Futurism · September 2024 (per reporting) "Google paid $2.7 billion to get a single AI researcher back" — an unprecedented industry re-hire over the DeepMind return of Transformer paper co-author Noam Shazeer.
The one takeaway from this article: the Character.AI soft acquisition is the third in 2024, and with three such cases inside six months, "hyperscaler consolidation as the standard pattern of frontier competition" was established. The reported licensing figure of $2.7 billion is the largest of the three, signaling Google's unusual willingness to pay for the re-hire of one Transformer paper co-author.
Industry Context — Why Character.AI is retro No. 3
Character.AI was co-founded in 2021 by Noam Shazeer (21 years at Google, Transformer paper co-author, in the first 100 employees of the company) and Daniel De Freitas (formerly of Google's LaMDA team). Total funding around $193M (with participation from Andreessen Horowitz and others), valuation around $1 billion as of March 2024. A character-based AI chat product letting users converse with any character (historical figures, anime characters, self-created personas, etc.). Monthly active users reportedly reached the tens of millions.
The founding story is important industry context — Shazeer had been the core researcher developing LaMDA (the predecessor to Bard / Gemini) inside Google in 2021, but resigned in frustration when Google leadership refused to release LaMDA externally, and co-founded Character.AI with Daniel De Freitas. In other words, Character.AI was an independent-startup implementation of "the product Google would not ship" in 2021. Three years later, in 2024, Google paid $2.7 billion to bring Shazeer back — an unprecedented industry case.
On August 2, 2024, Google and Character.AI issued same-day official statements — a non-exclusive licensing agreement between Google and Character.AI + the return of Shazeer + De Freitas + key researchers to Google DeepMind. The remaining Character.AI organization installed Dominic Perella (former General Counsel) as interim CEO and announced a pivot to "a consumer chatbot using third-party LLMs (models not developed in-house) alongside its own." Structurally identical to Inflection-Microsoft (March 19, 2024) and Adept-Amazon (June 28, 2024) — (a) a hyperscaler absorbs key technical staff via license + hire, (b) the remaining company continues as a separate business, (c) M&A regulation under antitrust is, formally, avoided.
Timeline (primary-source-based)
- Around 2000: Noam Shazeer joins Google (in the first 100 employees)
- June 2017: Transformer paper 'Attention Is All You Need' Published June 2017, arXiv:1706.03762, accepted at NeurIPS 2017. Eight co-authors: Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, Illia Polosukhin. The paper that formalized the Self-Attention mechanism — the foundation of modern LLMs (GPT, Claude, Gemini families) — and achieved SOTA on the WMT 2014 machine translation benchmark of the time. As of 2024, with over 120,000 citations, it is one of the most influential papers in AI. published, Shazeer one of eight co-authors who established the foundational technology
- 2017–2021: Shazeer leads Mixture of Experts (MoE) / Switch Transformer / LaMDA and other key Google research
- 2021: Shazeer + De Freitas resign in frustration when Google leadership refuses to release LaMDA externally, and co-found Character.AI
- 2022–2024: Character.AI grows rapidly as a consumer chatbot product, reportedly reaching tens of millions of MAU, total funding $193M, valuation $1B (as of March 2024)
- August 2, 2024: Google + Character.AI same-day official statements — Shazeer + De Freitas + key researchers return to Google DeepMind, Google and Character.AI sign a non-exclusive licensing agreement. Dominic Perella (former General Counsel, previously a Snap Inc. executive) installed as interim CEO. The reported licensing figure is $2.7 billion (sources: same-day reporting by The Information, WSJ, Entrepreneur — neither Google nor Character.AI has officially confirmed the figure)
- August 2024 – within 2024: The U.S. DOJ opened an investigation into the Google–Character.AI transaction (whether the acquihire avoids M&A regulation under antitrust law)
- 2025: Karandeep Anand appointed permanent Character.AI CEO (source: Character.AI official blog)
- 2025–2026: Character.AI continues operation focused on third-party LLM usage + user-base maintenance — but frontier model in-house development is halted
Why it disappeared (= exited the frontier) — structural analysis
Character.AI has not disappeared as a corporate entity, but it has exited frontier model in-house development. Three structural reasons MEMEX reads from primary sources:
(1) Rapid saturation of the consumer chatbot market. Character.AI held a first-mover advantage in consumer character chat in 2022–2023, but in 2024 the market saturated rapidly with ChatGPT persona / GPTs, the character expressiveness reinforcement in Anthropic Claude, Meta AI Studio (same kind of character chat feature), and other startup entrants. At a $193M funding scale, the economic logic of continuing frontier model development in a saturated market broke down.
