In a bull market, the smartest money doesn't chase hype; it audits the code of monetization. Bill Ackman’s Pershing Square just filed its 13F, revealing a stark shift: Amazon is now the fourth-largest holding, while Alphabet has been dumped entirely. This is not a macro bet on e-commerce vs. search. It is a technical verdict on which company’s AI revenue model is actually verifiable.
Truth is not given, it is verified. And the market is finally running the numbers.
Context: The 13F as a Philosophical Statement
Pershing Square’s move lands in a bull market where AI euphoria masks deep structural flaws. Every tech giant claims to be “AI-first,” but the difference between a revenue stream and a business model is the difference between hype and code. Ackman’s fund is known for concentrated, thesis-driven bets. The switch from Alphabet to Amazon signals a conviction that the two companies’ AI monetization paths have diverged irreversibly.
Amazon’s AWS has an annualized revenue run rate exceeding $100 billion, with AI services (model hosting, inference, training) contributing a rapidly growing share. By contrast, Alphabet’s core search advertising—still ~75% of revenue—faces an existential threat: AI conversational search (ChatGPT, Perplexity, Gemini) reduces click-through rates, directly undermining the ad-based monetization model. This is not a minor headwind; it is a structural rewrite of the value extraction mechanism.
Core: The Technical Divergence in AI Revenue Certainty
Let me deconstruct the core difference from a builder’s perspective.
Amazon’s AI infrastructure is built on a modular neutrality principle. AWS offers Bedrock, a platform that gives customers access to multiple models—Anthropic’s Claude, Meta’s Llama, Amazon’s own Titan, and even third-party models. This approach reduces client lock-in risk. You pay for compute, not for a model. The revenue is metered by token consumption and inference time, directly tied to AI usage. It’s a utility model, not a platform lock-in model.
In contrast, Google Cloud’s Vertex AI is optimized for Gemini, Google’s own model suite. While technically superior in many benchmarks, this creates a self-model conflict. Enterprise clients worry: “If I use Google Cloud to run a non-Gemini model, am I subsidizing a competitor?” This perception, whether real or not, dampens adoption. The architectural choice to prioritize proprietary models introduces friction that AWS’s model-agnostic stance avoids.
Skepticism is the first step to sovereignty. The market is learning that open, modular infrastructure commands a higher trust premium than vertically integrated stacks.
Furthermore, Amazon’s AI hardware—Trainium and Inferentia chips—is purpose-built for AI workloads, reducing dependency on NVIDIA and lowering costs for clients. Alphabet has TPUs, but they are tied to Google Cloud. AWS’s chips are available to any customer, reinforcing the neutral utility narrative.
From a regulatory perspective, Alphabet carries a heavier asymmetric risk. The US Department of Justice’s antitrust case against Google’s search monopoly could force the company to divest its ad tech or end default search agreements. That would directly puncture the core cash cow. Amazon faces antitrust scrutiny too, but it targets its retail dominance, not its cloud profit center—a less existential threat in the short term.
Contrarian: The Inverse Signal and Its Blind Spots
But let’s not mistake correlation for causation. Ackman’s move could be driven by factors unrelated to AI: valuation multiples, political alignment, or simple portfolio rebalancing. The 13F filing has a 45-day delay, so the market may have already priced in this shift.
More importantly, Alphabet’s technical capabilities in AI remain first-tier. Google DeepMind’s Gemini models rival OpenAI’s GPT-4, and its open-source Gemma models are gaining developer traction. The company is also investing heavily in AI-first search experiences (SGE, AI Overviews). If Google can successfully transition its search ad model to a conversational context—charging for clicks in a new format—the monetization gap could narrow.
Modularity is the architecture of freedom, but centralization can be efficient. Google’s unified stack (TPU + Gemini + Google Cloud) might eventually produce higher margins and faster innovation, if it overcomes the trust barrier. The contrarian bet is that the market is overreacting to short-term revenue visibility and underestimating Alphabet’s ability to adapt its business model.
However, the evidence from the current earnings cycle favors Ackman’s thesis. Amazon’s AWS AI revenue grew over 40% year-over-year in the most recent quarter, while Google Cloud’s growth slowed to 30% and its search advertising revenue growth decelerated. The numbers are not lying; they are code.
Takeaway: The Verdict is in the Audit Trail
In the bear market, only code remains. But in a bull market, the code that survives is the code that monetizes with verifiable certainty. Pershing Square’s move is a canary in the coal mine: the next cycle will reward companies that build open, modular, and trust-minimized AI infrastructure—principles that align with the decentralized ethos we champion in crypto.
Amazon’s AWS is currently the closest approximation of a permissionless AI utility. Alphabet’s Gemini is a walled garden with a beautiful lock. The market is voting with its capital. The question for builders is: which architecture will you bet on when the bull market euphoria fades and only code remains?
As always, we do not trust; we verify. The next 13F filings will tell us if this is a trend or a single trade.