Alphabet Is The Cheapest Way To Own The AI Infrastructure Trade

Alphabet invested in custom silicon in 2015 to reduce AWS and Nvidia dependency for Google-internal workloads. For most of the past decade, the TPU was an internal cost-optimisation tool. The revenue contribution was effectively zero. The accounting treatment reflected the internal-use nature: TPU capex was absorbed into the infrastructure line without a separate revenue disclosure.

That treatment is ending. Google Cloud's TPU revenue from external customers is estimated at $3-4 billion annualised as of Q4 2025, growing at 200%-plus year over year. The customer mix includes the expected AI-native companies (Anthropic, Character, others) plus a growing list of enterprise customers running Gemini API workloads. Chinese startups, limited in their access to Nvidia hardware by export controls, have increasingly turned to TPU via Google Cloud. The combined addressable market for TPU-as-a-service has expanded materially over the past twelve months.

The Valuation Desk models TPU-as-a-service revenue reaching $12-18 billion annualised by end of FY2026. Applied to peer cloud infrastructure multiples, that revenue stream is worth an incremental $150-220 billion of enterprise value that is not in consensus. That alone justifies a 5-7% multiple expansion in Alphabet's current share price. Cumulative TPU compute capacity is the underlying asset being monetised; the pricing model is converging toward cost-plus-modest-margin for external customers, which is still valuable at scale.

The consensus fear on Alphabet for the past three years has been that generative AI will destroy Search. That fear has been wrong. Search revenue grew 12% in FY2024 and approximately 11% in FY2025. The underlying query volume continues to grow, although click-through rates on traditional search results have modestly compressed. Alphabet has compensated through AI Overview monetisation, which shows ads within the AI-generated answer box rather than on traditional result pages.

The AI Overview monetisation rate is currently running at approximately 85-90% of traditional search rates. That gap is expected to close through 2026 as advertiser targeting tools improve within the AI context. The Valuation Desk models AI Overview monetisation reaching 100% parity with traditional search by late 2026. That represents a quiet upside to consolidated Search revenue that consensus models do not currently incorporate.

The dynamics look like the mobile search transition of 2012-2015. Mobile initially monetised at a lower rate than desktop, analysts worried about the monetisation gap, and the gap closed within three years. History does not repeat, but it rhymes; the Search monetisation engine has shown repeated ability to reprice new formats. AI Overview ads have also shown structurally longer dwell time per impression, which is a quality lever distinct from the volume story and has meaningful implications for Alphabet's pricing power with advertisers over the next eighteen months.

Google Cloud operating income hit approximately $18 billion in FY2025, up from $10 billion in FY2024. Operating margin expanded from 14% to 22%. That is a meaningful operating leverage move in a single year. The segment generates $53-55 billion of annual revenue at a $250-280 billion implied valuation based on peer cloud infrastructure multiples. That alone is 6-7% of Alphabet's market capitalisation.

Unlike AWS, which has been in the harvest phase of its capex cycle for five years, Google Cloud is in the build phase. The segment has been gaining meaningful share in the mid-market and in AI-native startups. Large enterprise wins have been more sporadic but trending in the right direction. The Valuation Desk models Cloud revenue reaching $85-95 billion by FY2027, with operating margin expanding to 28-32% at that scale.

At $90 billion of revenue and 30% operating margin, Cloud would generate $27 billion of operating profit. At a 22x forward multiple on that profit stream, the segment alone would be worth approximately $600 billion. That is a 270% expansion from current implied valuation. Even with conservative assumptions on the revenue and margin trajectory, the value creation runway in Cloud is larger than most investors appreciate. That growth trajectory is resilient to the competitive pressure from other streaming services.

YouTube advertising revenue hit $42 billion in FY2025 and is growing at 13%. YouTube subscription revenue (YouTube Premium, YouTube TV, YouTube Music) added another $16 billion at growing rates in the high teens. Combined YouTube revenue is approximately $58 billion annualised. Apply dedicated streaming and ad tech peer multiples, and the combined YouTube business is worth $400-450 billion of enterprise value on a standalone basis.

That is 11-12% of Alphabet's total market capitalisation. The consolidated multiple obscures this value. Analysts who have begun modelling YouTube separately in sum-of-parts frameworks arrive at fair value estimates 15-20% above consensus. The market has slowly begun rewarding this disaggregation; it has more room to run.

