Qualcomm shares climbed 3.2% on Tuesday after chief financial officer Akash Palkhiwala described the company’s chip agreement with Amazon as central to its plan to reach $15 billion in annual data-center revenue by fiscal 2029.
Speaking at the Goldman Sachs Communacopia and Technology Conference in New York, Palkhiwala said the first revenue from the Amazon agreement should arrive in the December quarter. The deal covers successive generations of custom processors built for Amazon’s large-scale AI data centers, where the chips will handle inference, the task of running trained AI models to produce predictions. The company first presented the $15 billion data-center target at its investor day in June, and Tuesday’s comments were the first update on timing since then.
Why the deal matters to Qualcomm
Qualcomm’s core business remains smartphone chips, and that business is mature. The company’s latest quarterly revenue totaled $9.9 billion, of which its CDMA Technologies division, covering handsets and automotive, contributed $8.5 billion. Qualcomm Technology Licensing added $1.3 billion. Phones pay the bills, and they will for years, which is exactly why management keeps pointing investors at everything else.
The data-center effort is the most visible part of a broader attempt to reduce dependence on handsets. At the June event the company lifted its fiscal 2029 projection for revenue outside handsets to $40 billion, nearly twice its previous forecast, a number that assumes the data-center line and automotive both compound at rates the handset market has never shown.
The near-term milestone is $5 billion from data centers in fiscal 2027, which begins at the end of this month. Palkhiwala expressed strong confidence in hitting that number, and the stock reaction suggests investors took the comment seriously. Qualcomm shares have re-rated through the year as the market warmed to the idea that the company can sell into AI infrastructure rather than only phones, a shift few analysts predicted two years ago.
The inference niche
Qualcomm is not trying to outbuild Nvidia in training clusters. Its pitch rests on custom inference silicon, where power efficiency matters more than raw compute and where cloud operators want alternatives to a single dominant supplier. Amazon already builds its own Graviton server processors and Trainium AI accelerators, so a Qualcomm role fits the company’s pattern of splitting workloads across multiple silicon vendors. Neither company has disclosed the financial terms, and neither has said which Amazon services will run on the chips first.
The custom-chip market has crowded quickly. Broadcom and Marvell sell design services to hyperscalers, and a wave of startups backed by venture money compete for the same contracts. Amazon, Google, Microsoft and Meta each run internal chip programs, and every one of them wants a second and third supplier to keep pricing honest. What Qualcomm adds is its low-power design heritage from mobile, which translates directly into inference economics, since data-center operators pay for electricity on every token a model generates. A chip that delivers the same throughput at lower wattage wins on cost per query, and that is the metric cloud operators actually track when they pick silicon.
What to watch
The December quarter will show whether the Amazon revenue lands on schedule. After that, the proof point is customer breadth: one hyperscaler agreement makes a story, two or three make a business line. Qualcomm has hinted at additional design wins without naming customers, and its fiscal 2027 target leaves little room for a single-client dependency. Missing the first milestone would undercut the entire diversification narrative the company has been building since June.
Execution risk is real. Custom silicon programs routinely slip a year or more between contract signing and volume production, and the company must hire and retain the engineering talent to tape out successive generations on schedule. The fiscal 2027 clock starts in weeks, not quarters, so any delay in the December quarter numbers will show up immediately against the $5 billion path.
For Amazon, the deal deepens a supply chain it has been diversifying for years. For the chip industry, it is another signal that inference has become a market of its own, large enough to sustain specialists alongside Nvidia’s grip on training.
discussion