Alphabet's stock rallied 160% in 12 months. The best month since 2004 was April 2026 — when the broader market was falling. That tells you something about what the market is actually pricing in.
It is not pricing in a search business. It is not pricing in YouTube ad revenue. It is pricing in a company that has quietly assembled the only complete AI stack in public markets: proprietary chips, frontier model, hyperscaler cloud, and the most profitable search monopoly in the history of the internet — all inside a single ticker. The Q1 2026 numbers are not the story. They are the evidence.
Google Cloud's $20B Quarter — and the $460B Backlog Behind It
In its Q1 2026 earnings call on April 29, Alphabet reported $109.9B in total revenue, beating the $107.2B consensus. Net income came in at $62.58B — up 81% year-over-year. EPS hit $5.11.
The number that matters most is Cloud. Google Cloud crossed $20B in quarterly revenue for the first time, growing 63% year-over-year. That is not a catch-up story anymore. AWS and Azure have been the dominant hyperscalers for years. A 63% growth rate at this scale suggests Alphabet is taking share, not just growing with the market.
The backlog figure reinforces it. According to Alphabet's investor relations disclosures, the Cloud backlog stands at approximately $460B — nearly doubled from the prior period. Cloud backlog is contracted revenue not yet recognized on the income statement; it is a forward visibility metric. A $460B backlog means customers have already committed. Revenue is waiting to be unlocked.
CEO Sundar Pichai described the company as "compute constrained" on the April 29 call. That is the most important three words in the earnings transcript. Demand is outrunning capacity. Alphabet is spending $180–190B in capex in 2026 to fix that problem.
What Alphabet's AI Stack Actually Includes
The thesis is straightforward: Alphabet is not an AI company. It is AI infrastructure. Here is what that means at the component level.
Search — the cash engine that funds everything
Search generated the cash flows that paid for DeepMind, for TPU R&D, for the cloud build-out. Google Search remains the most durable cash-generating asset in technology. It is also the surface where Gemini is being deployed at scale — the AI Overview feature now reaches billions of queries. The monetization model is evolving, but the traffic moat is intact.
TPU — building what the world will rent
TPU stands for Tensor Processing Unit — Google's proprietary AI chip, the in-house alternative to Nvidia GPUs. Alphabet does not disclose TPU revenue separately, but Citizens Financial analysts estimate the segment generated approximately $3B in 2026, with a forecast of $25B in 2027. That is an 8x increase in one year if the estimate is right. The model here is straightforward: build the compute you need internally, then lease the excess capacity to the world. It is exactly what AWS did with spare server capacity in 2006.
Google Cloud — the distribution layer
Cloud is the commercialization vehicle for everything else in the stack. TPU access, Gemini API calls, Vertex AI workloads — all of it runs through Cloud billing. The $460B backlog means enterprise customers are not just testing; they are committing multi-year contracts. That is structural revenue, not cyclical.
Gemini — the frontier model in the race
Gemini is Alphabet's answer to GPT-4o and Claude. It is embedded in Search, Workspace, Cloud, and Android. The competitive question — whether Gemini is ahead or behind OpenAI — is real. But it misses the point. Gemini does not need to win the model race. It needs to be good enough to keep enterprise customers inside Google's ecosystem. At current capability levels, it is more than good enough.
The TPU Thesis: From $3B to $25B in One Year
Let me be direct about what Citizens Financial analysts are suggesting with the TPU estimate. If $3B in 2026 becomes $25B in 2027, that is a new business line — inside one of the world's largest companies — growing faster than Google Cloud did in its best years. The way Ruslan Averin frames it: Alphabet built the compute it needed internally, then realized the world wanted to rent it.
The mechanism is not speculative. Alphabet built TPU infrastructure to train Gemini and run Search AI at scale. The marginal cost of that infrastructure is already sunk. Leasing excess TPU capacity to third parties — through Google Cloud — converts fixed cost into incremental revenue at high margins. Microsoft did this with Azure OpenAI. Amazon is doing it with Bedrock and Trainium. Alphabet has better chips, more of them, and a cloud platform already winning enterprise contracts.
I do not treat the $25B figure as guidance — Alphabet has not issued any. But I treat it as directionally credible given the infrastructure already in place. The TPU revenue line is the thing to watch in Q2 and Q3 2026 disclosures.
Is 28x Forward P/E Expensive — or Are You Pricing the Wrong Thing?
Forward P/E is the price-to-earnings ratio based on next-12-months consensus earnings estimates. At current prices, GOOGL trades at roughly 28x forward earnings, according to sell-side consensus estimates. The 10-year historical average for the stock is below 21x. That gap is real and worth taking seriously.
The bear case is simple: 28x is expensive for a search-and-advertising business. If AI disintermediates search — if people start asking ChatGPT instead of Googling — Alphabet's core revenue model faces structural pressure. The market is paying a premium for a company that may be disrupting itself.
The bull case is also simple: 28x is not expensive for the only company that owns the compute layer, the model layer, the distribution layer, and the cash engine simultaneously. You are not buying a search business at 28x. You are buying a full AI infrastructure platform that happens to generate $62B in quarterly net income.
My read is that the valuation is stretched relative to history, but the correct comparison is not history. The correct comparison is what Alphabet's earnings power looks like in 2028 if Cloud hits $100B annually and TPU becomes a disclosed segment. At that earnings trajectory, 28x today looks different.
The market cap as of May 10, 2026 stands at $4.8T — ahead of Apple at $4.3T, behind Nvidia at $5.2T. That ordering reflects the market's current ranking of AI infrastructure value. Alphabet is positioned between the chip maker and the device maker. That is exactly where the stack thesis places it.
My Position on GOOGL
I hold GOOGL in the portfolio. At 28x forward, I'm not adding aggressively — but I'm not trimming either. The stack thesis is intact.
The position reflects Ruslan Averin's position sizing discipline applied to high-conviction concentrated tech exposure: enough to matter if the thesis plays out, sized to survive if it takes longer than expected.
What would change my view? A sustained loss of Search query share to AI-native interfaces — measurable, not anecdotal. Or Cloud growth decelerating below 40% with no TPU segment revenue to offset it. Neither is happening now.
What am I watching in Q2? The TPU disclosure cadence. If Alphabet begins breaking out TPU revenue — even informally in the earnings transcript — that is the signal that management is preparing investors for a new segment. That is when the re-rating starts.
The broader market sold off in April. GOOGL gained 34% — its best month since 2004. A company that rallies hard when everything else falls is telling you something. The market has already decided what Alphabet is. The question is whether the stack delivers the earnings to justify it.
That is the bet I am holding.
— Ruslan Averin, averin.com