Qualcomm is facing dual pressures from a slowdown in its smartphone business and the loss of Apple’s in-house baseband chips, resulting in a 13% year-over-year decline in mobile revenue. However, automotive revenue surged 38% year-over-year to a quarterly record, and IoT revenue grew 9%. The company plans to begin initial shipments of customized data center chips to major cloud providers this year, and rumors suggest Qualcomm is collaborating with OpenAI to develop AI-powered smartphone processors. Qualcomm is betting on edge AI, edge inference, and automotive computing, aiming to transition from a leading mobile chipmaker to a comprehensive computing platform provider. The key challenges now are whether mobile profit margins can be stabilized, whether automotive and IoT businesses can continue scaling, and whether data center inference capabilities can be replicated.

Article author and source: Semi-Industry Spectrum

To participate in the next round of computing platform development, Qualcomm needs to prove itself again.

Redefine

Series B

Smart Security Solutions Provider

On April 27, Ming-Chi Kuo, an analyst at Tien Feng International, stated that OpenAI is collaborating with Qualcomm and MediaTek to develop an AI-centric smartphone processor, with potential mass production possibly occurring in 2028. The report also noted that the involved companies have not immediately responded to requests for comment, and OpenAI’s explored hardware form factor may not be a traditional smartphone, but rather what Sam Altman has referred to as the “third core device.”

Two days later, Qualcomm released its second-quarter fiscal year 2026 earnings. The results were not strong: revenue declined year-over-year, the mobile business faced pressure, and guidance for the next quarter was weighed down by storage supply constraints and weak customer demand. However, during the earnings call, CEO Cristiano Amon mentioned that the company expects to begin initial shipments of data center chips to a major cloud provider this year.

Placing these two moments together reveals the real context of Qualcomm’s “midlife crisis”: it remains deeply tied to the smartphone cycle, facing competition from Apple’s in-house baseband chips and the Android market; yet, advancements in edge AI, personal AI hardware, automotive computing, edge inference, and custom data center chips are causing the market to reassess whether Qualcomm can enter the core supply chain of the next computing platform.

Therefore, the real question is not whether Qualcomm has an AI story, but whether that story can overcome the pressure from its legacy businesses slowing down and translate into sustainable revenue and profits. To assess this, one cannot rely solely on rumors about OpenAI or isolated statements during earnings calls; instead, one must return to the underlying business structure.

Cycle 01 Pressure and Apple Loss

First, review the latest financial report. In the second quarter of fiscal year 2026, Qualcomm reported revenue of $10.599 billion, a 3% year-over-year decline; adjusted earnings per share were $2.65. Of this, chip product revenue amounted to $9.076 billion, down 4% year-over-year, while technology licensing revenue reached $1.382 billion, up 5% year-over-year. Within the chip product segment, mobile revenue was $6.024 billion, down 13% year-over-year, while automotive and IoT revenue increased by 38% and 9% year-over-year, respectively.

These figures show that Qualcomm's diversification has begun to appear in its financials, but has not yet replaced the smartphone business as its core. Smartphones still account for about two-thirds of chip product revenue and more than half of the company’s total revenue. While automotive and IoT segments are growing rapidly, they remain insufficient to fully offset the decline in smartphones. As long as the smartphone business weakens, the market will continue to value Qualcomm primarily according to the mobile chip cycle.

The guidance for the next quarter further reinforces this point. Qualcomm expects revenue for the third quarter of fiscal year 2026 to be between $9.2 billion and $10.0 billion, with adjusted diluted earnings per share of $2.10 to $2.30. The company stated that the guidance incorporates the impact of storage supply constraints and related pricing on demand from some smartphone customers, and expects smartphone revenue from Chinese customers to bottom out in the third quarter.

This is not merely a Qualcomm-specific operational issue, but rather a spillover effect from the AI infrastructure cycle onto the consumer electronics supply chain. Counterpoint Research reports that global smartphone shipments declined by 6% year-over-year in the first quarter of 2026; Gartner also forecasts that soaring storage costs will suppress PC and smartphone shipments in 2026. As data centers absorb storage capacity, material costs for consumer electronics rise, ultimately impacting smartphone procurement timelines and demand for mid- to low-end device upgrades.

