Serenity Analyzes Diverging AI Capex Paths Among Tech Giants
Claire Weston
Analyst Serenity ranks the Magnificent Seven's AI capex logic, calling Amazon the clearest case — 1.57 million employees make internal automation a quantifiable payoff anchor, while not every mega-spender can tell an equally solid story.
Is the money well spent — where's the real disagreement?
Serenity rejects the claim that "profits are being siphoned off by chip vendors." She sees massive AI capex as investment in future revenue growth or margin improvement, not pure cost drain.
This means → the right question isn't "how much are they spending," but whether each dollar points to a specific, measurable return.
The gap across the seven is wide — some can already quantify where savings land; others are still in "spend now, prove later" mode.
Why is Amazon ranked the top sample?
Amazon employs roughly 1.57 million people — a massive headcount. In plain terms = internal automation alone offers an enormous, concrete cost-optimization surface.
Path one: deploy LLMs — large language models, AI that processes and generates text — to restructure internal workflows, then extend into autonomous delivery and warehouse robotics. This maps directly to operating-cost reduction.
Path two: AWS's in-house chip Trainium is scaling up compute capacity, with potential to sell chips externally via a "Neocloud" strategy. This means → Amazon isn't just an AI buyer; it's trying to become an AI compute seller.
Serenity contrasts Tesla's Optimus robot: Optimus has a broader narrative, but Amazon has clear internal operating costs to cut, giving it stronger near-term execution certainty.
Google ranks second — what's strong, what's weak?
With ChatGPT pressuring its search business, AI spending is Google's necessary defense of the search moat and its push to close the gap with Gemini.
Google Cloud benefits from in-house TPU — Google's custom AI chip — gaining a cost-efficiency edge, with potential to sell TPUs externally and replicate Nvidia's model.
Gemini's user base keeps growing, and AI capabilities feed into ad-targeting optimization. But Serenity flags a gap: Google's physical-AI narrative — robotics, autonomous vehicles — remains vague. This reflects the key difference from Amazon: strong on the software-defense side, thin on the hardware-offense side.
Microsoft and Meta — still in "prove it" mode?
Both still need to convince the market that capex at this scale is genuinely necessary.
Meta's revenue is growing at over 30% year-on-year, partly validating its spend. Microsoft's situation is more complex: its in-house chip Maia is seen as behind schedule, and its large OpenAI investment is viewed as diverting internal AI R&D resources — weighing on sentiment.
Put simply = Meta can at least point to growth numbers; Microsoft is carrying two burdens at once — "chips are late" and "money went to someone else."
Serenity notes, however, that this pessimism could reverse — similar to the 180-degree sentiment turn Google experienced earlier.
Why are all the giants doubling down at once?
The core driver is a winner-take-most dynamic: the company with the strongest model keeps attracting users, harvests more training data, and reinforces its advantage — potentially creating a structural moat.
This means → this isn't a simple arms race. It's a land-grab under "winner-take-all" expectations — the cost of falling behind may far exceed today's capex.
But Serenity also notes that near-term monetization clarity varies sharply: Amazon can already do the math; Microsoft is still telling the story. The spending logic is the same; the spending confidence is not.
Content is for reference only, not financial advice.