In my last post, I opined on how the size of today’s smartphones have increased year over year to accommodate a bigger battery, and how that size has now maxed out, especially if you want your phone to remain pocket-sized — and your pants are (hopefully) not getting any bigger. And with the advent of AI and other heavy workload applications that Gen Z in particular is using more and more, smartphones desperately need a breakthrough in battery innovation.
When I joined Enovix one year ago, my goal was to solve this problem head on. Historically energy density in Li-ion batteries has increased less than 4% a year, a meager increase over three decades especially when you compare the advancements of other critical components in our phones — memory, CPU clock speeds, storage, camera resolution and cellular network speed. The battery has clearly held back our mobile experiences.
There is also constant pressure to drive more functionality into our smartphones, like 4K and 8K video upload/download, multi-player gaming and enhanced camera capabilities. All the while, there is an ever-increasing desire for high-demand applications, of which AI is the 800-pound gorilla.
To drive this point home, we did a few internal tests to demonstrate how AI is taxing devices even further. One test showed that we need 50% more battery capacity to support a popular AI program called Stable Diffusion. But we wanted to dig even deeper into this problem, so we commissioned a third party, Tirias Research, to conduct external tests. The results are eye opening.
What’s Driving the Need for a Better Battery
Using that latest smartphone models from leading OEMs, Tirias confirmed what we believe is driving the use case for a better battery: 1) Physical limitations of phone form factor 2) High compute use cases 3) Moore’s law unable to keep up with compute demands 4) Generational usage shift of AI driving demand.
Smartphones need more software and hardware optimization to support GenAI. These applications want access to the GPU, but it will tax the system and will be throttled. Power consumption is going up, but we’re using a tiny fraction of the model size expectations for GenAI.
The best of today’s AI-forward applications are hybrid on-device versus off-device computing environments. A good example is Zoom running its transcription and translation option. But many AI applications are being run locally, like TikTok, with content reaction, filters, editing and recommendations all being managed on-device. This trend will only continue with a massive generational shift when it comes to content creation and device usage. Gen Z and Millennials lead the charge on social media content creation, and they create and consume all of their content on their phones (forget the TV!)
Want Mobile Gaming? Find Your Power Cord!
In late ‘24, more GenAI video-based applications will arrive that take up 3–4X the processing power — a huge computational load. More demanding use cases (gaming, video enhancement) will push power consumption beyond feasibility.
The new generation of AI-based games are running the battery down at a rapid clip, with interactive video games draining the battery to zero in less than four hours. Moreover, not only do these games require quick video upload and download, but these apps often run in the background taking up a decent amount of system power when not in use, something that is not often considered.
User demand, as well as increased functionality, are pushing the need for more CPU and GPU performance, core count, frequency, and hardware accelerators. SoCs have added heterogeneous computing like NPUs and ISPs (image sensor processor), but of course, with Moore’s Law, demand will never be able to keep up with the limitations of physics.
You can see how the list is piling up.
But the kicker is that generative AI is forecast to increase more than 150-times by 2028. Let me repeat this in all caps: GENERATIVE AI IS FORECAST TO INCREASE BY MORE THAN 150X!
Given all of this, how will batteries be able to keep up? Simply put — battery technology needs to fundamentally change. Enovix battery architecture replaces graphite with silicon in the anode, which enables an increase in battery capacity up to 30 percent compared to batteries in today’s high-end smartphones. And we are on the path to exponentially increase performance into the future.
The industry must make improvements across every component of the phone, but perhaps most importantly, the battery must evolve. This is why I am so excited about 2024. We are planning many technological innovations and manufacturing milestones. Stay tuned as Enovix continues to revolutionize energy storage and offer solutions to these problems so that AI will enhance — and not kill — your phone!
Find the full read-out of the Tirias Research here, “Battery Technology Trailing Smartphone Innovation.”