Within the mobile business, Apple remains the most critical structural variable. In its Q2 2026 10-Q filing, Qualcomm disclosed that Apple has already begun using its own baseband chips in some smartphones and expects Apple to increasingly adopt its own baseband chips in future devices rather than Qualcomm’s products, which will have a materially negative impact on chip product revenue, operating results, and cash flow.

The issue with Apple isn't just about selling fewer baseband chips. Qualcomm also disclosed that Apple purchases standalone or slim baseband products that do not include Qualcomm’s integrated application processor technology, resulting in lower revenue and profit contribution compared to a full integrated platform. If Apple devices equipped with Qualcomm’s standalone basebands capture market share from other customers using integrated platforms, Qualcomm’s revenue and profit margins would also be affected.

This means Qualcomm has lost not only shipment volume but also a portion of the high-end mobile connectivity profit pool. More importantly, Apple’s in-house baseband chip development reinforces the trend of vertical integration among leading device manufacturers. Qualcomm also noted in its 10-Q filing that major customers such as Apple, Samsung, and Xiaomi are developing their own integrated circuit products, and some Chinese customers may adopt in-house chips due to supply security concerns or policy pressures.

However, Qualcomm is not without defenses. Its licensing business generated a 5% year-over-year revenue increase in the second quarter, with a pre-tax profit margin of 72%, enabling the company to continue investing in R&D and returning value to shareholders during a downturn in the mobile cycle. In the second quarter, Qualcomm returned $3.7 billion to shareholders and announced a new $20 billion share repurchase authorization. Yet, the licensing business can only help it navigate the cycle—it cannot alone answer where Qualcomm’s future growth will come from in light of AI’s revaluation of semiconductors.

02 From Edge Inference to Data Center Custom Chips

Qualcomm's disadvantage in this wave of artificial intelligence is that it has not positioned itself at the center of training compute. NVIDIA controls GPUs and their accompanying software ecosystems, while companies like Broadcom and Marvell benefit from custom accelerator and networking chips for cloud providers. AMD and Intel compete for market share in server CPUs and accelerators. Qualcomm has long emphasized edge-side AI but has not been a primary beneficiary of training cluster development.

However, the second phase of the AI industry is not just about training. As model deployment scales up, inference costs, latency, privacy, power consumption, and end-user interaction become more critical. Qualcomm is betting on models moving from the cloud to smartphones, PCs, cars, XR devices, robots, industrial terminals, and edge servers. In these scenarios, low-power heterogeneous computing, cellular connectivity, Wi-Fi, Bluetooth, image processing, and on-device NPUs are more important than merely pursuing peak computational power.

The rumors about an OpenAI processor have garnered attention because of this vision. If OpenAI indeed develops a native AI personal device, the chip platform would need to handle local inference, real-time voice and visual input, low-power persistent sensing, connectivity, privacy, and mass production supply chain requirements. Qualcomm’s extensive experience in smartphone SoCs, baseband chips, RF technology, edge AI, and carrier certifications makes it a natural candidate.

However, this line of thought must be viewed with restraint. The related collaboration has not been officially confirmed, OpenAI’s hardware form factor remains unclear, and potential mass production by 2028 does not provide short-term revenue support. For Qualcomm, the rumors surrounding OpenAI resemble more of an “entry option” than a guaranteed incremental opportunity.

Compared to rumors about OpenAI’s hardware, the data center business is more noteworthy, as the company’s management has already provided a timeline. According to the earnings call transcript, Qualcomm is entering the custom chip market, initially ramping up with a leading cloud provider, with initial shipments expected to begin in December. Management also stated that this project will enhance profitability. This aligns with recent market rumors that Qualcomm is developing a dedicated data center processor based on Arm architecture.

It should be noted that rumors cannot be treated as officially released products, but Qualcomm’s official data center webpage has already revealed its direction. The company positions its data center solutions around AI inference, energy efficiency, and total cost of ownership, outlining product roadmaps such as cloud AI inference chips, and states in the “Server Processors” section that it is developing data center processor solutions.

