Woofun AI reports that Alphabet's stock price plummeted by more than 7% following a cascade of high-profile executive departures, resulting in a market capitalization loss exceeding $300 billion. The immediate trigger was the announcement on June 18 that Noam Shazeer, co-author of the seminal Transformer paper, left X to join OpenAI, followed two days later by John Jumper, the 2024 Nobel Prize winner in Chemistry and leader of the AlphaFold team, who departed Google DeepMind for Anthropic. Analysts have attributed this severe market reaction to a "talent drain," raising urgent concerns that Google is losing its competitive edge in the global AI talent war. This exodus represents a historic milestone: all eight co-authors of the 2017 paper "Attention Is All You Need" have now left Google, marking the first time the founding team has been entirely dispersed from the company that birthed the architecture.

Shazeer's latest move concludes a volatile tenure defined by his second exit from the tech giant. In 2021, he initially left Google to found Character.AI after becoming dissatisfied with the company's reluctance to release the chatbot technology he had developed. The situation reversed in August 2024 when Google acquired Character.AI's technology rights for approximately $2.7 billion, subsequently bringing Shazeer back on board as Vice President of Engineering for the Gemini project.

However, less than two years after this high-profile return, he departed again for OpenAI, where he is reportedly leading "Architecture Research" as the organization prepares for an initial public offering. This cyclical movement underscores the intense friction between internal research constraints and external entrepreneurial opportunities.

The dispersion extends to the core team of Essential AI, co-founded by Transformer co-author Ashish Vaswani and Niki Parmar in 2023. Rumors indicate that NVIDIA is absorbing this core team to develop its open-source model, Nemotron. Essential AI had secured $175 million in Series B funding in early 2026, achieving a valuation of $1 billion.

However, financing bottlenecks and the strategic involvement of AMD as an investor may have prompted NVIDIA's aggressive acquisition strategy. Several Essential AI researchers have already updated their LinkedIn profiles to confirm their new employment at NVIDIA, signaling a rapid transfer of intellectual property and human capital away from the Google ecosystem.

Niki Parmar, who co-founded both Adept AI and Essential AI alongside Vaswani, executed her own strategic pivot by leaving Essential AI in late 2024 to join Anthropic. At Anthropic, she has contributed directly to the development of the Claude 3.7 Sonnet model. Her departure from the startup she helped build highlights the fragmentation of the original team, with members scattering across competing entities rather than consolidating within a single new venture. This pattern suggests that the gravitational pull of established AI leaders like Anthropic is currently stronger than the allure of building new startups from scratch.

Jakob Uszkoreit, the researcher who proposed replacing recurrent neural networks with self-attention mechanisms, took a different trajectory by co-founding Inceptive in 2021 to apply AI specifically to RNA design. Inceptive has raised about $1.2 billion and recently signed a strategic partnership with Alnylam Pharmaceuticals. The deal includes an initial payment of $30 million and potential total collaboration funding of approximately $2 billion. This financial structure demonstrates how the original Transformer authors are leveraging their foundational knowledge to secure massive capital in specialized biotech sectors, far removed from general-purpose large language models.

Llion Jones, who was responsible for the initial codebase and inference efficiency optimizations, co-founded Sakana AI in Tokyo in 2023. The company focuses on evolutionary algorithms and has released the Sakana Fugu model. Sakana AI raised $379 million, including a Series B round in March 2026, and secured a multi-year cooperation agreement with Mitsubishi UFJ Financial Group (MUFG). The involvement of a major Japanese financial institution indicates a global diversification of the team's influence, moving beyond the traditional Silicon Valley corridor into Asian markets with significant institutional backing.

Aidan N. Gomez, the youngest author of the original paper, co-founded Cohere in 2019 with a focus on enterprise AI services. As of mid-2026, Cohere's annualized recurring revenue exceeded $200 million, supporting a valuation of around $7 billion. Gomez has been vocal about "digital sovereignty" and AI regulation, positioning his company as a leader in compliant, enterprise-grade solutions. His success illustrates a viable path for the team members who chose to build independent commercial entities rather than rejoin major tech conglomerates.

Łukasz Kaiser, who contributed to TensorFlow and the tensor2tensor framework, joined OpenAI in 2021 and contributed to Codex, GPT-4, and the o1 reasoning model. He recently stated that Transformers can solve any problem given sufficient intermediate reasoning steps. His presence at OpenAI, alongside Shazeer, reinforces the narrative that the most advanced reasoning capabilities are migrating toward organizations prioritizing open-ended research over product constraints. Illia Polosukhin, who designed the original Transformer model with Vaswani, co-founded NEAR Protocol in 2018. NEAR has raised over $5.3 billion, and Polosukhin positions the platform as the "settlement layer" of the AI economy. He asserts that future blockchain users will be AI agents and advocates for robust regulatory frameworks for autonomous AI agents, bridging the gap between decentralized finance and artificial intelligence.

The complete dissolution of the Transformer team at Google represents a structural realignment of the global AI landscape. With members now leading initiatives at OpenAI, Anthropic, NVIDIA, and various high-growth startups, the concentration of foundational AI talent has shifted from a single corporate silo to a distributed network of competitors. This marks the third such incident this year where a major tech firm has lost a critical mass of foundational researchers to external rivals.