AI and Big Five: Revisited

AI and the Big Five: A 2025 Re-evaluation

In January 2023, an analysis was made regarding AI and the Big Five tech companies: Apple, Google, Meta, Microsoft, and Amazon. After two-and-a-half years, it's useful to revisit that analysis and re-evaluate these companies, especially considering Meta's situation.

Meta's Llama 4 release was disappointing, leading Mark Zuckerberg to recruit AI talent. He has contacted researchers, scientists, infrastructure engineers, product experts, and entrepreneurs for a new Superintelligence lab, offering significant financial incentives, even discussing buying a startup. Some candidates are hesitant because of Meta's AI challenges and restructures. There were accusations of gaming a leaderboard to improve an AI model's appearance and delays in unveiling a new AI model.

Zuckerberg's vision for the AI superteam, tasked with achieving advances and reaching a point of “superintelligence,” has been met with concern, with some finding the concept vague.

In the original 2023 article, evaluations were framed using Clay Christensen’s framework of sustaining versus disruptive innovation. The question of whether AI obsoletes everything is considered. While the most dire predictions are deemed unlikely, dismissing AI as hype is also a mistake. Changes may align with the idea that the impact of AI is overestimated in the short term but underestimated in the long term.

Apple

Apple has had a period marked by Apple Intelligence, with basic on-device LLM capabilities and private cloud compute infrastructure, but it is not at the cutting edge of models or products. However, its core business isn't immediately threatened by AI. Apps like OpenAI and Claude are used on iPhones or in browsers. Apple's local LLMs could differentiate apps for Apple platforms, and it has access to consumer data for building semantic indexes for AI operation.

Apple's position is compared to Microsoft and the Internet. The company should deepen its partnership with OpenAI and focus on being the best hardware provider for the dominant consumer AI company. This involves building AI-powered devices beyond the phone and investing in robotics and home automation. The obstacle is Apple's integrated hardware and software approach, where hardware is differentiated by Apple's software. Apple's differentiation has shifted to hardware, and its chips could offer local AI capabilities, augmented by cloud AI capabilities. If Apple chooses to go it alone, it needs a major acquisition, such as Mistral, or to invest billions of dollars.

Google

Google's last two years have been better than anticipated, but its position and concerns are unchanged. Its infrastructure is considered the best, integrated from chips to networking to models, enabling capabilities like Gemini's context window size and attractive pricing. However, dependence on TPUs means competing with the Nvidia ecosystem. Gemini has improved and scores highly in LLM evaluations, but real-world usage lags behind OpenAI and Anthropic's models. Google leads in media generation, with Veo for video generation. Google's advantage is data, drawing from YouTube video and its web indexing position. It has distribution channels like Android, offering the potential for an integrated device and cloud AI experience to challenge Apple. The challenge is whether Google can make its devices better by controlling the model, operating system, and consumer data.

AI's disruptive potential for Search is a problem for Google's core business. Its focus on AI Search Overviews and the Search Funnel aims to make Search better and maintain monetization. Cloud computing is promising, leveraging Google's infrastructure and model advantages, and Google Cloud Platform focuses on multi-cloud workflows, with AI capabilities potentially attracting enterprise cloud workflows.

Meta

Meta's positioning is between Apple and Google. While it is facing AI challenges, its strategic positioning is solid. Individualized content fits into Meta's distribution channels, and generative ads enhance its advertising base. Generative AI is key to realizing returns on XR investments by creating metaverses for VR and UI for AR. The scarce resource is attention, and LLMs consume a lot of it. Chatbots could be a problem for Meta, as time spent using Meta AI is time not spent on monetized formats. However, Meta may monetize new surfaces better, as there is no expectation of objectivity like with search. Zuckerberg's AI recruiting efforts suggest that Meta's core business is threatened, requiring new leadership and product thinking.

Microsoft

Microsoft's position seemed strong. It has a cloud service that sells GPUs and is the exclusive cloud provider for OpenAI. Incorporating ChatGPT-like results into Bing risks the business model for market share. GPT is coming to Microsoft's productivity apps. The success of GitHub Copilot, an AI-coding tool, should be imitated. Adding new functionality fits Microsoft's subscription model.

Microsoft's infrastructure and distribution are advantages. Its pivot after missing mobile positioned it to capitalize on AI. It is the most Nvidia-centric of the hyperscalers. Azure is the exclusive non-OpenAI provider of OpenAI APIs. Microsoft's priority should be securing this Azure advantage and deepening its relationship with alternative model providers like xAI and Mistral. Not having access to leading models would be costly.

Amazon

Amazon's position has become more optimistic. AWS remains the largest cloud provider, and enterprises prefer AI close to their existing data. There is potential in voice-controlled devices with Alexa. Foundation model makers are critical to AI.

Foundation Model Makers

OpenAI is viewed as a consumer tech company. ChatGPT has won the consumer AI space. OpenAI is in conflict with entities that seek to own the customer relationship, like Microsoft and Apple. The question is when OpenAI will figure out an advertising model to supplement its subscription business. Anthropic has a strong position with developers and API revenue. Its reliance on Amazon and its Trainium chips could mean cost savings. Amazon is a committed partner for Anthropic. xAI's acquisition by X makes it a less attractive investment. Meta may be wasting money on AI, but the ultimate capabilities of LLM-based AI are debated. Zuckerberg needs to promise superintelligence to attract talent. If LLM-based AIs are more like the microprocessor, then Meta wouldn't need to invest in building their own. Waiting-and-seeing is risky. China may commoditize chips and AI, benefiting Big Tech.