Serenity, the small-cap investor who back-maps the AI buildout to find obscure upstream suppliers, spent today on the demand side of his framework — the hyperscalers whose spending he treats as a telegraph for where his bottleneck bets pay off. He also flagged an amusing meme-stock episode tied to one of his core holdings.

A Reddit-fueled rally in Wendy's

Serenity noted with some amusement that retail traders on Reddit — the social platform behind one of his highest-conviction longs — had launched a viral campaign to "save" Wendy's (WEN), the US fast-food burger chain, sending the stock up around 20% overnight. The Wendy's name has long been a running joke in retail-trading circles, shorthand for working behind the dumpsters after blowing up an account on short-dated options. For Serenity, Reddit (RDDT) is a roughly 5% position and a "money printer" he believes is wrongly lumped in with AI-disrupted software; episodes like this reinforce his view that its engagement and network effects are durable and hard to displace. He took no position in Wendy's and treated it purely as a curiosity, wondering aloud whether the campaign would actually work.

Ranking the hyperscalers by capex payback

The more substantive post pushed back on the idea that hyperscaler AI capex is money being "siphoned off." Serenity frames that spending instead as an investment in future revenue or margin, and he uses the hyperscalers' roadmaps to justify his upstream chokepoint bets — the lasers, substrates and photonics suppliers that feed the optical buildout.

He named Amazon (AMZN) as his current favorite hyperscaler. His argument: Amazon's headcount of roughly 1.57 million gives it the clearest, most concrete path to a payback — using AI and eventually physical robotics to automate warehouses, shipping and delivery and cut operating costs, while AWS grows compute on its in-house Trainium chips and potentially sells silicon in a neocloud-style model. Amazon is a hyperscaler whose ~$200B capex and automation push he maps directly onto his upstream longs; he has never claimed a personal position, treating it as a demand driver rather than a holding.

He placed Alphabet (GOOGL) second. Google's AI capex, in his telling, was necessary to defend its search moat — Gemini reclaiming ground from ChatGPT — while Google Cloud grows on efficient TPUs that Alphabet could ultimately sell the way Nvidia sells GPUs. Alphabet is his demand anchor for the whole photonics trade, the money-printer whose TPU bill of materials underwrites his upstream picks; he owns it and rates it highly, though he sees a less obvious physical-AI angle than Amazon's.

He contrasted both with Tesla (TSLA), whose Optimus humanoid program he calls a very broad pitch with no single concrete use case yet — versus Amazon's specific, internal reason to scale robotics now. Consistent with his approach, he admires Elon Musk's execution but plays the robotics theme through suppliers rather than owning Tesla itself.

Serenity argued Microsoft and Meta are still working to convince the market their capex is justified — pointing to Meta's 30%-plus revenue growth as early evidence — and that Microsoft's Maia chip and AI sentiment sit near the bottom, much as Google's did before its turnaround. His underlying thesis: a "leader effect" in frontier models is pushing everyone to rush the buildout, which keeps demand flowing to the upstream suppliers he favors.

*This is derived commentary, not investment advice.*