Google Cloud’s revenue more than doubled in Q3, but Alphabet’s new filing shows why not every dollar should be valued like a recurring cloud subscription. TPU system sales bring hardware delivery, inventory funding and, in some arrangements, repurchase obligations. Our calculation of nine-month operating cash flow less cash equipment purchases fell from $48.715 billion to negative $4.804 billion. The opportunity is real; its cash requirements and retained risks deserve separate valuation inputs.
AHNIM STUDIO · News analysis · Alphabet · Nasdaq: GOOGL / GOOG · October 30, 2026 (KST). This review uses the September 30, 2026 Form 10-Q filed October 29. Quarterly income and revenue comparisons are Q3; the cash-flow comparison is the first nine months, not Q3 alone. All amounts are USD.
Cloud growth is strong, but the revenue mix has changed
| Measure | Q3 2025 | Q3 2026 |
|---|---|---|
| Alphabet revenue | $102.346bn | $127.577bn |
| Google Search & other revenue | $56.567bn | $63.350bn |
| Google Cloud revenue | $15.157bn | $30.709bn |
| Operating income | $31.228bn | $40.576bn |
| GAAP net income | $34.979bn | $26.468bn |
| Net income available to common stockholders | $34.979bn | $26.170bn |
| Diluted EPS | $2.87 | $2.12 |
Calculated from the reported filing amounts, Cloud revenue grew 102.6%, compared with 12.0% for Search & other. Group operating income rose 29.9%. The weaker bottom line is not evidence that operating profit collapsed: other income and expense moved from a $12.759 billion gain to an $8.221 billion expense. Preferred dividends then reduced the amount available to common shareholders by $298 million. Neither the investment-result swing nor preferred distributions should be silently merged into operating performance.
The filing describes Google Cloud as a combination of consumption-based infrastructure, subscriptions, Workspace and product sales, primarily TPU systems. A Tensor Processing Unit is Google’s custom AI accelerator. A server system sold to a customer is economically different from compute consumed over time on Google-owned infrastructure, even when both appear in the same Cloud revenue segment.
What the TPU accounting actually says
Alphabet says it signed a limited number of agreements to supply TPU systems for specialized, high-scale, on-premises workloads. Revenue recognition began in Q2 2026, with the significant majority expected in 2027. That is a company expectation, not recognized revenue or a guarantee that deliveries and cash collection will follow the same timetable.
| Component or condition | Filing treatment | Investment implication |
|---|---|---|
| Hardware and software | Generally recognized when hardware control transfers; hardware revenue is net of estimated return allowances | A delivery can bring revenue forward relative to a multi-year service contract |
| Installation, support and extended warranty | Recognized over the service period or as work is performed | Do not recognize all contract components at initial shipment |
| Certain repurchase options | Preset-price repurchases over a multi-year term, generally treated as rights of return | Residual-value and return assumptions remain relevant |
| Assignment to another party with repurchase right retained | Accounted for as a financial guarantee | Changing the owner does not necessarily remove Alphabet’s exposure |
Our interpretation is that the market should distinguish an accelerator product business from a purely recurring cloud service. Product sales can widen the addressable market and monetize Google’s architecture outside its own facilities. They can also require component purchases before delivery and expose Alphabet to return or guarantee economics after revenue is recorded. The filing does not provide a complete contract-by-contract cash schedule, an isolated TPU revenue total or a quantified loss estimate for these features. We therefore do not assign an invented liability or deduct the full contract value from earnings.
The cash test is broader than the capex headline
| Cash measure | 9M 2025 | 9M 2026 |
|---|---|---|
| Operating cash flow | $112.311bn | $127.149bn |
| Cash property-and-equipment purchases | $63.596bn | $131.953bn |
| OCF minus cash equipment purchases — our calculation | $48.715bn | −$4.804bn |
| Inventory cash-flow use | $1.154bn | $10.959bn |
| Net common-stock issuance proceeds | $0 | $42.870bn |
| Net mandatory-convertible preferred proceeds | $0 | $19.063bn |
| Net debt issuance proceeds, before repayments | $38.002bn | $85.222bn |
| Debt repayments | $26.094bn | $8.879bn |
| Share repurchases | $39.810bn | $0 |
Operating cash flow rose 13.2%, while equipment purchases rose 107.5%. Our residual fell by $53.519 billion. This is not a liquidity-crisis claim: September cash, equivalents and marketable securities totaled $265.098 billion. But that balance cannot be presented as proof that operating cash alone financed expansion; the statement records substantial debt, common and preferred issuance, as well as other investing flows.
Inventory also matters. The balance-sheet amount rose from $2.439 billion at December 31 to $13.308 billion at September 30, while the cash-flow statement shows a $10.959 billion inventory use during the nine months. Those are distinct measures and need not match. The policy note includes TPU systems, consumer devices and related components, so attributing the entire increase to TPU sales would exceed the disclosure. Nonetheless, the product model creates a second cash question alongside data-center capex: how quickly purchased hardware turns into delivered systems and collected customer payments.
The distinction changes how to compare Alphabet with Microsoft’s latest Azure and equipment-cash-flow analysis. Both face infrastructure spending, but one cannot assume identical revenue timing, inventory exposure or contract risk. Nor should customer financing be confused with product revenue, the issue examined in our NVIDIA–Meta financing analysis.
Backlog is a delivery obligation, not cash in the bank
At September 30, total revenue backlog was $567.7 billion, including $560.8 billion for Google Cloud. Alphabet expects just over 55% of total backlog to become revenue over the next 24 months. The note says recognition depends on contract duration, delivery and customer usage; some amounts have not yet been invoiced. It also says the company began including contracts with original terms of one year or less in Q1 2026. Historical backlog comparisons must preserve that scope change.
A large backlog supports evidence of demand, but it is neither current revenue nor a measure of prefunded infrastructure. Our next checks are deliveries against that backlog, the split between product and recurring services, inventory conversion, return allowances and guarantees, and operating cash after the capacity needed to satisfy contracts. An expansion funded externally can create value if incremental returns exceed its capital cost; the financing itself does not demonstrate that outcome.
Broker reaction and our valuation stance
In an October 29 report, Investing.com said DA Davidson cut its target from $350 to $300, keeping Neutral. The stated concerns included hardware-expense timing, weaker-than-expected Search and the funding overhang from higher capex. The reported new target uses 21× the firm’s FY2027 EPS estimate. We accessed the secondary report, not the broker’s original research. This individual action is not an analyst consensus.
We are not initiating a numerical AHNIM target in this event review. Our assessment is positive on the demonstrated Cloud demand but more cautious on treating its growth as a homogeneous, capital-light stream. A house valuation needs separate assumptions for Search earnings, recurring cloud returns, TPU product economics, financing cost and dilution. A verified, horizon-matched broker average and comparable AHNIM Base are not available, so a 20% target-gap comparison is not applicable. The reported quarterly EPS should not simply be multiplied by four.
A useful operating sensitivity, not a company forecast: one percentage point of margin on Q3 Cloud revenue equals $307.09 million of quarterly operating profit before any group-level allocation change. This illustrates why product mix, installation costs and support obligations can matter even when the revenue growth headline remains strong. It is not a forecast of TPU margin or an assigned fair value.
Sources and original documents
Alphabet — Form 10-Q for September 30, 2026, filed October 29; consolidated statements, Notes 1–2 and MD&A. Direct SEC document.
DA Davidson target action — October 29, 2026, secondary report via Investing.com.
Our research responsibility, AI assistance and corrections policy applies. Research is informational and educational, not personalized investment advice.
Leave a Reply