Large Cap tech AI strategies in flux.
Where will the pieces land?
The large tech companies are currently investing a total of over $200bn per year in AI datacentres and infrastructure. They may do this for the rest of the decade. However, recent events indicate they are not happy about their AI business models, personnel and strategy. They are looking to shuffle the pack of cards.
What is going on?
Things are moving too fast. The companies realise their investments are not winning them too much in terms of technology advances. There are at least two reasons for this.
First, as they are all investing billions, it is hard for any one player to pull ahead of the pack.
Second, there is rapid depreciation in the value of any advances made. The tech writer Kevin Kwok has described it thus
Tech, and especially AI, is increasingly deflationary. Every year the advances in AI are obsoleting the last year’s models. The knowledge gleaned from training the last generation of models or from building products that best utilized them might be essential for working with the latest models–but the actual old models or products will be outdated fast.
Conversely, as it gets easier to build software every year, the value of owning the legacy codebase falls or can even go negative.
This rapid depreciation indicates the brutal economics of the AI investment race. In order to prosper, companies believe you have to move very fast and you must hire people who are not just academic researchers but aggressive CEOs within their domain. There is increased demand for a few of the very best people who have these characteristics
Recent news headlines are evidence of the flux in AI world.
Microsoft and OpenAI are in very difficult re-negotiations of their partnership agreement.
Meta is paying very large packages to recruit AI talent.
Google is responding to the perceived threat to its core search by hiring talent from OpenAI among others.
Apple is widely believed to have fallen behind in AI and there are rumours they will engage inorganic expansion such as an acquisition of Perplexity Labs or Anthropic.
Microsoft and OpenAI
Tensions between OpenAI and Microsoft (MSFT) over the future of their AI partnership are flaring up.
OpenAI wants to loosen Microsoft’s grip on its AI products and computing resources and secure the tech giant’s blessing for its conversion into a for profit company. Microsoft’s approval of the conversion is key to OpenAI’s ability to raise more money and eventually go public.
Microsoft’s relationship with OpenAI began in 2019 with an investment of just $1bn. In total, it has invested $13.75bn in return for a 20% share of revenues and profits up to a total of $ 92bn. [1] If this is achieved Microsoft would see a return of at least 6.7X.
Microsoft Azure provides cloud hosting for Open AI’s rapidly growing business, which is currently at about $1bn of annual revenue, for Microsoft and likely to grow fast. This is subject to a 20% revenue rebate. Given likely margins, this is a highly profitable revenue stream for Microsoft, even after accounting for the rebate.
The agreement favours Microsoft. It may have been reasonable for OpenAI six years ago, but it now looks a poor deal given the exponential growth in OpenAI revenues and profits since then.
Open AI wants to renegotiate the agreement and look for a path to convert to a regular for-profit company. Microsoft is playing hardball. The negotiations have been reported as being difficult.
OpenAI’s executives have discussed what they view as a nuclear option: accusing Microsoft of anticompetitive behaviour during their partnership. This strategy could involve seeking federal regulatory review of the terms of the contract for potential violations of antitrust law, as well as a public campaign,
Microsoft fuelled OpenAI’s rise in exchange for early access to its technology, but their interests are now diverging and it is more difficult to find common ground.
The official statements have been optimistic but that may just be PR.
“We have a long-term, productive partnership that has delivered amazing AI tools for everyone,” representatives for the two companies said in a joint statement. “Talks are ongoing, and we are optimistic we will continue to build together for years to come.”
One subject of disagreement was OpenAI’s $3bn proposed acquisition of coding startup Windsurf. Microsoft currently has access to all OpenAI’s IP, as per their agreement. It offers its own AI coding product, GitHub Copilot, that competes with OpenAI. OpenAI did not want Microsoft to have access to Windsurf’s intellectual property. Microsoft objections means the acquisition did not happen.
Another dispute is over how much of OpenAI, Microsoft would own, if the former converts into a public corporation. Microsoft is currently asking for a larger stake (35%) than OpenAI is willing to give,
Under the current contract, MSFT has the exclusive right to sell OpenAI’s software tools through Azure and has preferred access to the startup’s technology. Microsoft has exclusivity as OpenAI’s only compute provider.
The two now compete on products ranging from consumer chatbots to AI tools for businesses. OpenAI wants to join with other cloud providers such as Google, Amazon and Oracle) so it can sell its technology to more customers and access additional computing resources.
Microsoft, meanwhile, wants access to OpenAI’s technology even after the startup declares its models have achieved humanlike intelligence [2],: This was originally designated a partnership-ending milestone. This is a strange aspect of the partnership agreement, and we will consider it in more detail.
As we understand it, the agreement will end when Artificial General Intelligence (AGI) is achieved. The contract only requires that OpenAI’s board declare AGI has been achieved, in good faith, though Microsoft could easily sue the company in response, risking a drawn-out legal battle.
