OpenAI, Meta and Their AI Rivals Ramp Up Turf Wars and Partnerships, in Three Charts
The Takeaway
- Meta increased its lead in AI-powered devices but fell behind in AI models.
- xAI made gains in LLM quality, consumer AI apps, and AI training clusters.
- All these AI rivals are more interconnected than ever, due to a string of new alliances.
The AI Stack
Competition is heating up among nine top makers of artificial intelligence and related hardware as they expand their product lines. Highlighted cells show new entries since last year.
You can call 2025 the year of the robot, at least for the major AI companies. While companies such as Anthropic, Meta and xAI filled out their AI hardware and software lineups, nearly everyone started to work on technology for humanoid robots.
The shift to robots says as much about the current competitive landscape in AI as it does about the dream of humanoids running around every factory and household. Nearly every major technology firm offers or is developing a full slate of AI hardware and software.
For a better picture of how AI companies have evolved over the past year, take a look at the second chart below. It shows how nearly all of them improved their position in developing a fuller AI “stack” to capture more revenue through their products or potentially saving money in the long run by gaining more control of their AI training servers, as was the case with OpenAI and Anthropic.
Yet as our third chart shows, these gambits for more independence have only deepened their entanglements. As companies reduced ties with one partner, they fell into the arms of another, creating an increasingly tangled web of alliances that have made industry more interdependent than before.
These nine companies are pushing to get bigger pieces of the fast-growing markets for consumer and enterprise AI services, including server chips and personal devices. They also want to be there at the creation of new industries, such as humanoids. Another trend underlying the data is a push to save money by lessening dependence on key suppliers like Nvidia or on cloud providers such as Microsoft.
We did a similar chart package this time in 2024. Some of our predictions from then came true:
- Meta released an application programming interface to sell its Llama models directly to customers.
- Companies including Google, Amazon and OpenAI took steps to develop humanoid robot software or hardware. These are all at a nascent stage, as the challenges are enormous. (We expanded the humanoid robot category in this year’s chart to also include software for the machines, not just robot hardware, as Google, Amazon and Nvidia appear to be focused on software for now.)
- And Amazon made big moves to develop wearable AI devices in the form of augmented reality glasses.
Additionally:
- xAI says it’s developing enterprise AI applications dubbed Macrohard, the semantic inversion of Microsoft.
- OpenAI and Anthropic both started ambitious plans to develop or control the server clusters for developing their technology.
- OpenAI also stepped up to develop wearable AI devices, spending $6.5 billion in stock to acquire a design group overseen by former Apple design chief Jony Ive. While OpenAI’s devices may not arrive in 2027, we’re sure to hear lots about them in 2026. (OpenAI has worked on smart glasses, a smart speaker, a wearable pin and digital voice recorder.)
The relatively few gaps left in the first chart—namely, cloud-server rentals—are probably not going to be filled anytime soon.
Several large tech companies are competing neck and neck to develop advanced AI hardware and software, while others are trying to catch up.
In the below chart, the darker the square, the more advanced the company’s business or technology is in that field. Blank or lightly colored squares show the gaps these companies have in developing core components for AI, which they might fill through new internal efforts or an acquisition.
The chart shows how most companies made progress in one or more areas of the AI stack.
- For as much as we and others write about the intense competition among companies on this list, they are as interconnected as ever.
- Google was already the most well-rounded of the bunch when we published our chart in 2024, and that’s still true in the 2025 version. During the past year, Google’s prowess in AI server chips jumped. Anthropic put in a $20 billion order for Google tensor processing chips and Meta Platforms is seeking to sign a deal to use Google TPUs as well. Google has taken steps to sell TPUs to other cloud providers for the first time. It’s also become clear that these chips help Google save a lot of money in powering its AI, according to employees of rivals such as OpenAI.
- The only other notable progress in server chips came from Microsoft, which continues to develop its Maia chips, though it remains behind most rivals.
- In the 2024 chart, we had already considered Google’s large language models to be state-of-the-art. The company proved it again in 2025 with Gemini 3, which like Google’s other models was trained on TPUs.
- Meta made gains in developing AI-powered devices with its Meta glasses, and it’s far ahead of rivals such as Apple, albeit in a tiny market.
- xAI made gains in both LLM quality and with its Nvidia-powered training clusters, though its LLMs still lag behind Google, OpenAI and Anthropic by most real-world measures.
- We also showed gains for xAI in consumer AI apps. XAI’s Grok powers key features for the X app and has improved the experience for many users—especially in deciphering posts on X that would otherwise make no sense to half the people reading them. Grok also appears to have made inroads with people who want to use a chatbot for adult content.
- Elon Musk’s other company, Tesla, appears to be in the lead in humanoids among the biggest AI firms on our list, though the “hands problem” and other issues persist with its Optimus machines.
- Anthropic’s products have been booming, according to its unreleased financials, and our chart reflects its progress with both businesses and individual customers.
The chart shows two setbacks:
- Meta has lagged in developing its own state-of-the-art models, as Llama 4 failed to deliver big performance gains and compete against AI model incumbents. The company spent much of the year paying up to acquire talent to turn things around, so we expect it to improve in this regard in 2026.
- Nvidia stepped back from competing head-on with Amazon Web Services, Google Cloud and Microsoft Azure through a recent reorganization, though it hasn’t completely left the market.
For as much as we and others write about the intense competition among companies on this list, they are as interconnected as ever, as this chart shows.
OpenAI’s focus in 2025 was to expand its cloud relationships far beyond Microsoft, which has been its primary provider. For instance, Amazon struck a deal to provide $38 billion worth of servers, and possibly lots of cash, to OpenAI, while OpenAI plans to work with Amazon on commerce initiatives. (OpenAI announced even bigger cloud deals with Microsoft Azure and Oracle Cloud.)
Microsoft, meanwhile, struck a major deal to rent out Nvidia servers to Anthropic. Meanwhile, it is also buying Anthropic models for its products, such as Office 365 Copilot, even though it can get OpenAI models for free through its multifaceted partnership with the ChatGPT maker.
As the chart shows, Google is now a tech provider to at least five companies it also competes with. In 2025, Google struck a deal to power Siri queries for its longtime business partner Apple. (That effectively replaces the role currently held by OpenAI, but it appears to have had little effect on OpenAI’s business.) Google had previously rented out TPUs and other cloud servers to Apple developers, which is why we didn’t highlight that box in the chart as a new entry.
Google is also working on a deal to provide TPUs to Meta, another longtime business partner. And Google in 2025 started providing Nvidia-powered servers to OpenAI in a big way, leaving open the possibility it will someday sell or rent out TPUs to OpenAI too.