Nvidia, OpenAI and Their Rivals Expand AI Turf Wars
The Takeaway
• Meta could follow other open-source AI developers by selling Llama directly to customers
• It’s probably a matter of time before Google and OpenAI start developing humanoid robots themselves
• Don’t be surprised if Microsoft and Amazon develop wearable AI devices.
Nvidia, a chip giant, says it can develop a cloud services and software business to rival Amazon Web Services. OpenAI, an artificial intelligence developer, has been developing server chips and a browser to reduce its dependence on Nvidia and Google, respectively. And Elon Musk hired a data center team to build one of the biggest supercomputing server clusters for developing conversational AI.
It seems like every major competitor in the AI field is moving into the territory of its rivals or customers. They’re trying to either take a piece of the growing market for consumer and enterprise AI services, including server hardware and personal devices, or create a new industry, such as humanoid robots. In many other cases, especially when it comes to AI server chips, companies are trying to lessen their dependence on key suppliers to save money.
As the two charts below show, several large tech companies are competing neck and neck to develop an AI hardware and software “stack,” 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 squares 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.
For instance, OpenAI began exploring developing its own AI chips to lessen its reliance on Nvidia barely a year after launching its first major business, ChatGPT, in late 2022. It has continued that effort with the help of Broadcom, a longtime chip designer, and plans to launch sophisticated chips by 2026. Apple also recently embarked on development of its first AI server chip with Broadcom’s help as Apple barrels into consumer AI services for the iPhone.
Llama Cloud Service?
Meta Platforms, meanwhile, runs a major open-source AI model, Llama, but doesn’t have a way to distribute it directly to businesses. Meta could follow in the footsteps of other open-source AI developers by hosting the Llama software in its own data centers and selling it to businesses as a cloud service, as well as providing extra security and other features such buyers might want.
Meta already is going outside its traditional comfort zone by developing a web search engine so its AI chatbot can answer questions about current events without needing help from Google and Microsoft.
The frenzy in AI mirrors the expansion Google, Amazon, Apple and Microsoft have made over the past 15 years to compete with one another in selling cloud services, mobile devices and video—with varying degrees of success.
In AI, however, the attempted land grabs have happened faster, as leaders in the field don’t know which businesses or technologies will be most valuable in the long run.
OpenAI recently took control over the planning of a data center in Abilene, Texas, because it felt that Microsoft, which is supposed to be its exclusive data center provider, wasn’t moving fast enough to set it up. OpenAI also has been developing a web browser so it can have more control over the way people interact with ChatGPT. And through a separate company, OpenAI CEO Sam Altman has been developing an AI-powered personal device that could possibly use OpenAI software.
OpenAI technically doesn’t own or control the Abilene facility, but don’t be surprised if it gets more involved in setting up data centers and supercomputing clusters for AI down the line. Chief Financial Officer Sarah Friar has already talked with colleagues about the company’s ability to raise debt for such facilities.
Meanwhile, Nvidia, which was feeling some heat after customers such as OpenAI and Amazon ramped up development of their own custom AI chips, has been expanding a cloud and software business that could compete with Amazon Web Services. Nvidia has typically sold products through intermediaries, so its shift to direct sales may not be smooth.
AWS, for its part, is making a big push with its own AI chips, which it hopes will eat into Nvidia’s share of the market, though that’s far from guaranteed. And AWS recently launched its own large language models, similar to OpenAI’s, so it doesn’t have to entirely rely on Anthropic. AWS resells Anthropic models to cloud customers and uses them to power new Amazon AI products and features.
Then there’s Musk, who unexpectedly developed an advanced data center to aid his startup, xAI, as he felt that cloud providers such as Oracle wouldn’t move fast enough to set up such a facility. The bold and risky bet appears to be paying off and has set into motion a scramble among rivals including OpenAI to catch up. Some of them have even felt compelled to spy on the facility from the air.
Chip Flop
Not all the companies’ new moves have gone well. Microsoft last year finally launched an AI chip to potentially lessen its reliance on Nvidia, but the chip hasn’t performed as well as the company hoped. Microsoft isn’t selling it to cloud customers yet, though it is continuing to develop new versions.
Where companies have weaknesses, they have turned to partners. The most famous example in the AI bunch is Microsoft’s reliance on OpenAI models, though it is trying to develop comparable ones on its own and has been striking deals to use models from other providers such as Anthropic. Apple is also partly relying on OpenAI to power features for Apple Intelligence.
Acquisitions are another option. As Nvidia has increasingly developed a nascent cloud and software business to complement its chip largesse, for instance, it has made several small startup acquisitions to buttress those newer products.
Anthropic, OpenAI’s biggest startup rival, remains a potential acquisition target for companies such as Amazon or Apple that don’t have state-of-the-art LLMs that can compete with those of OpenAI.
Of all the companies in our charts, Google has established its technology or business bona fides in nearly all of the software and hardware categories in our chart, so it seems the least likely to pursue a major acquisition or partnership. And now it's looking to leverage major strengths in some areas, such as consumer apps like search, to point its billions of users to Gemini, a chatbot that competes with ChatGPT, The Information reported last week.
However, Google has been pushing to break apart the Microsoft-OpenAI cloud deal so that Google Cloud might host OpenAI models in its data centers and resell them to cloud customers. Right now, only Microsoft can do so, meaning Google and other non-Microsoft clouds are missing out on potential revenue.
In the area of wearable AI devices or robots powered by LLMs, a lot could happen in the coming year as Musk parades around his nascent Optimus robot prototypes and talks about them as a $1 trillion-per-year revenue opportunity. More rivals undoubtedly will get in the game.
The Robots Are Coming
For instance, Google on Friday announced it would provide AI software to humanoid robot startup Apptronik. Google previously owned and later shut down or sold numerous robotics firms, including Boston Dynamics. But now, some of the AI that powers services like ChatGPT can help robots understand objects around them. It’s probably a matter of time before Google and OpenAI—which earlier this year said it restarted its robotics software division—start developing humanoid robots themselves.
Meta sells smart glasses that it is pairing with its Llama AI so the wearer could verbally ask the AI about their surroundings or request other information, getting responses through speakers positioned next to their ears. Meta has also considered developing earbuds resembling Airpods, but with an outward-facing camera that can “see” and analyze the person’s surroundings.
Apple and Google are incorporating AI into mobile phones to perform functions similar to those of the glasses, and Apple could also make its mixed reality Vision Pro device a hub for AI assistants and apps. Google has also said it has developed prototype glasses for AI.
Microsoft and Amazon haven’t announced any new AI devices or robots based on LLMs, but the chart suggests we shouldn’t be surprised if they do. Microsoft is no stranger to wearable hardware, given its experience developing Hololens, and an Amazon Alexa team last year discussed developing a new AI-powered device.