Can Qualcomm become the NVIDIA of the AI PC race?
/In the world of AI, and the computing infrastructure that powers AI, there is only one king: Nvidia. The company has skyrocketed to one of the largest in the world on the back of it success leading the transition to an AI computing ecosystem. It has a dominant hold on the chips that power the data centers and servers that enable companies like Microsoft, Amazon, and Google to deliver the AI services that gather the headlines.
But that position of strength is limited to servers and data centers today. There is an emerging battle for the AI market at the edge, defined as a combination of laptops, smartphones, and individual consumer and commercial devices. It is on these devices where this AI content will be consumed and will likely see a growing need for processing to offload servers and lower the delay from request to action.
This segment is absolutely still in play, with combatants ranging from Intel to AMD to Qualcomm, and yes even Nvidia with its gaming and mobile graphics chips. Everyone is fighting to control what most believe is the next growth frontier for AI and for chips to power it. Who will become the NVIDIA of edge AI?
Qualcomm is in a unique position of opportunity with AI at the edge, thanks to a combination of its communication background, strong investment in new computing architectures, and an initiative to create a software springboard to rival what Nvidia did with CUDA all those years ago.
Though Qualcomm and CEO Cristiano Amon consistently refer to the company’s transition from a communications to a computing company, its heritage and leadership in communications is an area often undervalued. Even if more AI computing happens on local devices, that will drive the need for more data and bandwidth consumption. So having the best wireless technology is a critical factor. At the Mobile World Congress show in Barcelona last week, Qualcomm announced its newest cellular modem called the Snapdragon X80 5G, as well as a new FastConnect 7900 chip that combines the latest Wi-Fi, Bluetooth, and wideband technologies. These will get integrated into high performance smartphones throughout the year as well as top end laptops.
Qualcomm’s lead in modems and wireless technology is an advantage that is often overlooked. Despite the best attempts of both Intel and Apple, companies that invested billions of dollars in modem technologies to stop paying Qualcomm for its offerings, failed at building competitive parts. Intel has abandoned the initiative all together and Apple’s solutions have been underperforming so much that it was recently announced it had signed a multi-year deal with Qualcomm to continue using Snapdragon modems in iPhones.
If Qualcomm wants to become in the consumer and device AI space what Nvidia is for the data center, it will take a significant effort in the form of software. Nvidia’s CUDA is a combination of software development tools, drivers, pre-configured and updated models that make writing software for its GPUs as simple as possible. Software that is developed using CUDA is more likely to be deployed at scale on those same architectures in the cloud.
Another move that Qualcomm recently made is the introduction of its AI Hub, a toolset of its own for deploying and optimizing AI for its processors. This takes Qualcomm beyond just enabling on-stage demos and promotional content for social media. It’s a developer engagement platform that provides simple to integrate tools and instructions to make sure the most important AI models run on Snapdragon hardware and run in an optimized, maximum performance state.
Today the Qualcomm AI Hub offers over 75 unique AI models that are ready for software teams to integrate and it supports the Snapdragon chips for smartphones, laptops, and even automotive. The company hopes that by making integration and support of its chips as easy as possible, it can entice developers to target its hardware with emerging applications that will be a part of the AI revolution at the edge and create the same inherent advantage that NVIDIA has with CUDA.
The third piece to this is about the hardware itself. In the smartphone market, Snapdragon is the clear leader in market share and performance, powering most of the highest end devices on like the Samsung Galaxy S24 Ultra. This leadership position means that Android application developers will really need to target Qualcomm chips for any AI compute. The company has also teased that its latest CPU core architecture, called Oryon and stemming from its Nuvia acquisition a few years back, will be coming to its smart phone chips soon as well, keeping Qualcomm in the driver’s seat in this market.
For the upcoming AI PC boon that most are projecting to happen in the second half of this year and into 2025, even though Qualcomm has minimal market share today, the Snapdragon X Elite platform announced in October of last year offers more than 4 times the AI performance of the currently shipping chips from Intel or AMD. The first laptops featuring this chip won’t be available until June, but all indications are that Qualcomm has some incredibly influential design wins and system partners that will raise eyebrows and drive some impressive volume in the market. Intel and AMD, and even NVIDIA, won’t go down without a fight of course, and Qualcomm has a steep hill to climb in the PC space.
It’s clear to me that there is a battle looming in 2024 for the mind share and wallets of consumers looking to bring AI to their phones, laptops, and almost everything else at the edge. Every consumer-facing technology company is trying to surf this next wave of AI, including Intel, Nvidia, AMD, Arm, Qualcomm and even smaller startups like MemryX or Rabbit. Qualcomm has a solid foothold to make the case for itself here: strength in the smartphone market, a growing portfolio of products for PCs, and a new software initiative that will drive developer adoption.