Alphabet's stock dropped more than 6 percent on Monday and continued under pressure Tuesday — its worst stretch in more than a year — as investors absorbed a week that produced two departures from the upper ranks of Google's AI organization, landing against a backdrop of growing evidence that the company's search transformation is eating the ad revenue that funds its entire business.

On Thursday, June 18, Noam Shazeer, Google's vice president of engineering and co-lead of its Gemini model family, announced he was leaving to join OpenAI. On Friday, June 19, John Jumper, a Google DeepMind vice president who shared the 2024 Nobel Prize in Chemistry for developing AlphaFold, said he would join Anthropic. Two of the most recognizable names in Google's AI organization were gone within 48 hours — one to each of its two closest competitors. The cracks forming in Google's core business are not simply the result of who left. They exist because the company is trying to fight two wars simultaneously: one against AI rivals recruiting its best researchers, and one against its own product, which is quietly straining the advertising engine that pays for everything.

Why Google Is Fighting a Two-Front War It Did Not Choose

Google's dilemma is structural and partly self-inflicted. Ads account for roughly three-quarters of the company's revenue. That model depends on users clicking through search results to advertiser-backed pages. But when Google's AI Mode is active, 93 percent of queries end without a single click to an external website, according to a study by Seer Interactive analyzing 25.1 million AI Mode impressions. Across all Google queries — not just AI Mode — 68 percent now end without a click, up from roughly 45 percent a decade ago, according to Similarweb data analyzed by SparkToro.

The threat is both that Google loses market share to competitors and that, in deploying the AI tools it needs to stay competitive, it damages the revenue model those tools are supposed to protect. At its I/O 2026 developer conference, Google announced what it called the biggest redesign of its search box in 25 years, embedding AI Mode directly into the interface. The message to the market was clear: the old search paradigm is ending. What the ad economics of the new one will look like remains unresolved.

Lily Ray, vice president of SEO and AI search at marketing firm Amsive, warned that Google's I/O search overhaul would have a "devastating impact on the Internet," particularly for the publishers and independent websites whose traffic depends on Google-driven clicks. Condé Nast CEO Roger Lynch put it more directly in a recent interview: "Last year, I told our teams assume there's no search. You have to have your businesses plan as if search is zero."

When Google Paid $2.7 Billion and Still Lost

In August 2024, Google struck a deal valued at approximately $2.7 billion — structured primarily as a license for the technology of Shazeer's AI startup Character.AI — that also brought Shazeer and a group of researchers back into Google DeepMind. Google installed him as vice president of engineering and co-lead of Gemini, with responsibility for the foundational pretraining phase of model development — the stage where the model learns from massive datasets before fine-tuning or deployment. Multiple people inside Google DeepMind credited him with meaningful improvements to Gemini's performance during that period.

That deal produced a departure less than two years in. Shazeer joins OpenAI as its lead for AI architecture research — a title confirmed by OpenAI's chief research officer Mark Chen — at a moment when OpenAI is preparing for a public offering. OpenAI CEO Sam Altman said on X the morning of the announcement that Shazeer was "one of the people I have most wanted to work with since the very beginning of OpenAI." Acqui-hires have a documented failure mode: the most valuable assets can walk out the door once retention periods end. In this case, Google spent $2.7 billion, kept Shazeer for 22 months, and watched the institutional knowledge he accumulated about what works at Gemini scale join a direct competitor.

How the Transformer Authors All Left Google in the End

What makes Shazeer's departure especially significant is not just the cost of retaining him. It is the historical pattern his departure repeats.

Shazeer is a co-author of the 2017 paper "Attention Is All You Need," which introduced the Transformer architecture — the design that underpins virtually every major large language model today, including GPT, Gemini, and Claude. The paper has been cited more than 173,000 times and ranks among the ten most-cited scientific papers of the 21st century. Its core technical innovations — multi-head self-attention, which allows a model to process all tokens in a sequence simultaneously rather than one at a time, and positional encoding, which tracks token order without sequential processing — are the engineering foundation of the current AI era.

After the paper was published, all eight of its co-authors eventually left Google. Several founded or joined the companies now competing most directly with it: Ashish Vaswani and others co-founded Adept AI; Aidan Gomez co-founded Cohere; Illia Polosukhin co-founded Near Protocol. Shazeer himself left Google in 2021, founded Character.AI, was brought back in 2024, and has now left again. The architecture that made Google the dominant AI company of this era was built by people who did not stay — and the current talent exodus follows that pattern with an explicit competitive dimension absent in 2017.

