At Qualcomm’s 6G Leadership Day held at its HQ in San Diego last month, the company talked in detail to industry analysts and the press about its latest views on 6G. While discussion of 6G in previous years has been fairly nebulous and lacking in specifics, this year we are starting to hear more concrete details about deployment and capabilities.

Qualcomm also used the 6G Leadership Day to highlight the importance of spectrum and of working with carriers and governments to ensure there’s enough spectrum to go around for 6G. Qualcomm, like many others in the industry, is taking a more tempered approach to 6G and talking about it in a way that is more easily absorbed by the whole ecosystem. At the very core of Qualcomm’s thesis is that 6G will be the AI-native generation of cellular.

Qualcomm’s AI-Native Approach to 6G

Everyone seems to agree that 6G will be AI-native, but specifically how that is implemented depends on who supplies the compute. For example, 5G Advanced is already carrying some uplink-heavy agentic traffic that’s been pitched for 6G. This really shouldn’t be a surprise, because most of the later releases of any cellular generation start to blur the lines between Gs and offer some of the capabilities of the next one, much like we saw in the transition from 3G to 4G. Qualcomm sees 6G as the formalized approach of AI-native cellular networks, and during the event again mentioned its longstanding “AI is the new UI” mantra. Qualcomm also talked about 6G dynamic QoS (quality of service), which allows the device to talk directly to the network to make changes based on shifting device needs and workloads. This is an advance beyond 5G QoS, which kept traffic types within a single QoS flow.

In this agent-driven future, Qualcomm envisions multi-device, agent-centric experiences across drones, vehicles, robots, watches, glasses, and phones as a single orchestrated system. Besides AI, Qualcomm sees three main pillars to 6G: connectivity, wide-area sensing, and high-performance, energy-efficient compute. Qualcomm also still has a strong viewpoint that XR will grow demand for 6G, even though I think we can all agree that the same thing didn’t happen in 5G as some inside the industry thought it would. During its presentation, Qualcomm did show some compelling XR demos that paired with local AI compute, in this case to offload some of the compute from a pair of smart glasses while enhancing their capabilities.

6G’s Need for More Spectrum

During the 6G Leadership Day, Qualcomm repeatedly emphasized 6G’s ability to support wider bandwidth channels, up to 400 MHz. The company talked about 5G being a mostly 100 MHz channel technology (assuming spectrum availability), but added that 6G’s increased spectrum would give it enhanced sensing capabilities.

Qualcomm is focused on the FR3 bands from 6 GHz up to 8.4 GHz, which happens to be where Wi-Fi 6E and Wi-Fi 7 reside today (from 5.925 GHz to 7.125 GHz). So, naturally, a lot of Qualcomm’s testing for 6G and its new Giga-MIMO prototype antennas are in the 7 GHz to 8 GHz range, as well as at 13 GHz. The purpose of Qualcomm’s Giga-MIMO antennas (an upgrade from Massive MIMO) is to enable 7 GHz coverage to behave similarly to 3.5 GHz without any densification. This also means that the spectrum needs to be available to multiple carriers to allow for anything close to 400 MHz of coverage — even though I believe it’s unlikely any U.S. operator will get 400 MHz of contiguous spectrum. Even if operators don’t acquire 400 MHz in new spectrum, 6G still promises to deliver an up to 50% gain in spectral efficiency from a baseband-only upgrade.

6G’s Timeline Is Accelerating

Just like with any new cellular generation, everyone is pushing for things to happen sooner, and while 6G was originally anticipated for 2030, we already have some ideas of how much sooner it might become a reality. The 3GPP standards body is scheduled to finalize the Release 21 specification in March 2029, but we should see early foundational work done by March 2027. Some of the earliest assumptions for these releases were locked in as early as 2025, and it’s worth remembering that the 2028 L.A. Olympics have already been touted as among the earliest 6G use cases and demos.

Qualcomm’s own commitment is to have a commercial 6G modem available in the second half of 2029, setting up 2030 as the year of true 6G. It seems definite that we will get some pre-6G standard demos and trials before then, but I believe it’s unlikely we will see nationwide 6G networks in 2029, considering the commitment by the 3GPP and its members to go the standalone route for 6G. Recall that 5G was implemented with a non-standalone approach to ease carriers’ transition, with the result that most users initially experienced 5G deployments as equivalent to 4G — or worse.

The Changes to 6G’s Architecture

With features like Dynamic QoS and the network’s ability to understand device workloads and anticipate network conditions, there are many opportunities to improve the user experience without needing to push higher speeds. That said, the 6G demos we saw at the Qualcomm event still showed it performing exceptionally well with 16 Gbps downlink and 4 Gbps uplink. These context-aware communications should allow for reduced latency, improved throughput, and bandwidth savings, according to Qualcomm’s early prototypes.

But this is far from being the main area where things will change in 6G, especially because integrated sensing and communication (ISAC) gives operators an opportunity to negotiate with the government for spectrum. Qualcomm gave an incredible live demo of ISAC using its Giga-MIMO antennas, which showed not only drones flying by but also cars and their speeds, and could serve as an additional sensor for security and safety applications. Qualcomm demonstrated this capability with one of its own AI accelerator add-in cards, which I took as a direct jab at NVIDIA.

AI-RAN was also a very hot topic, with Qualcomm talking about the company’s capabilities across CPU, NPU, and other accelerators that will be necessary at the RAN (radio access network) level. Qualcomm sees the future RAN as an agentic problem that will require CPUs, NPUs, and what it calls 6G accelerators working in conjunction to manage the network. Qualcomm, like NVIDIA, sees the 6G network utilizing AI-RAN for both radios and compute. How that looks in practice, I believe, will be very different. (For more context, see my recent analysis of Nokia’s AI-RAN partnership with NVIDIA.) It’s also important to understand that AI will not only be running on top of 6G networks; it will also be pivotal in running the networks themselves, which is why the AI-native story is being emphasized.

Qualcomm’s 6G Story Is a Counter to NVIDIA

Qualcomm did a good job of telling a cross-company story, complete with concrete dates and a strong focus on improved economics. The demos felt more real than ever before and validated many of the assertions made during the presentations. There’s still a very long road ahead for Qualcomm, OEMs, and cellular operators to make 6G a reality. A lot of the implementation may hinge on how much spectrum becomes available through government cooperation, since we’re already in quite a spectrum-starved environment for 5G.

The AI-native vision landed with me in a meaningful way and showed a more comprehensive view than NVIDIA’s, with much more detail. Ultimately, this was Qualcomm’s opportunity to counter the 6G messaging of the NVIDIA-Nokia partnership, and I think Qualcomm did a good job, even though we didn’t hear or see much from network vendors. I suspect those developments might still be saved for MWC 2027 or 2028, and until then, we will have to make do with Qualcomm’s and NVIDIA’s prototypes and demos.