An alliance of arch rivals and $14 billion recruiting: how the AI industry kicked off the summer

The AI industry began 2025 with the announcement of the unprecedented $500 billion Stargate mega project, and then followed it up with an equally unprecedented $1 trillion drop in markets following the release of China's DeepSeek AI model, which turned out to be just as good and significantly cheaper than its American competitors. We are, of course, fortunate to witness in real time a technological revolution comparable in scale to the invention of the Internet. Apparently, there's a reason why the word «unprecedented» is mentioned 64 times in a report on AI by renowned technology analyst and investor Mary Meeker (though I teased about it when wrote about her last week). And now, the year that started so brightly hasn't even reached the halfway point, and the ever-burgeoning AI market continues to be shaken by events that, in the apt expression of TechCrunch, have investors and analysts «scratching their heads.»

On June 10, Reuters dropped a bombshell of information on the market - it reported that OpenAI, creator of the most famous and advanced AI GPT, agreed with Google to use the Google Cloud service to train its AI models.

Analysts at Scotiabank called this development «somewhat unexpected,» Reuters writes. The agreement, which is certainly a victory for Google in the cloud computing market, at the same time means that the company with its computing power will support the worst competitor - OpenAI, with which it is fighting for leadership in the AI market;

But that's not the worst of it - AI could also threaten Google's monopoly on Internet search, which thanks to advertising links generates $175 billion in revenue a year for the company.

Last month, Apple's vice president of services, Eddy Cue, made a lot of noise when he said that for the first time in 22 years, the number of Google searches in the Safari browser had fallen. He attributed this to the fact that users are increasingly using AI to search for information, which understands the essence of the question and gives the answer immediately, rather than a bunch of links, half of which are advertising. Analysts from eMarketer predicted that Google's share of the U.S. search advertising market could fall below 50% for the first time in a decade, Business Insider wrote last year. And on top of that, younger generations no longer use the verb «google» as a synonym for Internet search, indicating a profound cultural shift, the publication added.

«The deal ... underscores the fact that the two companies are willing to ignore serious competition from each other to meet huge computing needs,» Reuters quoted Scotiabank analysts as saying.

There's another aspect to it - the deal with Google is, to put it bluntly, a spit in the face of Microsoft, which invested $1 billion in OpenAI in 2019, when no one outside of a narrow circle of specialists knew or heard of the company and its CEO Sam Altman;

In 2023, Altman was attempted to be ousted from OpenAI in a failed corporate coup attempt, and Microsoft CEO Satya Nadella supported him by announcing that he would then hire him to head a new AI lab. About 700 OpenAI employees said they would also leave for Microsoft following Altman.  In the end, the «mutiny» failed and Altman remained at the helm of OpenAI.

In addition, Microsoft competes fiercely with Google on several fronts - from the same internet search to the introduction of AI tools based on licenses from OpenAI and the market of cloud services. The latter is dominated by Amazon and Microsoft, while Google Cloud is in a catch-up position, and using unfair practices in doing so, as Microsoft's general counsel Rima Eliley alleged last year. But apparently, OpenAI's new alliance with Google is, as they say, just business, nothing personal. Microsoft, OpenAI and Google have not yet commented on the deal, writes Reuters.

Now let's move on from the drama, on the contrary, to a happy reunion. At the very end of May, Anduril, a company that makes various kinds of innovative AI-based weapons, announced that it had formed an alliance with Meta to create a cutting-edge augmented reality system that soldiers will be able to use in combat to gain a variety of tactical information: the position of vehicles, drones, allies, enemies, and so on. «This integration will transform how warfighters see, experience and integrate battlefield information, providing immersive technology solutions that improve tactical decision-making in combat scenarios,» Anduril said in a statement.

In principle, a logical alliance: Anduril specializes in weapon systems and combat AI, while Meta has been investing billions of dollars in virtual reality for years and developing its own family of large AI models, Llama.