By comparison, Netflix has a $340 billion market cap on $40 billion of revenue. YouTube's combined advertising-plus-subscription revenue base is 45% larger than Netflix with comparable growth characteristics. The valuation multiple implied for YouTube inside Alphabet is less than half of the comparable Netflix multiple. That asymmetry is the unexploited opportunity. Waymo's insurance and liability cost structure has been improving as the fleet matures, which changes the unit economics materially over time.

Across three prior hyperscaler repricings, the pattern has been consistent. First, the market anchors on the core business (Search for Alphabet) and prices everything else at a discount. Second, a single external validation event (major customer win, regulatory clarity, competitor misstep) shifts attention to the underappreciated asset. Third, sell-side models rebuild sum-of-parts frameworks and the aggregate multiple expands over twelve to twenty-four months.

The Microsoft repricing in 2017-2019 is the textbook case. The stock spent 2017 at 22x forward earnings, expanded to 28x through 2019 as Azure and Office commercial cloud gained analyst visibility. The path added approximately 55% to total return beyond earnings growth alone. Alphabet's current setup rhymes. The core business is Search and YouTube advertising. The underappreciated asset is TPU and broader Cloud. Waymo is the optionality overlay.

The external validation event is arriving in real time. Chinese AI startup adoption of TPU-as-a-service is the validation. Sell-side models are beginning to ask whether TPU revenue should be disclosed separately. The second-stage narrative shift is in process. The valuation re-rating will lag by six to twelve months.

Waymo completed approximately 2 million commercial driverless rides in Q4 2025 and is expanding to nine metro areas by end of calendar 2026. Revenue from commercial operation is estimated at $400-600 million annualised today, growing at 150%-plus year over year. Management has not disclosed segment profitability, but industry estimates place the segment at roughly break-even on fully-loaded economics.

The Valuation Desk models Waymo revenue reaching $4-6 billion by end of FY2027 on the current geographic expansion path. At a generous multiple (matching the most optimistic Tesla robotaxi valuations), Waymo could be worth $40-80 billion standalone. At a conservative multiple (matching established ride-share unit economics), Waymo could be worth $15-25 billion. The midpoint of that range is $28-50 billion, representing 1-1.5% of Alphabet's market cap.

That is a modest contribution to the overall valuation argument. But it is a concrete option value that is increasing in probability as Waymo builds its operational footprint. The option is particularly valuable because it is counter-correlated with Tesla's robotaxi narrative; a Waymo leadership position invalidates a significant component of Tesla's bull case. That competitive dynamic alone supports a valuation premium on Waymo beyond the standalone revenue analysis.

Alphabet introduced its first-ever dividend in FY2024 and has since grown it in line with management's stated framework. Combined dividend plus buyback capital return reached approximately $75 billion in FY2025, absorbing just over 100% of free cash flow during the capex-heavy year. That discipline matters. Cash and equivalents sit at $111 billion against $30 billion of debt; net cash of $81 billion remains.

The capital return framework creates a specific valuation floor. At the current $335 share price, the combined dividend yield plus buyback yield produces a total shareholder yield of approximately 4.2%. Investors who evaluate large-cap compounders on total yield metrics find Alphabet attractive at these levels even before accounting for the growth optionality. The yield alone supports the case against meaningful downside from here.

Most importantly, the capital return programme demonstrates that management is not running a pure growth-at-any-cost strategy. The decision to initiate the dividend signalled discipline. That discipline has been maintained. For a mega-cap that could easily default to retention of all free cash flow, the capital return pattern is differentiated and deserves a modest multiple premium versus peers.

At 29.9x forward earnings, Alphabet trades at approximately a 30% premium to Microsoft and a 25% discount to Meta on an earnings basis. On free cash flow yield, Alphabet is the cheapest of the three. The Valuation Desk argues that Alphabet deserves a discount to Meta (given Meta's superior growth rate) and a premium to Microsoft (given Alphabet's better strategic positioning on AI infrastructure). The current pricing places Alphabet in approximately the correct position between these two peers, meaning the undervaluation argument runs through absolute multiple expansion rather than relative repricing.

The absolute multiple expansion would be driven by the TPU, YouTube, and Waymo narratives proving out over the next eighteen months. Each is independent of the others. Any two delivering above-consensus outcomes would shift the forward multiple to 33-35x earnings. All three delivering would reach the 35-38x range. On current consensus earnings, that translates to fair value of $400-475 per share.

Historically, Alphabet has traded in a wide 20-34x forward earnings range, depending on growth perception and capex cycle positioning. The current 29.9x multiple sits at the upper-middle of that range. The path to the upper end is narrow but visible.