Summary
Qualcomm Incorporated reported Q2 2023 earnings after the close on August 2, 2023.
Qualcomm forecast a very positive long-term demand outlook for on-device AI inference capabilities.
Qualcomm sees the potential for a major upgrade cycle for devices, but this potential will take time to materialize.
da-kuk/E+ via Getty Images
Qualcomm Incorporated $Qualcomm(QCOM)$ reported fiscal Q3 earnings after the close on Aug 2, 2023. During the earnings call, the company provided some upbeat new commentary about the impact of artificial intelligence ("AI") on its business. In this article, I discuss some significant takeaways about Qualcomm's commentary on AI.
The Outlook For On-Device AI Inference
Qualcomm's commentary about AI primarily focused on the company's very positive long-term outlook for on-device AI inference. Qualcomm CEO Cristiano Amon highlighted several dimensions along which on-device inference may be preferable to cloud-based inference, including "context, immediacy, privacy, security, application reliability and personalization capabilities." He also emphasized the importance of on-device inference for "everything that is real-time AI, always on AI."
Qualcomm foresees that demand for on-device inference will be high enough that "it could create an upgrade cycle [for handsets] in between 5G and 6G." Per Amon, "As multibillion parameter gen-AI models run pervasively and continuously on-device, we believe our NPUs' [Neural Processing Units] unparalleled AI processing performance and power efficiency will become a requirement."
Qualcomm has some impressive competitors, so the claims about unparalleled performance should probably be taken with a grain of salt as typical marketing speak. Still, if generative AI models come to be used "pervasively and continuously on-device," it is possible that AI hardware features in chips (whether Qualcomm's or otherwise) may become a requirement for consumers.
Qualcomm also sees AI features potentially lifting ASPs as the capabilities of its chips increase.
This commentary from Qualcomm aligns with a similar forecast about on-device inference from Intel. During Intel's recent earnings call, CEO Pat Gelsinger had likewise forecast that the demand for on-device inference capabilities is expected to drive a major upgrade cycle for client CPUs—citing latency, bandwidth, and cost structure as primary drivers for demand. The fact that Intel has provided similar commentary about on-device AI inference should lend more credence to the outlook being forecasted by Qualcomm.
Readers can see a more detailed analysis of Intel's commentary in my recent article "Intel Q2 2023 Earnings: Key Updates About The Impact Of AI."
This Story Will Take Time To Play Out
As we can see with Qualcomm's mostly flat guidance for Q4, as well as the general tenor of management's commentary, the potential effects of AI on Qualcomm's business will not materialize immediately. In fact, all things considered, it makes sense that demand for consumer AI chips should emerge much slower than demand for data center AI chips, and investors should not expect an Nvidia-like short-term explosion in revenue.
After all, there is a very clear and pressing use case driving the current insatiable demand for data center AI chips: training large models. Admittedly, this is not the only use case, but it should be the primary driver for demand at the moment. By contrast, consumer AI chips for devices like handsets cannot train large models, so there is less immediate need for them.
Rather, for on-device AI chips, the primary use case is inference once the aforementioned large models (and their lighter versions for consumer devices) are trained and fine-tuned and packaged into applications. Only then should we expect inference workloads to truly take off. It therefore makes sense that chips intended for inference will see demand growth later than chips for training.
In time, though, one can see the possibility of an ecosystem developing where more and better AI applications drive demand for on-device AI chips and vice versa. But this process cannot happen overnight, so investors will have to be patient for the AI revolution to significantly affect Qualcomm's business.
Does The Market Undervalue Qualcomm's AI Potential?
We are still in the very early stages of the AI revolution, so we are all still discovering how AI workloads are likely to shape the future of the chip industry. Inference workloads, in particular, have been a significant question mark. We do not yet have a very clear picture of demand for inference—it is even possible that inference will drive significantly more demand for AI chips than training. Nor do we yet know how inference workloads will be distributed between the cloud and edge.
It is probably still too early to say with certainty, but the commentary from Qualcomm is another credible data point suggesting that inference will be a very significant workload, and that it may drive very significant demand for devices with on-device AI features.
If Qualcomm's predictions turn out to be accurate, then the company's current valuation may be too low. Qualcomm trades at far lower multiples than most other firms considered to be likely beneficiaries of the AI revolution. As such, depending on how seriously we take this updated commentary, Qualcomm could be one of the most affordable AI plays on the market.
Conclusion
As we have seen with a number of firms this earnings season, Qualcomm's commentary about the impact of AI on its business seems to have turned significantly more positive turn since last quarter. Although the short-term impact of AI is likely to be limited for Qualcomm, the company may eventually turn out to be an under-appreciated and undervalued AI play.
For long-term investors, the short-term slow start should not matter much, and the focus should be on Qualcomm's execution over the next couple of years. For such investors, Qualcomm Incorporated's post-earnings slump may represent an attractive buying opportunity as Qualcomm moves closer to 52-week lows.
Source: Seeking Alpha
Comments