The Age of the Pseudo-Acquisition?
What do CharacterAI, Inflection, and Adept tell us about the M&A and AI markets?
Please subscribe, and share the love with your friends and colleagues who want to stay up to date on the latest in artificial intelligence and generative apps!🙏🤖
Last Friday Google acquired the buzzy AI app Character.AI for $2.5B. Well actually no, it didn’t.
According to the Information:
On Friday, after chatbot developer Character announced that its founders and some other employees were going back to Google, Character told its roughly 130 employees that Google would pay a licensing fee (ostensibly for Character’s technology) that effectively valued the company at about $2.5 billion, as Kalley, Natasha and Stephanie first reported.
So basically the Character.AI co-founders Noam Shazeer and Danaiel De Freitas struck a deal where they would rejoin Google (where they had worked until 2021, when they left to found Character.AI) alongside “more than two dozen Character AI researchers.” Google will also sign a non-exclusive license to access Character.AI’s models and buy out venture investors at “around a $2.5B valuation”. So while this arrangement isn’t technically an acquisition (where Google would assume the entire company, business assets, intellectual property, etc.), it certainly feels like something close to it.
As industry observers have noted, this is the third “pseudo-acquisition” made in the AI space in recent months.
In March, Microsoft paid Inflection $650M for a “licensing deal” and hired Inflection’s cofounders and much of their senior staff (Inflection had raised north of $1.5B from investors to date).
In June, Amazon hired Adept’s cofounders and several employees while also striking a deal to license Adept’s technology. Semafor recently reported that investors in the company, last valued at $1B, would receive their money back.
So why are we starting to see this novel type of deal-making happen in AI, and should we expect more? (The short answer is yes, and we’ll explain our reasoning below).
Quick Disclaimer - we are not investors in any of these companies and do not have any insider information one way or another. We are purely speculating as keen and interested industry observers :)
Tech M&A Has Become Difficult
M&A has traditionally been a rich part of the technology industry. The five-year stretch from 2015-2020 saw some of the largest acquisitions of all time:
Dell acquiring EMC for $67B (2015)
Salesforce acquiring Slack for $28B (2020)
Microsoft acquiring LinkedIn for $26B (2016)
Not to mention the many hundreds of sub $5B acquisitions that barely made the headlines.
As you can see above, the high-water mark for tech M&A peaked in 2021, when $800B of deal value was transacted across 4K+ deals. That figure then decreased sharply in 2022 and 2023, and 2024 year to date is similarly off to a slow start.
Why? Certainly a cooling economic climate plays a role. However one of the major factors is the regulatory environment making M&A, particularly initiated by “BigTech”, extremely difficult. Regulators ranging from the DOJ to the FTC have scuttled several high-profile transactions, including the $20B Adobe/Figma deal, Illumina’s acquisition of Grail for $8B, and Amazon’s purchase of robot vacuum maker iRobot for ~$1.3B.
This leads to an environment where acquirers feel the traditional path to buying companies, either for the company itself or for particular talent (i.e. an acquihire), is no longer one they intend to pursue.
So Why Are These Deals Happening?
We see a few main reasons.
1. BigTech wants to protect and expand its dominance without upsetting the regulatory applecart.
BigTech companies - notably Microsoft, Amazon, and Google - recognize that their massive cloud businesses will benefit from growth in the AI ecosystem, but that they also need to protect their entrenched dominance. They want to own the entire stack, ranging from specialized chips to the application layer. Traditionally they would have attempted to acquire into areas where they lacked presence, but in a hostile M&A environment, they need to find workarounds.
This means hiring instead of acquiring, and licensing instead of buying.
Now, even these novel maneuvers are coming under the crosshairs of regulators. In July several US Senators called for an investigation into these practices, accusing the tech giants of finding yet another way to skirt around regulations.
The Cloud Giants (AWS, Azure, and GCP) all reported Q2 earnings in the last week, and the three businesses combined make up $220B+ of run rate revenue (!). It is no surprise that they would be looking to keep their toehold position. See figures below from Q2 ‘24, provided by our friend Jamin Ball from Clouded Judgement.
AWS (Amazon): $100B run rate growing 19% YoY (last Q grew 17%)
Azure (Microsoft): ~$81B run rate (estimate) growing 30% YoY (last Q grew 31%)
Google Cloud (includes GSuite): $41B run rate growing 29% YoY (last Q grew 28%)
The cloud giants want to protect their incumbent status, and ultimately grow their position, by partnering on innovation and training the foundation models themselves. Building these models requires an immense amount of talent and knowledge.
2. In AI, talent is > > > everything
Another reason we are seeing these giant price points for talent is that in the AI race, top talent is extremely scarce and more in demand than any other resource. While fads change and the demand for Character.AI’s specific product may recede, there are very few people in the world who understand AI better than Noam and Daniel. It is unclear what the future of agents will look exactly like, but David Luan is in the top 1% of AI engineering talent.
So while the headline prices that the acquirers pay for this talent looks immense at first glance, it is not without reason. If Microsoft/Amazon/Google believe AI can add trillions to their market caps over time, they are comfortable paying billions to attract the right teams to join their ranks (and consequently weaken their competitors).
During the mid-2000s, software engineering salaries surged due to the proliferation of smartphones and cloud computing, which created high demand for engineers with specialized knowledge. Today, a similar trend is occurring with AI. There are very few people who have true depth in AI, particularly at a PhD/academic level, and that makes this group even more sought after by large players.
Over the last decade, we have seen many talented software engineers elect to join early-stage start-ups, attracted by enticing equity packages and the opportunity to work on small high-impact teams. However, over the last two years, some fatigue has set in. With limited liquidity in the market for many SaaS companies, sought-after employees are returning to Big Tech, where they can receive substantial equity packages, access to major computing resources, and underlying stability. This “boomerang effect” is now in full swing.
3. Bloated valuations paired with low PMF will make fundraising difficult
Building a startup is just hard. Its even harder in an environment that looks very different from the one in which you first started raising money. In 2022 and 2023, especially in the wake of ChatGPT, the fundraising market for AI was frothy and plenty of startups and founding teams were thrown cash despite barely having a business plan. Now ~2 years later, many of those same companies are sitting on bloated valuations while still not having found the vaunted “product market fit”. This makes fundraising even more difficult, and can propel a vicious spiral where founders must decide whether to keep building, return capital, or find a soft landing.
This isn’t a commentary on Character, Adept, or Inflection specifically, but we expect many unicorns to sell in the coming months and quarters as cash begins to run out and options dwindle. We were quoted in a recent Information article that we expect several “fire sales” to occur in an environment where growth slows down and prior valuations become unsustainable to support.
This is particularly true for companies training their own models. We are recognizing that this is an incredibly expensive game, and few outside of the largest model providers (e.g. OpenAI, Anthropic, Meta) have the resources to make the unit economics work.
Conclusion
In conclusion, we greatly expect that both BigTech as well as large private acquirers (e.g. Databricks, OpenAI) will attempt to find creative ways to hire the best talent and access innovative technologies without tripping regulatory lines. On the other side of the equation, as startups spend large sums of money and feel they no longer have a sustainable path to independent success, their openness to “finding a home” becomes more pronounced.
In other words - get ready for more “pseudo-acquisitions” to hit AI in the coming months!
We hope you enjoyed this edition of Aspiring for Intelligence, and we will see you again in two weeks! This is a quickly evolving category, and we welcome any and all feedback around the viewpoints and theses expressed in this newsletter (as well as what you would like us to cover in future writeups). And it goes without saying but if you are building the next great intelligent application and want to chat, drop us a line!