(2) Imbalanced compute cost structure. Character.AI carried massive inference demand toward the consumer market, with structural double-bearing of per-user compute cost — (a) in-house frontier model training + (b) large-scale inference. An independent startup without hyperscaler backing carries this double burden under hard structural pressure.
(3) Strategic value of Shazeer himself to Google. The combination of Transformer paper co-author + first-100 Google employee + LaMDA core researcher is irreplaceable in the AI industry. As the Futurism framing — "desperate Google paid $2.7 billion to get a single AI researcher back" — spread in industry coverage, Google's strategic priority on Shazeer's DeepMind return was, in MEMEX's reading, itself worth $2.7 billion (editor's inference, with a hedge given that Google has not officially confirmed the figure).
What the industry foreshadowed — confirming the "pattern" with the third case of 2024
The Character.AI-Google case matters because three structurally identical cases — Inflection (March) + Adept (June) + Character.AI — landed within the same year. With the third case, the observational hypothesis becomes confirmed: "this is not coincidental, it is the industry's standard exit pattern." For the detailed structural analysis, see the 2024 soft-acquisition trio meta-analysis.
What MEMEX observes as distinctive about the Character.AI case is the reported $2.7 billion figure. Compared with Inflection's reported $650M and Adept's undisclosed amount, it's an order of magnitude larger. This can be read as Google rating the strategic value of Shazeer the individual + team accordingly high — and it became a reference point that influenced subsequent startup valuations as well. From the second half of 2024 onward, this also became a catalyst for the widening view among VCs evaluating independent frontier startups that "in the end, soft acquisition by a hyperscaler is the realistic exit route" (editor's inference, aggregating secondary reporting).
Editorial Observation — Character.AI as a MEMEX industry axis
(1) The way Shazeer returned indicates a shift in Google's internal recognition. Shazeer resigned in 2021 because "Google didn't release LaMDA." Three years later, Google paid $2.7 billion to bring him back — that fact shows a shift in Google's internal "strategic call on releasing frontier models externally." Consistent with Gemini's active expansion during 2024–2025.
(2) MEMEX's reservation about the long-term sustainability of consumer chatbots. The remaining Character.AI continues consumer chatbot operation with third-party LLM use, but whether a consumer chatbot that has given up frontier model in-house development can maintain its user base long-term is an open question. As Meta AI Studio and OpenAI GPTs catch up on features in the same consumer character chat space, the likelihood that Character.AI's distinct advantage shrinks is high. MEMEX takes the position of re-examining the state of the remaining Character.AI on a 12–18 month observation window.
(3) The reasons MEMEX is treating Character.AI in a retro article: completing the "pattern confirmation" as the third case of the trio + archiving the career trajectory of Shazeer, one of the eight co-authors of the Transformer paper. The latter is an extension of MEMEX's Karpathy line and Anthropic Fellows Program series — "people who built modern AI" — with archival value in tracing the industry's human network.
Related Resources
- Inflection AI End Archive (Soft Acquisition No. 1, March 19, 2024)
- Adept AI End Archive (Soft Acquisition No. 2, June 28, 2024)
- 2024 Soft Acquisition Trio Meta-Analysis
- Karpathy's Software 3.0 / Agentic Engineering
- Anthropic Fellows Program
- The Inverted Triangle of the AI Economy
- Person profile: Noam Shazeer (Google DeepMind, former Character.AI CEO)
Sources
Primary (official)
- Character.AI official blog: Our Next Phase of Growth (August 2, 2024)
- Character.AI official: Names Karandeep Anand as CEO
- Transformer paper "Attention Is All You Need" (arXiv:1706.03762)
- Google DeepMind official
Secondary (reporting — distinguished from official statements)
- TechCrunch: Character.AI CEO Noam Shazeer Returns to Google (August 2, 2024)
- The Information: Google Confirms $2.7 Billion Deal to Hire Character Co-Founders
- Entrepreneur: Google Rehires AI Pioneer Noam Shazeer in $2.7 Billion Deal
- SiliconANGLE: Google rehires founders of consumer chatbot startup Character.AI (August 2, 2024)
- Ctech: Google's $2.7B AI deal with Noam Shazeer's Character.AI draws DOJ attention
Editor's note: The licensing figure between Google and Character.AI ($2.7 billion) is an estimate from reporting by The Information / WSJ / Entrepreneur and others — neither Google nor Character.AI has officially confirmed it. This article follows the editorial policy of explicitly separating source tiers (official vs. reporting), handling the official statements (the existence of the transaction + the relocation of key people) and the reported estimate in separate paragraphs.