"The race at the frontier right now appears between Anthropic and OpenAI," D.A. Davidson analyst Gil Luria told investors in the days following the departures — a comment that reflects a broader market question about whether Google's enormous resource base can compensate for who is leaving through the door.

John Jumper Won a Nobel for Google's Science, Then Left for Anthropic

John Jumper's departure follows a different but equally pointed logic. Jumper co-created AlphaFold, the AI system that predicted the three-dimensional structure of proteins for more than 200 million molecules — solving a core problem in structural biology that had resisted decades of research. He shared the 2024 Nobel Prize in Chemistry for that work with DeepMind CEO Demis Hassabis.

Jumper announced his departure on June 19 after nearly nine years at DeepMind. In his post on X, he credited Hassabis for "taking a real chance letting me lead the AlphaFold team just six months after finishing my PhD." Hassabis responded publicly: "What we achieved with AlphaFold changed the world, and showed the field what was possible with AI for science and medicine."

The warmth of the farewell did not soften what it signals. Anthropic has been building life-sciences infrastructure throughout 2026 — including wet laboratories, biological agent research, and partnerships with the Allen Institute and Howard Hughes Medical Institute — that aligns with Jumper's expertise. Bloomberg reported that engineers at DeepMind have been moving to Anthropic at a ratio of nearly 11 to 1, according to industry talent analyses, with Anthropic drawing staff specifically from a division where Google has struggled to build commercial traction in AI coding tools.

What AI Mode Actually Does to the Clicks That Pay for Everything

The zero-click problem requires some technical clarity, because the mechanism is counterintuitive. Google's AI Mode works by running an inference pass over indexed content at query time — the large language model reads relevant pages, extracts the key information, and assembles a synthesized answer that appears directly on the results page. The user receives the answer without visiting the source. From a user experience perspective, this is an improvement: fewer steps, faster answers. From the perspective of the click-based ad model that has sustained Google for two decades, the math is unfavorable.

Every time an AI Mode response resolves a user's query, the click that would have occurred — to an advertiser-backed result or to a publisher earning traffic — does not. Seer Interactive's study of 25.1 million AI Mode impressions found that 93 percent of AI Mode queries generate no outbound clicks. Across all Google search types, SparkToro and Similarweb's June 2026 data puts the overall zero-click rate at 68 percent — the steepest decade-of-change the researchers have measured.

That said, Google's ad business has not collapsed. Ads now appear in 25.5 percent of AI Overview results, up from roughly 3 percent in January 2025, according to BrightEdge data. Google Search revenue grew 19 percent year-over-year in Q1 2026 to $60.4 billion. The zero-click structural pressure is real, but Google is simultaneously building ad formats for the AI era. The unresolved question is whether those formats can grow fast enough to replace what they are replacing.

Shazeer's specific technical contributions to this system were substantial beyond the 2017 paper. His later architectural work introduced Sparse Mixture of Experts, which allows a model to activate only a fraction of its parameters for any given token — dramatically reducing inference cost at scale — and Multi-Query Attention, which reduces memory bandwidth requirements during token generation, enabling faster inference without proportional hardware increases. Both are mechanisms running inside Google's frontier models today. That knowledge is now at OpenAI.

Google Earned Record Search Revenue in Q1 While Building the Tool That Threatens It

The financial picture is more complicated than the talent headlines suggest. Alphabet's Q1 2026 results were strong across the board: total revenue reached $109.9 billion, up 22 percent year-over-year, with operating income of $39.7 billion. Google Cloud exceeded $20 billion in revenue for the first time, growing 63 percent from a year earlier, with a backlog that nearly doubled to over $460 billion. Gemini's share of AI chatbot web traffic grew from 5.7 percent to 13.1 percent over the past year, according to First Page Sage's market analysis — the largest year-over-year gain of any major AI chatbot platform, driven by integration across Search, Android, and Workspace.

Google's global search market share remains at approximately 90.5 percent, based on May 2026 StatCounter data. ChatGPT Search processes 250 to 500 million weekly queries and Perplexity handles around 50 million, per Similarweb's 2026 AI Search report — meaningful volumes but a fraction of Google's roughly 14 billion daily searches. AI Mode itself crossed 1 billion monthly users, with queries more than doubling every quarter since its launch, according to Pichai's characterization at I/O.

