The largest technology companies all agree that AI matters. They disagree on almost everything else: whether to build or partner, whether to open their models, and where AI should run.

Microsoft: the partnership play

Microsoft’s AI strategy has been anchored in its partnership with OpenAI, which began with a $1 billion investment in 2019 and was extended in January 2023 with a multi-year, multibillion-dollar commitment. Azure supplies the compute; Microsoft distributes the models through Copilot features across Windows, Office and GitHub, and through Azure’s model catalog.

Google: research to products

Google invented much of the modern stack, including the Transformer architecture and the TPU chip. In April 2023 it merged Google Brain and DeepMind into Google DeepMind to speed up product work. Its Gemini models now power Search features, Workspace and Android, and Google is one of the few companies that controls every layer: chips, cloud, models and consumer distribution.

Meta: open weights at scale

Meta took the opposite path to most rivals by releasing the weights of its Llama models. Llama 3.1, released in July 2024, included a 405-billion-parameter model that anyone could download under Meta’s license. The bet: an open ecosystem builds tooling and talent around Meta’s models, while Meta’s own products — Facebook, Instagram, WhatsApp — provide distribution to billions of users.

Amazon: the infrastructure provider

Amazon focuses on being the place where AI runs. AWS offers many third-party and in-house models through Bedrock, builds its own Trainium and Inferentia chips, and invested a total of $8 billion in Anthropic, which named AWS its primary training partner.

Apple: on-device first

Apple announced Apple Intelligence in June 2024, emphasizing models that run on the device and a “Private Cloud Compute” system for heavier requests. It also integrated ChatGPT as an optional extension in Siri, showing that even vertically integrated companies partner where it makes sense.

What it means

  • Partnerships trade control for speed.
  • Open weights trade direct revenue for ecosystem influence.
  • Custom chips reduce dependence on NVIDIA and lower the cost of inference.
  • Distribution — operating systems, apps, clouds — may matter as much as model quality.

Comparing the strategies at a glance

Company Core AI strength Main distribution Model approach
Microsoft Enterprise software and cloud Windows, Microsoft 365, Azure, GitHub Partnership with OpenAI plus its own models
Google Research and search Search, Android, Workspace, Google Cloud Gemini models built in-house
Meta Social platforms and open models Facebook, Instagram, WhatsApp Open-weight Llama models and in-house research
Amazon Cloud infrastructure AWS, retail, Alexa Model marketplace, own models and partner investment
Apple Devices and privacy iPhone, Mac, iPad On-device models plus cloud and partner models

Where they compete and cooperate

These companies compete for developers, cloud customers, consumer attention and AI talent — but they also cooperate: one company’s models may run on another’s cloud, and device makers integrate assistants from partners. Competition authorities watch both the competition and the partnerships closely.

Questions for the next few years

  • Which consumer AI assistant becomes the default for most people?
  • Will open-weight models commoditize model capabilities?
  • How quickly do enterprise customers move from pilots to large-scale use?
  • Can custom chips reduce dependence on NVIDIA?
  • How will regulation in the EU, US and elsewhere shape product strategies?

See why big tech builds its own AI chips and AI data center capex explained.

Bottom line

Each company plays to its existing strengths, and these strategies keep evolving.

What to watch next

  • Capital spending on data centers by each company.
  • Assistant integration into operating systems and devices.
  • Open-weight model releases and licensing.
  • Regulatory scrutiny of partnerships and market power.

Key takeaways

  • Microsoft combines its OpenAI partnership with enterprise distribution.
  • Google builds models in-house and integrates them across search and products.
  • Meta releases open-weight models and uses AI across its social platforms.
  • Amazon focuses on cloud infrastructure and offering many models.
  • Apple emphasizes on-device AI and privacy, complemented by cloud and partners.

Frequently asked questions

Which big tech company is winning in AI?

Each leads in different areas — cloud, research, distribution, open models or devices. The race remains open.

Why does Meta release open models?

To build an ecosystem around its technology, attract developers and avoid dependence on competitors’ models.

Does Apple build its own AI models?

Yes, it develops on-device and server models, and also works with partners for some features.