MW Qualcomm takes a big tax charge that hits earnings, and the stock falls
By Britney Nguyen
The chip company's results were upbeat when excluding the charge, with its CEO cheering strong business momentum and new opportunities in data centers
Qualcomm's fiscal fourth-quarter revenue beat the analyst consensus.
Qualcomm Inc. took a big tax-related charge that dragged down earnings in its latest quarter, but which could improve its tax positioning going forward.
The semiconductor company disclosed a $5.7 billion noncash charge for its fiscal fourth quarter, which it said will let it "establish a valuation allowance against our U.S. federal deferred tax assets" in the future.
Qualcomm Chief Financial Officer Akash Palkhiwala said on the company's earnings call Wednesday that it expects "lower cash tax payments relative to prior expectations."
That charge - which follows a similar one that hit Meta Platforms Inc.'s (META) earnings last week - meant Qualcomm swung to a GAAP net loss of $3.1 billion, versus the $2.9 billion in net income that it posted a year earlier. On a per-share basis, the company posted a $2.89 loss - though after adjustments for one-time impacts like the tax charge, the company earned $3.00.
Qualcomm is taking the charge now in conjunction with new U.S. tax legislation, but expects "to be subject to the U.S. corporate alternative minimum tax beginning in fiscal 2026" now that it's set up the valuation allowance.
Shares of Qualcomm $(QCOM)$ fell 2.6% in after-hours trading Wednesday.
The company reported revenue of $11.3 billion for the fiscal fourth quarter, topping the FactSet analyst consensus of $10.8 billion. In a release, Chief Executive Cristiano Amon pointed to record revenue for the company's Qualcomm CDMA Technologies business, which designs and sells chips, software and hardware for wireless devices.
The company saw 18% growth from the previous year in total QCT revenues, when excluding Apple Inc. $(AAPL)$, Amon noted.
The company's handset segment saw revenue of $6.9 billion, which came in higher than expectations for $6.6 billion, while automotive revenue was $1.05 billion, slightly ahead of the $1 billion analyst estimate.
"We are excited about our business momentum, the availability of our automated driving stack and our expansion to data centers and advanced robotics," Amon said.
The company set its fiscal first-quarter revenue guidance between $11.8 billion and $12.6 billion, which was above the analyst consensus of $11.6 billion. The midpoint of Qualcomm's outlook for adjusted earnings of between $3.30 and $3.50 per share for the December quarter also topped consensus estimates for $3.32.
See more: Qualcomm ups its game in AI, sending its stock to a 15-month high
In October, Qualcomm's stock reached its highest close since July 2024 after the company announced two chip-based accelerator cards and data-center racks as part of a road map to launch multiple generations of solutions for artificial-intelligence inferencing, or the process of running AI models after training, on an annual cadence.
The AI200 is expected to be commercially available starting next year, and marks the company's introduction of a "purpose-built rack-level AI inference solution" with lower total cost of ownership while still optimizing performance for large language and multimodal models, Qualcomm said.
The AI250, which is expected to follow in 2027, will have "an innovative memory architecture based on near-memory computing," the company said, allowing it to deliver "greater than 10x-higher effective memory bandwidth" and lower power consumption.
Qualcomm announced that Saudi Arabian AI firm Humain will be the first customer for the products starting next year, and has a goal of deploying 200 megawatts' worth of the rack-scale systems.
"With a deep-pocketed customer like Humain, the product has a potential to generate billions in revenue," Melius Research analyst Ben Reitzes said in an October note. He noted that "Humain's ability to scale isn't yet proven."
Reitzes also said that Qualcomm's specifications for its offerings "seem to be much lower than" products from Nvidia Corp. (NVDA) and Advanced Micro Devices Inc. $(AMD)$ given its use of a type of random-access memory called low-power double data rate, or LPDDR, which is typically used in mobile devices, instead of high-bandwidth memory. He also noted that Qualcomm is using the Peripheral Component Interconnect Express, or PCLe, high-speed interface standard for scaling up, while Nvidia uses its NVLink interconnect.
The Melius team modeled Qualcomm's business with Humain scaling to between $1 billion and $2 billion in annualized revenue in the latter end of the decade, although Reitzes noted the company's margins could trend lower than average.
With Qualcomm seemingly wanting in on the coming inferencing boom, Reitzes said "the key for [Qualcomm] from here is whether any other nonsovereign customers will be interested, indicating broader adoption."
Bank of America's Tal Liani said in an October note that the announcement was "a needed diversification away from the low-growth smartphone market," which makes up about 75% of the company's QCT revenues. The move also "is addressing a market that we expect to grow to $114 billion by 2030, with key customers looking for vendor diversity outside of the likes of Nvidia," Liani wrote.
However, Liani said Qualcomm's opportunity going into 2026 "is limited with a single deal" and it needs to show it can execute with its technology.
Meanwhile, Reitzes said that Qualcomm's stock multiple has been negatively impacted by concerns over the loss of its modem business with Apple. Despite having "great chipset and modem technology," that multiple has been compressed to 10x, he noted.
-Britney Nguyen
This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.
(END) Dow Jones Newswires
November 05, 2025 20:20 ET (01:20 GMT)
Copyright (c) 2025 Dow Jones & Company, Inc.