Apple hopes its iPhone 16’s AI features will be enough to revive flagging sales

Apple announced yesterday that its new iPhone 16 will go on sale later this month. The phone is the first to be designed specifically to support the company’s “Apple Intelligence” features, including new AI photo editing tools, an improved version of Siri, generative AI writing aids, and the ability to summarize emails and notifications. Access to OpenAI’s GPT models through Siri is coming too. It’s all powered by Apple’s new operating system, iOS 18, and inside the phone by Apple’s custom A18 chip which has been designed to support AI workloads.

Many of these new AI features won’t roll out to customers until later in the year, or even next year. Overall though, the phone represents a big bet that demand for AI features will convince consumers to upgrade, awakening iPhone sales from a somnolent period that has been a drag on Apple’s revenues and its share price. Sales of the iPhone 16 are also poised to be a bellwether for the overall consumer generative AI market. And so far, the strength of that market is far from certain.

Consumer chatbot use is hot, but the money is on enterprises

Yes, OpenAI and Meta both announced seemingly impressive user numbers last week. OpenAI said 200 million people now use ChatGPT each week. Meta said its AI chatbot had 400 million active monthly users and 185 million weekly users. Those numbers are an indicator that there is a real consumer market for gen AI products.

But it’s unclear that an upgraded digital assistant and some new gen AI writing and photo editing tools will really be enough to convince people to shell out between $799 and $1,200 (depending on the version) for a new phone, especially when you can already access ChatGPT or Meta’s AI on your existing devices.

The real money from gen AI is being made in the enterprise, even as some chief information officers complain they aren’t seeing enough value from gen AI software. OpenAI revealed last week that it now has 1 million paying business customers for its service and that 92% of the Fortune 500 are using OpenAI’s models in some capacity. Meta has said its Llama large language models—which it makes available for free—have been downloaded some 350 million times. That figure is an indicator that developers are trying to build AI applications based on Llama. Moreover, it’s evidence that the generative AI boom is still far from petering out. (Meanwhile Anthropic has belatedly entered the enterprise AI space as my colleague Sage Lazarro pointed out in this newsletter on Thursday.)

AI reinforces the dominance of today’s tech titans

One thing that is apparent from this past week’s news is the extent to which the next big technology platform—AI—is likely to be dominated by the same set of giant tech companies that currently rule the roost: Google, Microsoft, Amazon, Meta, and Apple, alongside AI chip titan Nvidia. There was a brief moment, 18 months ago, when a few of these giants looked vulnerable. ChatGPT, some argued, was poised to be a Google-killer. Apple and Amazon both seemed hopelessly behind in their efforts to build large language models and integrate them into their product lineup.

But that moment was vanishingly brief. It is now becoming apparent that while the pecking order of these giants might be scrambled by AI, as a group their preeminence is unlikely to be usurped.

OpenAI is the closest of any upstart to breaking into this top tier, but even it is completely dependent on the largesse of the tech giants. Its success is largely due to the funding and data center infrastructure Microsoft has provided it. OpenAI seems to be trying to break its dependency on Microsoft—reports are that it is looking to raise billions in new funding in an investment round that could include Apple and Nvidia, as well as Microsoft and a gaggle of Silicon Valley venture capital firms. But the fact OpenAI needs more cash, despite having received a $10 billion commitment from Microsoft a little more than a year ago, is a testament to just how high its current burn rate is and how tricky its path to ever escaping the orbit of its Big Tech patrons will be.

This concentration of power should be a concern. As with previous tech platforms, there is a tendency towards monopoly or at least oligopoly. That has costs—in innovation and choice—even if these harms are not always as immediately apparent as higher prices are. In the U.S. and the EU, governments are only now seeking to address the concentration of power inherent in previous platform shifts—to the internet and mobile. Maybe we won’t have to wait as long for regulators to do something about the control these few companies exert over AI’s development. But I’m not holding my breath.

Here’s more AI news.

Jeremy Kahn
jeremy.kahn@fortune.com
@jeremyakahn

Before we get to the news. If you want to learn more about AI and its likely impacts on our companies, our jobs, our society, and even our own personal lives, please consider picking up a copy of my book, Mastering AI: A Survival Guide to Our Superpowered Future. It’s out now in the U.S. from Simon & Schuster and you can order a copy today here. In the U.K. and Commonwealth countries, you can buy the British edition from Bedford Square Publishers here.

