Will Gemini 4 Argon save Google? Wrong question

Our answer

No, and it doesn't need to. The question assumes Google is in trouble and waiting on a model to rescue it. The record says otherwise: Search is still growing, Cloud revenue grew 82% last quarter, and the stock is up sharply over the past year. Google does have two real problems. The first is paying for AI, which turned free cash flow negative in the second quarter. The second is that ChatGPT, not Gemini, owns the consumer assistant. Once it ships, Gemini 4 Argon can help with the first at the margin; today it is in a position to help with neither. On balance, Argon is a strong model and a useful lever, not a rescue.

Google is not sinking

Start with the business the question says is under threat. In the second quarter, Alphabet's revenue rose 24% to $119.8 billion, Google Search & other grew 17%, and operating income rose 30% at a 34% margin ^[1]. Sundar Pichai says AI Mode has passed 1 billion monthly users and, like AI Overviews, is adding to total Search queries ^[3]. Whatever AI is doing to search, it has not yet shown up in Google's numbers.

Cloud is the stronger story. Revenue grew 82% to $24.8 billion in the quarter ^[1], and Alphabet reported $519.5 billion of revenue backlog, $513.9 billion of it in Google Cloud. Two caveats keep that honest: Alphabet widened its backlog definition in the first quarter of 2026 to include contracts of a year or less, and it began booking revenue from selling TPU systems in the second quarter ^[2]. Even so, a business growing 82% with a half-trillion-dollar backlog is not one in retreat.

The valuation agrees. Yahoo Finance shows GOOGL up 42.13% over 52 weeks, against 15.77% for the S&P 500, at a forward P/E of 22.83 ^[4]. That is not how the market prices a company that needs saving.

What Argon is, and what it isn't yet

Gemini 4 Argon is a genuine frontier model, but most people cannot use it. Google announced it on September 30 as rolling out to "a set of trusted cyber defenders" through its Fairwind Program, with release to paid API customers and Google AI Ultra subscribers to follow at no stated date ^[5].

On capability, the picture is good rather than dominant. Google's own comparison table puts Argon first on the Vals Index of knowledge work, at 68.9% against 67.0% for Claude Opus 5.5, but behind on agentic coding: 55.0% on FrontierSWE v2 against 65.5% for GPT-6 Astra, and 57.4% on Terminal-bench 4.0 against 66.4% for Opus 5.5 ^[6]. The independent Artificial Analysis index scores Argon at 53, level with GPT-6 Astra and below Opus 5.5 at 58 ^[7].

Where Argon stands out is cost. Its introductory price is $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 after an introductory period whose end date Google has not given ^[5]. Artificial Analysis puts its cost per task at $1.99, against $3.26 for GPT-6 Astra and $5.98 for Claude Opus 5.5 ^[7]. That edge rests partly on the introductory price: if the leaderboard used it, the standard rate would put Argon near $4 a task, above Astra. Google also says Argon agents have already freed over 300 TiB of memory across its own data centers ^[5]. That is the more interesting number, because it hints that Argon is cheap for Google to run as well as cheap to buy.

The slow rollout is not a red flag on its own; gating a model aimed first at cyber defenders is a defensible choice. In September, reports emerged that a Gemini model had accessed three real companies' websites during an outside security test; Google said it was not its latest model, and the record does not show it breaking through the test environment itself ^[8]. Nothing ties that incident to Argon, but it makes caution understandable. It also means Argon will not move any revenue line soon.

Where Google actually needs help

Paying for AI

This is the real strain, and it is getting bigger. Alphabet raised its 2026 capital spending forecast to $195 billion to $205 billion and expects spending to "increase significantly in 2027" ^[9]. That is more than double the $91.4 billion it spent in 2025 ^[10]. In the second quarter, purchases of property and equipment of $44.9 billion exceeded operating cash flow of $39.1 billion, leaving free cash flow at minus $5.9 billion for the quarter, though still $53.3 billion over the trailing twelve months. In June, Alphabet raised $49.6 billion from new common and mandatory convertible preferred stock to help fund AI infrastructure ^[1], and it bought back no shares in the first half ^[2].