The problem is there is no objective criteria to determine whether AGI has been achieved.
Microsoft is restricted from developing AGI on its own under the companies’ contract, which runs through 2030.
Some Microsoft executives had objected to the inclusion of the AGI clause in 2019, arguing it was arbitrary and unenforceable. However, at the time MSFT was so behind on AI at the time that they agreed to it.
Microsoft is hoping to remove the clause entirely from the contract as part of its recent negotiations, or, at least, secure exclusive access to OpenAI’s IP even after AGI is declared.
AI experts see AGI as the point at which generative AI systems achieve humanlike intelligence.
OpenAI executives including Sam Altman believe they are close to being able to declare that AI tools have achieved AGI.
Microsoft Chief Executive Satya Nadella has expressed scepticism that reaching such a benchmark is possible. This is not surprising since it is in MSFT’s interest the partnership continues, and MSFT has exclusive access to OpenAI’s IP.
Meta’s Mega hiring spree
Meta Platforms (Meta) has been aggressive in AI. It has invested billions of dollars in AI datacentres and on developing the Llama family of models.
Meta’s first version of Llama was produced by its Fundamental AI Research Team, which consisted of academics and PhD researchers.
11 of the 14 original researchers have left the company. The subsequent models have been developed by a different team and progress has been problematic
Two later-stage Llama models came under fire after it was revealed that Meta submitted different versions of the models to a third-party performance test than they delivered to the public.
The most recent LLM called Behemoth has been delayed. Meta engineers have struggled to improve significantly on the Llama 4 model.
Meta has claimed Behemoth outperforms similar technology from OpenAI, Google and Anthropic on some tests. The delay in its release suggest progress has been hard to achieve.
Mark Zuckerberg, the CEO of Meta has grown frustrated by the poor performance and has started hiring top talent from outside aggressively.
The first step was the $14bn acquisition of Scale.AI. The rationale was to secure the services of the latter’s CEO, Alex Wang.
As part of the deal, other Scale AI employees would also join Meta as part of a new effort to develop advanced AI dubbed “Superintelligence”.
Zuckerberg wrote on Facebook “For our superintelligence effort, I'm focused on building the most elite and talent-dense team in the industry. We're also going to invest hundreds of billions of dollars into compute to build superintelligence. We have the capital from our business to do this.”
Superintelligence according to Meta, requires a small number of very smart people and huge resources such as datacentres the size of Manhattan(!).
Zuckerberg wrote “We're actually building several multi-GW clusters. We're calling the first one Prometheus and it's coming online in '26. We're also building Hyperion, which will be able to scale up to 5GW over several years. We're building multiple more titan clusters as well. Just one of these covers a significant part of the footprint of Manhattan.”
Antitrust scrutiny has made it difficult for large tech to acquire companies. They are instead resorting to large minority investments in startups that allow them to hire their key employees. This has been dubbed an “acquihire” strategy.
This is not new. Most large tech companies have been doing it in the last two years. Google paid $2.7bn to hire its former researcher, Noam Shazeer, and license technology from his startup, Character.AI in 2024. Microsoft and Amazon have also struck similar deals. Nevertheless, Meta has raised eyebrows with the scale of its activity.
Zuckerberg has become the company’s recruiter-in-chief. He has been approaching leading engineers at rival firms on WhatsApp. One person assumed it was a hoax and didn’t respond for several days. He believes an email from him is a more powerful weapon than outreach from a faceless head-hunter.
The Meta CEO gave an interview with The Information in which he outlined the rationale for his recruitment strategy.
“I think that the physics of this is you don’t need a massive team to do this. You actually kind of want the smallest group of people who can fit the whole thing in their head. So there’s just an absolute premium for the best and most talented people.”
“I think we’ll see how the technology trends, and we’ll see what the results are. In running the company, I’m sort of always looking for ways that I can convert capital into a higher-quality service for people.”
“And one of the benefits of reinforcement learning is it gives you a venue to, you know, potentially convert very large amounts of capital into a better and better service, and potentially a better service than other less well-funded or less bold competitors will be able to do so. I view that as a competitive advantage.”
“I think we’re going to have the largest compute fleet of any company, and focusing on that on being powered by a small and talent-dense team, I think we’re gonna have by far the most compute per researcher to do leading edge work.”
He strongly believes Meta will be an attractive place for the brightest researchers.
And one of the biggest is that you can just have more leverage as a researcher. You have more compute right?.. when I was recruiting people to different parts of the company, you know, people are like, OK, what’s my scope going to be? And, you know, here, people say, I want the fewest number of people reporting to me and the most GPUs. And so having basically the most compute per researcher is definitely a strategic advantage, not just for doing the work, but for attracting the best people.
Mark Zuckerberg has tasked the team, which will consist of about 50 people, to achieving “superintelligence.”
It was reported that some found the concept vague or without a specific enough execution plan beyond the hiring blitz.
