The News: Apple is reportedly exploring integrating Google Gemini and Meta AI into Apple Intelligence, a suite of generative AI features unveiled earlier this month at WWDC. The splashy announcement featured a ChatGPT integration, though Craig Federighi, Apple’s SVP of Software Engineering, said during his presentation that the company will also consider other AIs. Read the full story here.
Has Apple Missed the On-Device AI Market Opportunity in Mobile and PC?
Analyst Take: I keep being asked about Apple Intelligence and recent announcements (and rumors) about Apple’s partnerships with AI service providers, ranging from OpenAI to Meta and Google, so let me share some of my initial thoughts below.
Overview of Apple’s AI Strategy in the Consumer Electronics Segment
If you have followed our coverage of the intersection of the semiconductor, devices, and AI market segments, you will have no doubt noted that we have spent a good amount of time drilling down into the expansion of Cloud-based AI solutions to the Edge, and onto devices in the past two years.
Yes, we were already covering on-device AI innovation before generative AI blew up. But now that AI has become the most critical driver of technology spend, reframing the resulting AI ecosystem as Cloud-to-Edge-to-Device makes sense. What also makes sense is to think of the AI ecosystem as this multi-layered hybridized model, with each layer offering its own benefits and opportunities for semiconductor and component companies, device OEMs, and service providers.
Just as NVIDIA jumped on the Cloud and data center AI opportunity early, thanks to the role that its GPUs could play in accelerating AI workloads, Qualcomm leveraged the early advantage it also had in integrating AI workloads into its Snapdragon SOCs, and became the de-facto driver of on-device AI in the Mobile (Android), XR, Hearables and Wearables, Automotive, and PC spaces. Fast-forward to today, and one of the most important AI trends (and market opportunities) is in bringing on-device AI capabilities to every device imaginable, from phones and PCs to cars and wearables.
Qualcomm, as I have pointed out, is the company to watch here, as its Snapdragon platform family is currently, in my view, the most capable SOC ecosystem when it comes to delivering the best AI performance per watt in every IOT and device segment. Again: Mobile, XR, Hearables, Wearables, IOT, Automotive, and now PCs.
To take full advantage of the emerging hybridized AI ecosystem’s system and cost efficiency, UX and security advantages, enterprise technology customers, and consumers alike need a mix of AI-capable devices, AI-capable edge solutions, and AI-capable Cloud solutions. The Cloud services part is, because it benefited from a head start, the lowest hanging fruit in this model. AI-capable devices are, at present, the bleeding edge piece of that AI technology trinity, which is why virtually every semiconductor company working in the Mobile, PC, IOT, and Automotive segment has been focusing so many resources on developing powerful (and preferably energy-efficient) on-device AI capabilities.
Most recently, Microsoft and every major PC OEM (Acer, Asus, Dell, HP, Lenovo, Microsoft (Surface), and Samsung) partnered with Qualcomm to launch the first wave of truly AI-capable PCs, branded as Copilot+ PCs. These PCs are currently all built on Qualcomm’s Windows-on-Arm Snapdragon X architecture, which can compete directly against Apple’s MacBook PCs in ways that X86 Windows PCs can’t. The short of it is that Windows-on-Arm PCs deliver significantly better performance per watt and thermal characteristics than their X86 counterparts, even before the introduction of NPUs. But with the addition of NPUs (neural processing units) capable of 40+ TOPS of performance, not only can those Copilot+ PCs perform a variety of AI-intensive workloads directly on the device, they can also deliver additional performance per watt and thermal efficiencies that can, depending on the OEM implementation and the use case, outperform MacBook devices.
In other words, on-device AI is reshuffling the deck in the Mobile and PC segments, and semiconductor companies with the most advanced solutions AND the broadest, deepest bench of industry partnerships find themselves at a distinct advantage this early into what is shaping up to be a significant refresh cycle in several device segments, particularly PCs.
