Alphabet

Alphabet Inc. (NASDAQ: GOOGL) is Google’s parent, and the two names still mean almost the same thing: search and ads drive most of the $127.6 billion Alphabet reported for the third quarter of 2026, up 25% year over year. Google Cloud revenue rose 103% to $30.7 billion; Search grew 12%, down from 17% in Q2. Capital spending of $51.4 billion outran $42.3 billion of operating cash, so free cash flow was negative $9.1 billion for the second quarter running. The capex bull case models that spending. This is the business most headlines mean by “Google.”

It isn’t the whole business. Read Alphabet through TEXXR’s knowledge graph instead of its income statement, and a second company appears: a roughly 14% stake in the AI lab it competes with, a decade-old bet now inside the largest IPO in history, a self-driving unit Alphabet keeps re-funding, and a Nobel-winning lab whose star scientist left in June for the rival Alphabet partly owns. Little of that has a line of its own on an earnings release. The graph tracks what a company owns and deals with, not what it books as revenue.

The 10-Q shows how much that second company now earns. Other income and expense, net, was $127.5 billion for the nine months, 51% of $248.5 billion in pre-tax income, “largely from a private company, and SpaceX prior to its initial public offering.” The filing does not name the private company. Cumulative upward adjustments on Alphabet’s non-marketable equity doubled in nine months, from $44.5 billion to $88.8 billion. Coverage puts a 14% Anthropic stake among Alphabet’s holdings (below), but the filing does not say which holding produced the gain. Stakes that move half of pre-tax income are no longer quiet, which is the question this page ends on.

TEXXR logged 385 Google articles in the second quarter of 2026, 1.7 times the average of the eight quarters before it (232).

This is DeadRisk’s mag7 tier: the deepest corpus we hold on any single name, and not a stock-picking exercise. Benjamin Graham’s whole method was refusing to let a quoted price stand in for what a business actually owns. Alphabet is a clean modern test of that habit: half its nine-month pre-tax income came from what it owns, not what it sells.

The record’s other throughline is three years of coverage that ran as a company caught behind, then recovering. In December 2022, sources told the New York Times that ChatGPT’s release had led Google to declare a “code red”.

Gemini 3 was the answer. It arrived on November 18, 2025 with the Gemini app already at 650 million users, a figure Google put past 900 million by May 2026.

One caution for the chart below. In 2026Q1 TEXXR’s graph changed how it labels Google’s coverage: logged edges rose 3.6-fold, from 265 to 959, while articles rose by about half, and a “competitor” label that had never appeared took 11% of the 182 labelled edges. That quarter’s mix shift is a change in the ledger, not in the news.

The recovery’s second act is easy to miss, because it does not look like winning. Through the first half of 2026 the frontier race — the benchmark leaderboard everyone watches — was not going Google’s way. Bloomberg reported that Gemini 3.5 Pro had fallen months behind schedule, held back for missing Google’s own internal targets, and on the July 22 earnings call CEO Sundar Pichai named the weak spot himself: coding and agentic coding, the exact ground where rivals lead. Measured only on the frontier, Google looked like it was slipping again.

The revenue said otherwise. The day before those earnings, Google shipped not a frontier model but Gemini 3.6 Flash, priced below the model it replaced, at $1.50 per million input tokens against $7.50 per million output, and better at coding and knowledge work while spending fewer tokens. Flash is the workhorse — the model companies run at volume — and the volume is the tell. On the Q3 call Pichai put API throughput at 32 billion tokens a minute, up 40% on the quarter, while Cloud revenue grew 103% at a 36.2% operating margin, against 23.7% a year earlier. The market grades the benchmark that names the smartest model. The income statement scores a different contest — who serves the most useful token at the lowest cost — and on that one Google is not visibly behind. The Power Wall tracks the same fight from the hardware side: when power binds, a cheaper token is a cheaper megawatt. The economics of that token, not the leaderboard, is the number that clears.

The bill lands on the same statement. In Q2, capital spending of $44.9 billion passed the $39.1 billion of operating cash and pushed free cash flow to negative $5.9 billion, its first negative quarter in at least 18 years of filings. Q3 repeated it at negative $9.1 billion, and the 10-Q shows $12.4 billion of stock sold under the at-the-market program. Management held 2026 capex at $195–205 billion, raised in July from $180–190 billion, and said 2027 will “increase significantly,” with the figure left to the Q4 call. Google is a low-cost producer at the model layer and spends hard enough on the physical layer to run cash-flow negative. The two are the same bet: the cheapest token wins only if you own enough capacity to serve it. It is the AI-capex trade written into one company’s cash statement.

Google paid $650 million for DeepMind in 2014, beating a Facebook bid. A decade later Demis Hassabis and John Jumper shared the 2024 Nobel Prize in Chemistry for AlphaFold. In June 2026, Jumper said on X he was leaving for Anthropic, and Bloomberg wrote that the exit strains Google’s effort to compete with Anthropic and OpenAI. Alphabet’s stock closed down 5% the next trading day. TEXXR’s graph logs it as a competitor edge linking Google, Anthropic, and OpenAI, not a personnel edge alone.