AI IN THE NEWS

Countries sign Council of Europe Framework Convention on AI. The U.S., U.K., and the European Union, as well as Japan, Canada, Mexico, Australia, Israel, and others signed the convention at a meeting in Vilnius, Lithuania. It commits countries to take specific steps to ensure AI does not impinge on human and civil rights, that personal data is protected, and that AI is developed safely and responsibly. You can read more here.

German AI startup Aleph Alpha pivots to providing AI support. The company had been developing its own powerful large language models and raised $500 million in funding. But it will now focus on helping companies and governments deploy generative AI models from any vendor through an operating system product it calls PhariaAI, according to a Bloomberg story.

Elon Musk claims world’s largest GPU cluster built in record time. The billionaire said he had completed a data center, which he dubbed Colossus, with 100,000 of Nvidia’s H100 graphics processing units. The data center, which is designed to support the training of massive AI models by Musk’s xAI startup, was built in just four months, according to the billionaire, a claim that has raised eyebrows among rival AI companies and the data center industry, not just for its size, which would eclipse the clusters available to many other AI companies, but also for the construction timeline, which is much faster than the 12 months it typically takes to erect, equip, and wire together such a large cluster. Skeptics also questioned whether the local power grid could actually support so many GPUs in a single location. You can read more in this story from The Information.

EYE ON AI RESEARCH

Could crowd computing substitute for large GPU clusters? Some AI researchers and developers are hoping that it might be possible to harness the collective computing power of internet-connected machines across the globe to train large-scale AI models without the need for expensive, centralized clusters of graphics processing units. Now researchers at Nous Research have demonstrated that a model with 1.2 billion parameters can be effectively trained in this way. If the results hold for even larger AI models, the technique Nous pioneered could make it possible to train state-of-the-art AI models without the need for costly AI computing clusters. At the same time, such crowd-sourced computing efforts make it much harder for governments to regulate the development of powerful AI models. You can read the Nous research on GitHub here.

FORTUNE ON AI

Nvidia is a bargain now that AI is going beyond the hyperscalers, says Goldman Sachs —by Christiaan Hetzner

AI startup Glean aiming to build the ‘Google for Work,’ raises $260M at $4.4B valuation —by Sharon Goldman

Elon Musk’s xAI startup could help Tesla with FSD, Optimus, and Siri-like feature, report says —by Jason Ma

Would a gen AI-powered Alexa return real profits for Amazon? —by Jason Del Rey

AI CALENDAR

Sept. 10-11: The AI Conference, San Francisco

Sept. 10-12: AI Hardware and AI Edge Summit, San Jose, Calif.

Sept. 17-19: Dreamforce, San Francisco

Sept. 25-26: Meta Connect in Menlo Park, Calif. 

Oct. 22-23: TedAI, San Francisco

Oct. 28-30: Voice & AI, Arlington, Va.

Nov. 19-22: Microsoft Ignite, Chicago

Dec. 2-6: AWS re:Invent, Las Vegas

Dec. 8-12: Neural Information Processing Systems (Neurips) 2024 in Vancouver, British Columbia

Dec. 9-10: Fortune Brainstorm AI San Francisco (register here)

BRAIN FOOD

The race to build the $100 billion data center is on, and its starting line is in North Dakota. The largest AI companies are anticipating needing ever larger, more powerful data centers stuffed with graphics processing units or other AI chips. The Information reports this week that North Dakota’s Commerce Commissioner Josh Teigen said he and the state’s governor have been approached by two companies hoping to build supercomputing clusters in the state that each would cost as much as $125 billion and eventually draw 5 to 10 gigawatts of power. Five gigawatts is as much power, the publication points out, as the entirety of Microsoft’s Azure cloud currently consumes in a year. While the public official did not say which companies were involved, the obvious candidates are Microsoft, Google, Meta, Amazon, Nvidia, and perhaps Apple or Musk’s xAI. Increasingly, some communities are turning against these huge data centers, concerned they draw too much power and deplete precious groundwater resources, while not doing enough to contribute to either jobs or the tax base.