Note that MediaTek is also shipping AI-capable SOCs in Mobile, and that both Intel and AMD are also on the way to enter the Copilot+ PC ecosystems in the coming months, albeit with X86 chipsets.
What, you ask, does any of this have to do with Apple? Well, Apple appears to have failed to grasp the importance of on-device AI to its product and UX roadmap early enough to be in the race with its Mobile and PC competitors. And while Apple is no doubt working as fast as it can to catch up, its public-facing AI strategy right now seems to hinge almost exclusively on establishing strategic partnerships with Tier-1 providers of AI solutions – in other words, Apple’s AI strategy is mostly Cloud-based.
The State of the On-device AI Competitive Landscape
What this suggests is that Apple, at least for the time being, may be losing the AI arms race in the consumer electronics segment: In Mobile, Android chipmakers like Qualcomm and MediaTek are several generations in with on-device AI SOCs – Qualcomm’s Snapdragon 8 Gen 3 leading the market. In PC, Microsoft and every major PC OEM (Acer, Asus, Dell, HP, Lenovo, Microsoft and Samsung) are out with Snapdragon X-powered “Copilot+ PCs,” which sport 40+ TOPS (trillions of operations per second) on the NPU (neural processing unit), with Intel and AMD coming in hot behind Qualcomm in the segment. Apple doesn’t appear to have an answer to this on-device AI capability in either segment just yet.
Apple’s recent partnership announcement with OpenAI and subsequent rumors about partnerships with Meta and Google’s AI solutions speak to this on-device AI gap in the iPhone maker’s ability to compete against the Android (mobile) and Windows 11 (PC) ecosystems with regard to on-device AI capabilities.
In my view, whether Apple ultimately partners with every major AI services provider is irrelevant. Apple must also have on-device capabilities equivalent to or better than Snapdragon in Mobile and PC to remain competitive in the flagship and premium tiers.
The real winners in these AI services partnerships aren’t Apple. They are AI services companies. OpenAI, for instance, will benefit from Apple’s enviable install base (a major, frictionless sales funnel). By being integrated into the Apple ecosystem, I also presume that OpenAI and ChatGPT will avoid the dreaded “App Store tax,” at least for now. Other winners are the Cloud companies, in this instance Microsoft Azure.
Apple is basically just an access point for Cloud AI services, in this model. It’s a stopgap. Aside from the Apple Intelligence ecosystem itself, Apple still appears to be outsourcing much of its AI capabilities to Cloud partners rather than offering its own or integrating as many language or mixed models as possible into its own platforms for now.
What Is the Difference and Why Does It Matter?
On-device AI delivers an additional layer of performance and UX improvements that access to cloud-based AI services doesn’t.
Cloud-based AI services require a Wi-Fi or broadband connection, which means that workflows with AI-assisted features might only work when a user is connected to the cloud. Devices with AI capabilities built-in, on the other hand, can handle those workloads regardless of connectivity, which eliminates workflow interruptions.
Cloud-based AI services also come with two additional disadvantages compared to on-device AI options: Latency and Security vulnerabilities. Latency comes from having to send queries to the cloud for processing, and for the responses to make their way back to the user’s device. In instances where a conversational pace is important to the UX (like translations or talking with an AI assistant), that’s a problem. The security disadvantage is that every time datasets have to move from a device to the cloud and back, they can be intercepted. Shifting as much of these workloads as possible to the device solves both of these problems, and here, Apple is being left behind, at least for now.
Even Elon Musk, with whom I tend to disagree more often than not these days, astutely pointed out that Apple might open its users up to security challenges by outsourcing so many of its AI “features” to partners and cloud services rather than integrating them into Apple’s own stack. He isn’t wrong. Apple partnering with OpenAI/Microsoft, or Meta, or Google for AI processing means that Apple may not have complete control over its users’ data. This doesn’t mean that the data isn’t safe. It just means that Apple isn’t as much in control of its security as it usually is, and for many Apple users, that may be an issue.