Anthropic is the clearest case of Alphabet holding equity in a company it also fights. Google put roughly $300 million into Anthropic in February 2023, the FT reported, for about a 10% stake. By March 2025, SEC filings showed the 14% stake, with no voting rights, no board seat, and no board-observer rights. In April 2026, Google agreed to invest up to $40 billion more — $10 billion in cash now, $30 billion tied to performance targets. Google is also Anthropic’s landlord: an October 2025 cloud deal gave Anthropic access to a million TPUs and a gigawatt of capacity; an April 2026 expansion with Broadcom took that to roughly 3.5 gigawatts, and Google is building a $3.2 billion New York data center to rent Anthropic its own TPUs, using what the Journal called “Nvidia’s playbook” against Nvidia. Anthropic’s S-1 puts its Google commitment at $114.5 billion through July 2033, plus an estimated $125–150 billion of TPUs from Google and others for 2028–2031. Investor, landlord, and rival, in the same filings.

Anthropic is no longer the only tenant. The Information reported that OpenAI plans to buy Google’s TPUs; Oracle added Google TPUs to Oracle Cloud; and on the Q3 call Alphabet said more than 100 customers use TPUs, Apple among them. Some TPU deals are sales, not rentals: the 10-Q says Cloud product sales “generally consist of the sale of TPU systems,” some with options to repurchase the hardware at pre-set prices. Crux AI, a venture with Blackstone in which Google holds a minority stake, signed a multi-year deal that the 10-Q counts for $16.8 billion of Alphabet’s backlog and calls a related-party transaction: investor and customer again.

SpaceX is the oldest stake, and the one the record can actually price. In early 2015 Google put $900 million into SpaceX — a $1 billion round — when SpaceX was valued above $10 billion. By April 2026, filings showed the stake near 5%, down from 6.11% at the end of 2025 — a position Bloomberg estimated could top $100 billion if SpaceX’s planned IPO hit its target. SpaceX went public two months later as the largest IPO in history, and Alphabet’s 10-Q valued its shares at $94.1 billion on June 30, on an outlay under $1 billion. The Q3 10-Q shows the other side. Other income and expense swung from a $12.8 billion gain in Q3 2025 to an $8.2 billion loss, on unrealized losses “largely from SpaceX.” SPCX closed at $150.86 on September 30 against $170.86 on June 30, down 11.7%, and $75.2 billion of Alphabet’s shares were still under “short-term restrictions on the ability to sell.” Alphabet had already booked an $8 billion unrealized gain in early 2025 on a private holding that Bloomberg’s source said was SpaceX. A week before SpaceX’s debut, its filing showed Google had agreed to pay it $920 million a month for compute on Nvidia chips, October 2026 through mid-2029. Alphabet books losses on its SpaceX shares while it pays SpaceX rent: investor and customer, the same pattern as Anthropic.

The rest of the web is Alphabet declining to let go. Waymo has taken Alphabet money in every major round since the 2017 “Other Bets” reorganization, most recently in February 2026, when the FT’s sources put Alphabet’s share at $12 billion or more of a $16 billion round at a $126 billion valuation. Isomorphic Labs, launched by Alphabet in 2021 with Hassabis as CEO, raised $2.1 billion in 2026, led by Thrive Capital rather than Alphabet.

Positioning: how much of GOOGL is sold short, on the record brokers file twice a month. It lags by design — read it as where the crowd stood, not where it stands.

Signals in the coverage record worth tracking, not trade ideas:

• Answered on October 28. Free cash flow stayed negative, at $9.1 billion. Flash volume reached 32 billion tokens a minute. The SpaceX mark came in lower, with $75.2 billion of shares still restricted.
• Anthropic’s listing. Its S-1 became public on October 26 for a Nasdaq listing as ANT, with share count and price range blank. Reuters’ sources said investors discussed about $1.5 trillion, down from $2 trillion. A listed price lets the stake move daily, as SpaceX’s did. Watch the price range, and whether the prospectus states Alphabet’s share.
• The 2027 capex figure on the Q4 call. Management said only that it will “increase significantly.” Against negative free cash flow, it shows how far the bet runs on outside money.
• The SpaceX restrictions. The 10-Q gives no date when they end. The Q4 filing shows how many shares remain restricted, and the next mark.
• Talent flow out of DeepMind. Jumper’s exit moved the stock. Watch whether more follow.
• Isomorphic Labs and Waymo. Isomorphic is in early talks at a $40–50 billion valuation, Bloomberg’s sources said on October 8; an outside round dilutes Alphabet’s share, as SpaceX’s did.

Alphabet lives or dies by the same constraint as these names — small equal-weighted groups, one constraint each. Not investment products: see all categories.