None of that means Google is in free fall. Alphabet stock is still up more than 100 percent from a year ago, even after Monday's selloff. Analysts at Jefferies wrote after the departures that they "don't read the recent departures as a signal that Google is doing less with AI, but rather as another data point in an industry-wide war for talent." The war for talent and the war for the ad model are happening simultaneously, and Google is not winning either cleanly.

Google Search and the Alternatives: What the Past Week Changed

Google's traditional search market share has narrowed from a 2023 peak of 92.9 percent to approximately 90.5 percent today — modest in raw share terms, but consistent in direction. AI platforms are absorbing the informational and research queries that historically generated the most organic click traffic. ChatGPT Search, Perplexity, and Google AI Mode collectively handle hundreds of millions of queries per week that would have gone to a traditional results page four years ago.

DuckDuckGo's install rate spiked roughly 30 percent week-over-week in the days following Google's I/O 2026 search overhaul, with iOS installs peaking at nearly 70 percent. The privacy-focused engine's CEO, Gabriel Weinberg, described Google's approach as "force-feeding AI with no way to opt out." DuckDuckGo holds a fraction of a percent of global search share — Lily Ray of Amsive called it "microscopic" — but Microsoft launched a Bing extension called "Bing AI Search Choice" giving users the option to disable AI features, a direct acknowledgment of the same user frustration.

For publishers, marketers, and businesses built on organic search, the practical consequences are already visible. Google search traffic to publishers fell 33 percent globally in the year to November 2025, according to Reuters Institute and Chartbeat data. The ad revenue that web publishing depended on has not migrated to AI platforms — AI referrals remain a fraction of a percent of total site traffic. The economic model that sustained the open web for two decades is under structural pressure, and the tool applying that pressure is Google's own product.

Frequently Asked Questions

Is Google's search market share actually declining?

Traditional search market share figures show Google at approximately 90.5 percent globally as of May 2026, down from a 2023 peak of 92.9 percent — a real but modest erosion. The more consequential measure is what fraction of informational queries now get resolved inside an AI interface without a click. SparkToro and Similarweb data show 68 percent of all Google queries now end without a click to any external website, up from about 45 percent a decade ago. Google's raw share of search is largely intact; its share of the click traffic that funds web publishing and advertising is declining significantly faster.

What is the Transformer architecture, and why does Noam Shazeer's departure matter technically?

The Transformer is the neural network architecture introduced in the 2017 paper "Attention Is All You Need," which Shazeer co-authored with seven colleagues at Google. It replaced sequential language processing with a parallel mechanism called multi-head self-attention, which computes relationships between all tokens in a sequence simultaneously — a design that can be distributed across GPUs and enables the large model sizes that power today's AI systems. The paper has been cited more than 173,000 times and underlies every major large language model, including GPT, Gemini, and Claude. After the 2017 paper, all eight co-authors eventually left Google. Shazeer's latest departure follows that pattern — but this time he is joining the competitor most directly building against Google's frontier model position, and he is arriving with knowledge of what works at Gemini scale.

Does the talent exodus mean Google's AI is falling behind?

Not necessarily. Google DeepMind remains one of the best-funded AI research organizations in the world, and losing individual researchers — even landmark ones — does not immediately degrade the models they helped build. What the departures signal is that Anthropic and OpenAI are competing successfully for senior talent at the level of Nobel laureates and Transformer co-authors, and that the retention mechanisms Google has relied on — including multi-billion-dollar acqui-hire deals with extended periods — are insufficient to hold the most exceptional researchers once their commitments expire. Jefferies analysts noted the departures should be read as evidence of "an industry-wide war for talent" rather than a sign Google is technically retreating.

What can publishers, advertisers, and businesses do now that 68 percent of Google searches end without a click?

The strategic adjustment SEO professionals are advising is a shift from click optimization to citation optimization — what the industry is calling Generative Engine Optimization. Being cited in an AI-generated answer, even without generating a click, builds brand awareness and shapes purchasing decisions. Practically, businesses should audit which of their top revenue-driving queries now trigger AI Mode or AI Overviews, assess how much of their organic traffic model depends on those query categories, and begin testing structured content formats that AI systems can extract and cite. Condé Nast CEO Roger Lynch's advice to his team — to plan as if organic search traffic is zero — is an extreme formulation, but the directional logic is one every publisher and digital marketer should be running.