Apple may also be forced to delay many of its AI “features” in the EU due to DMA (Digital Markets Act) regulations, further highlighting the holes in the company’s AI strategy… making it seem increasingly cobbled together.
To recap, Apple not being able to offer on-device AI workload capabilities means more latency, lackluster UX, and security vulnerabilities compared to premium Android phones and Copilot+ PCs. This is on trend with Apple’s resistance to new technology adoption outside of its own innovation bubble. (You may recall that Apple also initially balked at 5G and resisted the transition to USB-C until the EC forced its hand.) Apple innovates well but Apple also tends to innovate late. Many of the new features announced at WWDC this year have been pretty standard in Android phones for years. It seems that on-device AI may have gotten caught into similar filters, perhaps to the detriment of Apple’s competitive position in Mobile and PCs as on-device AI continues to accelerate its takeover of popular market segments.
Apple is Apple though. It has the talent, financial resources, and industry partnerships it needs to catch up, so I have no doubt that even as early as next year, we may see the company begin to close the gap with regard to on-device AI. There is no reason why Apple can’t do this if it sets its mind to it, and it seems rather obvious that Apple decision-makers understand the need to make this a priority.
For now, however, Apple is trying to make the best of a bad competitive situation by partnering with the biggest names in AI to have some kind of consumer-facing AI story.
Apple’s Tactical Calculus
I suspect that Apple’s tactical calculus takes into account the fact that the vast majority of its customers don’t know the difference between on-device AI and cloud-based AI, and will be more impressed with partnership announcements with big AI names such as Open AI, Google, and Meta than they would be with spec sheets showing NPU TOPS. Apple’s marketing engine is second to none, and the company should be successful in spinning this as a win instead of an embarrassment.
Market perception: The initial announcements during WWDC (Apple Intelligence and OpenAI) gave Apple’s stock a pretty significant boost, but half of those initial gains have already dissipated in the past week. Something to keep an eye on, in terms of Apple’s AI story perhaps not having as much traction as the public’s growing understanding of on-device AI’s advantages. DMA roadblocks in EU countries may also act as a brake on Apple’s AI narrative, at least for now.
Copilot+ PCs may be hotter than MacBook with major retailers: We are monitoring retail pre-orders for Copilot+ PCs, at least the models powered by Qualcomm’s Snapdragon X series chips, and early signals point to much stronger per-orders than we’ve seen for any MacBooks in years. It’s still early but it is an early signal that Arm-based Windows PCs with powerful AI chips are a threat to Apple’s momentum in retail and perhaps in the enterprise.
Mobile remains a question mark for now: On the phone side, we may have to revisit in the fall, as Qualcomm and MediaTek unveil their next generation flagship AI mobile platforms, but we may also see a softening of iPhone demand if the on-device AI story sticks with consumers, especially in Europe and Asia, where Apple may not be on as firm a footing as it is in the US.
To be continued… Follow this space for more analysis of the intersection of semiconductors, devices, and AI.
Disclosure: The Futurum Group is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of The Futurum Group as a whole.
Other Insights from The Futurum Group:
The PC Segment’s AI Inflection Point: Your 2024-2025 Copilot+ PC Cheat Sheet
The Copilot+ PC Disruption Is Here: What Happens Now?
Talking Top Tech Trends 2024, Apple, Salesforce, NVIDIA & AMD, Adobe, CES 2024
Author Information
Research Director Olivier Blanchard covers edge semiconductors and intelligent AI-capable devices for Futurum. In addition to having co-authored several books about digital transformation and AI with Futurum Group CEO Daniel Newman, Blanchard brings considerable experience demystifying new and emerging technologies, advising clients on how best to future-proof their organizations, and helping maximize the positive impacts of technology disruption while mitigating their potentially negative effects. Follow his extended analysis on X and LinkedIn.