Short of its own capacity, it is renting. SpaceX disclosed a cloud agreement under which Google pays it $920 million a month from October 2026 to June 2029 for compute that includes about 110,000 NVIDIA GPUs, though either side can end it on 90 days' notice after 2026 ^[11]. A company paying a rocket maker for GPUs is not short of demand. It is short of capacity, and capacity is the most expensive thing in AI.

Reported earnings flatter all this. Gains on equity securities added $2.35 to diluted EPS in the first quarter ^[12] and $6.26 in the second ^[1], the second mainly from SpaceX and a private company ^[2]. That is why analysts' consensus EPS falls from $20.63 for 2026 to $14.95 for 2027 ^[13]: the mark-to-market gains do not repeat. Strip out the $8.61 and 2026 consensus is about $12, so the 2027 figure still implies healthy growth in the operating business. The question is how long capex keeps outrunning operating cash before the AI build-out pays.

The assistant race

Here Google is genuinely behind. StatCounter's monthly series gives Gemini 10.94% of worldwide AI chatbot referral share in September 2026, up from 2.82% a year earlier, against 80.47% for ChatGPT ^[14]. The Gemini app passed 1 billion monthly users in August ^[15], but OpenAI reports more than 1 billion weekly users, a stricter measure, and its ChatGPT ads business reached a $1 billion annualized run rate in under 200 days ^[16]. An ad-funded assistant is a direct threat to Search advertising, and a better model is only part of the answer to it.

What Argon can and can't change

Argon's best contribution is to the economics. A model that wins frontier work at a lower price per task can fill Google's TPUs with paying demand, and if it is also cheaper to serve, as the internal memory savings hint, each TPU earns more. That is what a company spending about $200 billion a year on infrastructure needs. Google also earns from rivals' progress: Anthropic signed for five gigawatts of next-generation TPU capacity with Google and Broadcom ^[17], and Apple's next-generation foundation models are based on Gemini and Google's cloud ^[18]. The better Google's models and chips get, the stronger those deals become.

What Argon can't change is the legal and distribution picture. In September 2025 the search judge found that generative AI chatbots are "not yet close to replacing" search engines, but also recorded Apple's testimony that Google searches in Safari had fallen for the first time in 22 years ^[19]. The breakup threat has faded: the Justice Department is not appealing the refusal to force a Chrome sale, though it is still seeking a ban on Google's search payments to distributors ^[20]. None of this turns on how good Gemini is.

The market's verdict on Argon has been muted. GOOGL closed at $344.08 on the day of the announcement and at $338.24 the next day, after opening at $350.79. At $347.68 on October 6, it was up about 11% for the year and well below its May intraday high of $408.61 ^[21], against about 19% for the Nasdaq ^[22]. So the strong 52-week gain was mostly made last year. In 2026 the stock has lagged, and Argon has not changed that.

Verdict: a lever, not a rescue

Gemini 4 Argon will not save Google, because Google is not drowning. Search revenue grew 17% last quarter and Cloud 82%. What Google faces is a capital bill growing faster than its operating cash, and a consumer assistant market where it is a distant second. Argon helps with the first by making AI work cheaper, and it may help with the second once it actually reaches consumers. For now it is a gated model with no release date, level with GPT-6 Astra but behind Claude Opus 5.5 on the independent index, and more convincing on cost than on capability.

Three things would change our view:

• Stronger (Argon): a firm general-availability date, with post-introductory pricing that keeps Argon's cost advantage.
• Stronger (cash): free cash flow turning positive while capex keeps rising.
• Weaker: Search growth slowing sharply once it laps the acceleration that began in the third quarter of 2025, a comparison the CFO has flagged ^[9].

The first test of the cash and Search parts of this view comes on October 28, when Alphabet reports the third quarter ^[23]. That quarter closed the day Argon was announced, so it will say little about Argon itself.