But there is a nuance: Anduril co-founder Palmer Luckey in 2014 for $2 billion sold his VR startup Oculus to Facebook (now Meta) and went to work there himself. And in 2017, the company fired him for supporting Donald Trump, who was running for his first presidential term. And not even Trump himself, but just an initiative group that was spreading pro-Trump memes online.

Lucky himself back in February of this year described those events on his blog as nothing less than «exile from Silicon Valley because of sneaky snakes.»

However, he managed to return to the Valley as co-founder of Anduril, and now he and Meta CEO Mark Zuckerberg are showering each other with compliments;

«We are proud to partner with Anduril to help bring these technologies to the U.S. military personnel who protect our interests at home and abroad,» quotes Zuckerberg on Anduril's website.

«I'm excited to be working with Meta again,» Palmer Lackey echoed him. - My mission has long been to turn fighters into technomages. And the products we create with Meta do just that.»

«It's time to end the fight,» wrote Lucky on Web X on May 29 and posted a photo of himself with Zuckerberg.

However, the reason for this suddenly rekindled love is quite clear - at stake is a $22 billion contract with the US Army, which Lucky managed to take over from Microsoft. The latter will remain in the deal, but only as a provider of cloud computing power, TechCrunch writes. As a sign that the hatchet is finally buried, Anduril has even activated its Page on Facebook, the portal snickers.

That's not all the news about Mark Zuckerberg, however. Aside from trying to revitalize his chronically lossy metasearch with the help of Lucky and military money, he seems to be seriously concerned about Meta's lagging behind in the AI race. Llama's latest models have received a less than warm reception in the market, with users disliking them and analysts criticizing performance, wrote Bloomberg.

According to agency data, as of mid-May, referrals (links from the chatbot to other sites) were 80.6 percent coming from GPT, 11.3 percent from Perplexity and nearly 5 percent from Microsoft's CoPilot. Launched only in April, the MetaAI chatbot has not yet appeared in the statistics, but is already being criticized for privacy issues.

Zuckerberg went into «founder mode» and decided to take the problem solving into his own hands. His ambitious goal is to be the first to create AGI, a general artificial intelligence that is expected to be able to solve most tasks as well as or even better than humans.

To that end, he has personally taken up recruiting - at his homes in Lake Tahoe and Palo Alto, Zuckerberg hosts dinners with potential candidates for the team - AI development stars, Bloomberg writes. He has created a special Whatsapp chat room called «Recruiting Party» for Meta's top executives, where potential candidates for the working group, dubbed the «Superintelligence Group,» are discussed day and night.     

«Zuckerberg intends to hire about 50 people for the new team, including a new head of AI research... He has rearranged desks at the company's Menlo Park headquarters so the new hires will sit next to him,» Bloomberg quoted sources who wished to remain anonymous as saying.

It looks like the head of the new superintelligent team will be Alexander Wang, co-founder of Scale AI, a startup that focuses on data partitioning for AI - the foundation of big model training. As he said himself, «We sell picks and shovels during the gold rush». 

How much Zuckerberg wants to attract new talent is evidenced by the fact that for a 49% stake in Scale AI Meta agreed to pay $14.3 billion - one of the largest investments in Zuckerberg's company history, or maybe just the largest because it paid $19 billion for Whatsapp in 2014 but bought the messenger outright.

«There's no shortage of investors and founders left scratching their heads this weekend over Meta's latest acquisition,» surprised TechCrunch. . on June 11;

And the clue, it seems, is simple. As written on the X network by Wang himself, the condition of the deal was his transfer to Meta, he was proud and happy to work at Scale AI, but now he got a unique chance, he has a new stage and so on, which is usually written in such letters. That is, in simple language instead of corporate language, Zuckerberg paid $14.3 billion without blinking an eye to get a star executive to join his team. Now that's recruiting! To put it bluntly, not even «in funder mode», but royal.