through 2028, framed as roughly 38 large nuclear reactors that do not exist.
Bloom Energy is pitched as a potential winner because data centers need more compute and energy is the missing piece for the next 5 years, even though BE still relies on natural gas.
CEG and VST are favored because they generate power and sit inside AI power-purchase agreements via long-term contracts with Meta, Amazon, and MSFT.
VST is described as a natural-gas utility trying to diversify into nuclear; its earnings outlook is 40% growth this year and 20%-30% for 2027.
Bloom is framed as a technology company monetizing a decade of R&D, but the bear case is high valuation, negative cash flow, and the risk that large deployments expose scalability limits.
Bloom’s setup is cleaner than gas turbines or generators: it needs only a natural-gas pipeline at the data center, avoids direct grid draw, cuts noise, and bypasses local “town hall meeting” opposition.
Bloom’s energy cost is higher today, but the cost is said to be falling every year and could reach parity over the next few years; the company also offers faster ramp-up than alternatives.
GEV, VST, and CEG are all framed as part of an “energy basket” because hyperscalers are trying to lock in electricity for the next 20 years, with some contracts potentially worthless in about a year and others turning into revenue.
GEV is described as boring but durable, with backlog sold out to 2032 and capacity already effectively booked 10 years out for some names.
Reddit’s traffic is weakening because search increasingly ends in AI mode and chat answers, reducing clicks even when links are shown.
OpenAI is buying Reddit information under a contract, while some other AI systems scrape Reddit directly; the result is that Reddit is seen as giving away data for pennies on the dollar.
Reddit’s Google deal signed in February 2024 is up in February 2026; the OpenAI deal signed in May 2024 is a $70 million deal and is up in May 2026.
Google’s $60 million and OpenAI’s $70 million deals are treated as too small for the value of Reddit’s data, with 2027 framed as the year to see what frontier models will pay for continued access.
Reddit is argued to be worth billions in AI deal value and possibly the most valuable acquisition target for AI companies, because a competitor would need billions to rebuild the product and ad base.
The Reddit trade is tricky because the next pricing reset depends on when a leak or renewed deal hits; once one AI provider pays up, similar terms could propagate across the others.
A long REDDIT position is held without leverage or options, but the position is constantly re-evaluated because timing is the main issue and the stock could re-rate sharply on new contract terms.
AMZN’s earnings trade is tied to a potential capex shock: if next-year spend comes in at $300 billion-$350 billion versus the last 2026 full-year guide of $220 billion, the stock is expected to get smoked.
The concern on AMZN is that investors will sell first and worry later if capex rises, because higher spend is seen as delaying cash flow next year; the response is to avoid being heavily levered into the print.
AMZN’s consumer hardware push includes a revamp of home products, including a laptop-like kitchen-counter device with a 16-inch screen priced around $700.
The Alexa-based products embed ads, and without a way to pay to remove them, the purchase case is viewed as weak.
TTWO is being cut back hard after the narrative shifted from just GTA 6 hype to whether AI will democratize game creation and pressure legacy publishers’ multiples; about 75% of the position was sold, leaving 20%-25% of the prior size.
The new TTWO question is whether there is still value in a gaming publisher by 2035 if tens of millions of people can create games on the fly and play them almost for free.
The video rejects using emotional market reactions as a guide: conviction only works when the information does not change, and any answer from AMZN that “pisses me off” is a cue to clip the position.
TSLA remains in focus via Optimus and FSD: a 2026 Model Y order kept getting pushed back, a pre-owned 2026 Model Y was bought instead, and a demo showed self-driving on Dallas traffic without destination input until navigation was entered.
On that Dallas-to-Austin drive, FSD is described as smooth and low-stress in Friday rush-hour traffic on 75, with the driver saying it was not nervous at all and that steering-wheel contact was only needed for takeover testing.
The movie “You’ll See Everything” is said to be playing only in theaters, with multiple viewings planned at North Park; Elizabeth Holmes is cited as having a 123-month sentence, reduced from 11 years 3 months to 10 years, and may be moved to a different facility.