Given his control over the voting shares of Meta, Zuckerberg can move faster than most other CEOs.
He is offering hundreds of millions of dollars, sums of money that would make them some of the most expensive hires the tech industry has ever seen.
In addition to Scale and Wang, Meta has also offered to buy a minority stake in funds of the venture firm, NFDG, founded by two of its key artificial intelligence recruits. This aim is to recruit NFDG founders, Nat Friedman and Daniel Gross.
Meta has also hired Lucas Beyer, Alexander Kolesnikov and Xiaohua Zhai, who were all working in OpenAI’s Zurich office. Prior to that they worked together at Google DeepMind.
Not all Zuckerberg’s approaches have been successful.
Meta approached OpenAI co-founder John Schulman and Bill Peebles, the co-creator of OpenAI’s Sora video generator. They have tapped OpenAI co-founder Ilya Sutskever and invested in his new startup Safe Superintelligence recently. None of these have joined Meta.
OpenAI CEO Sam Altman says his best people remain at his company. OpenAI has given counteroffers to people Meta has tried to poach, promising them more money and scope in their jobs if they stay. The effort has not been cheap. The Information reports that OpenAI expects to spend $6bn on stock compensation and another $1.5bn on additional cash compensation in 2025 as it tries to hang onto its people.
One possible reason for candidate caution is that Meta’s chief AI scientist is a sceptic of the current strategy. Yann LeCun, who started Meta’s AI research division in 2013, does not believe the LLMs currently being built will get the world to AGI.
In summary, Meta is paying large sums of money to hire the best AI researchers and developers. These people are not pure academics but start up CEOs.
Meta’s strategy is to take a “talent-dense” handful of researchers, surround them with a large GPU supply and subject them to minimal management responsibilities.
The hope is its AI researchers operating with founder-level autonomy utilising their large GPU fleet will generate breakthroughs faster than the competition.
Time will tell whether this the right strategy. Meta stock has been stable, in recent weeks which suggests that, so far, the market is giving the Meta mega hiring strategy the benefit of the doubt.
Alphabet (GOOGL)
GOOGL’s share price and valuation has been relatively low in recent. One reason is a worry the rise of AI is a threat to its critical Search revenues.
However, Google is not standing still. It has incorporated more AI in its Search function. It is also investing a lot in AI products including the Google Gemini multimodal LLMs.
Goggle hired the Windsurf CEO and a group of senior employees while taking a licence on its technology for $2.4bn. This was after OpenAI’s effort to take over Windsurf failed due to Microsoft’s objections. Google is not investing in Windsurf equity. Most of Windsurf’s existing employees will remain at the company.
Apple (AAPL)
Apple’s stock has fallen 20% stock this year so far, the worst run over this period since 2010.
Apple is facing a lot of issues.
Apple devices are manufactured outside the USA, in China, India and the Far East. US sales (42% of the total sales) will be hit by the high tariffs proposed by the Trump administration.
It is facing court cases about payments to app developers and payments received from Google.
In addition, Apple is believed to be behind in the AI race. It is tough for them given their hardware-led strategy.
Apple Intelligence introduced last year is a work in progress. The Siri digital assistant is awaiting a promised AI makeover. An AI-powered Siri might not hit the market until late 2026: two years after Google Assistant got its first major AI overhaul.
The other large tech companies can distribute AI through their massive global networks which are designed to deliver their core business services. Apple must distribute AI through its devices, which account for 75% of its revenues
“On-device AI” hasn’t yet proved to be a major selling point for PCs, Smartphones and other devices.
Apple’ does not have its own cloud infrastructure and so requires powerful allies.
Apple’s business model needs AI to be a selling point for its devices. The delayed Siri means iPhones is unlikely to get any sort of AI boost in the next years.
Credible press reports suggest that Apple has talked to Perplexity and Anthropic about taking them over to jump start their AI capacities.
Conclusion
The massive investments being made in AI are not generating required technological advances.
Management are frustrated and looking to change their business models and upgrade their personnel.
OpenAI has made progress and achieved much success but is now chafing against the restriction of their 2019 partnership agreement with Microsoft.
Mark Zuckerberg is not happy with progress made by Meta and is aggressively leading the charge to build a new core team to achieve “Super Intelligence.” Meta is offering very high remuneration for the right people.
Google is making steady progress, but its CEO cannot move as fast as Meta Platform as the latter’s CEO has effectively control of the company thanks to a dual-share structure.
Apple hardware-driven business model puts them at a great disadvantage, and they are widely believed to be behind in the AI race compared to their rivals.
As the quarterly reporting season begins we look forward to hearing what the managements of these companies have to say about AI strategies on their earnings conference calls.
[1] When Microsoft first invested, OpenAI was a small research-focused non-profit. Today it is a fast-growing entity generating billions of dollars in revenues and profits.
[2] This point is generally known as artificial general intelligence or AGI.