Qualcomm's return to the data center does not equate to a direct challenge against NVIDIA’s training GPUs. A more realistic path is to focus on customization around inference workloads, specific cloud providers, and dedicated systems. The barriers to training clusters lie in GPUs, high-bandwidth memory, interconnects, and software ecosystems; the inference market is more fragmented, with customers selecting based on model size, latency, cost per token, and power consumption. If Qualcomm can leverage its low-power NPUs, optimized memory access, and SoC integration capabilities to build rack-level inference systems, it has an opportunity to differentiate itself.

Changes in the Arm server ecosystem have also created an opportunity for Qualcomm. Arm is set to release a general-purpose AI processor for data centers in March 2026, developed in collaboration with Meta as a key partner, featuring up to 136 Arm Neoverse V3 cores, targeting agent-based AI infrastructure. This signals that major cloud customers are increasingly adopting Arm processors and custom silicon to pursue greater energy efficiency and cost optimization.

This is also the biggest difference between Qualcomm’s current approach and its Centriq era. Back then, it tried to challenge the x86 general-purpose server market, facing barriers in ecosystem, distribution, and customer trust; today, cloud providers are more receptive to custom silicon, and AI inference has created new efficiency demands. However, data center validation cycles are long and software stack requirements are high—Qualcomm still needs to prove it can deliver sustainable, scalable system-level solutions.

03 Automotive and IoT: The Second Curve

Among all non-mobile businesses, automotive is Qualcomm's clearest second growth curve. In the second quarter of fiscal year 2026, Qualcomm's automotive revenue reached $1.326 billion, a 38% year-over-year increase and a quarterly record. The earnings call transcript also showed that automotive revenue for the quarter surpassed the $5 billion annualized level for the first time, with management expecting annualized revenue to exceed $6 billion by the end of fiscal year 2026.

However, what’s more worth watching is not the numbers for this quarter, but the industry developments over the past few months. During CES 2026, Qualcomm announced an expanded partnership with Google on automotive software and AI experiences; simultaneously, its flagship cockpit and advanced driver assistance platform secured projects with leading automakers including Li Xiang, Leapmotor, Zeekr, Great Wall, NIO, and Chery, bringing the total number of design wins to ten. Platforms like Ride Flex, which integrate cockpit and ADAS workloads onto a single SoC, have also entered multiple mass production projects.

In April, Bosch and Qualcomm expanded their smart cockpit collaboration to include advanced driver assistance systems. Bosch disclosed that it has delivered over 10 million vehicle computers based on Qualcomm’s cockpit platforms; the new partnership will leverage Qualcomm’s driving platform to enable mass production of ADAS features and explore integrating the cockpit and ADAS functions onto a single SoC using Ride Flex. This indicates that Qualcomm’s automotive business is not merely selling cockpit chips, but is actively participating in the automotive industry’s transition toward centralized computing architectures.

IoT is not merely an revenue category in financial reports. In January, Qualcomm announced the completion of its expansion in the industrial and embedded IoT business, integrating multiple acquired assets into its portfolio with the goal of bundling processors, software, developer tools, and industry-specific solutions. Particularly in video security, industrial endpoints, local inference, and offline AI scenarios, Qualcomm aims to create an edge AI platform by combining CPUs, NPUs, connectivity, and software tools.

From a financial perspective, automobiles and IoT cannot yet fully offset the decline in smartphones; from an industrial standpoint, they have already demonstrated that Qualcomm is not solely dependent on smartphones. What Qualcomm truly needs to do is reuse underlying technologies across smartphones, automobiles, IoT, PCs, and data center inference, unifying Oryon CPU, Hexagon NPU, baseband RF, and connectivity technologies into a single platform rather than keeping them fragmented across different product lines.

04 Conclusion

Qualcomm's situation is not complicated: its smartphone base is affected by market cycles and Apple's in-house modem chips, but rumors of OpenAI hardware, timelines for data center custom chips, record growth in its automotive business, and the expansion of IoT toward edge AI have led the market to see its potential to enter the next computing platform.

Therefore, Qualcomm has neither missed the AI wave nor fully turned things around. Over the next two to three years, the key lies in three factors: whether smartphone profit margins can be stabilized, whether automotive and IoT segments can continue to scale, and whether data center inference can evolve from a single-customer project into a replicable business. Only if these are achieved can Qualcomm transform from a leader in mobile chips into a more comprehensive computing platform company.