Saturday, June 27, 2026
EN·DarkSubscribe
AI Infrastructure · News & Analysis
HomeChips & HardwareReport
Chips & Hardware · Report

Qualcomm promises $15 billion data-center AI chip sales by 2029, but chips remain in design—execution risk flagged.

Market skepticism around Qualcomm's ability to ship competitive data-center CPUs in volume; AMD's EPYC and Ampere's Altra set high bar.
Trade pressSlicast · June 27, 2026 · US · Source: Google News
importance 62

Cristiano Amon walked into Qualcomm's Investor Day on June 24, 2026 with a number designed to be repeated: over $15 billion in annual data center AI chip sales by fiscal 2029, up from $5 billion in 2027, alongside a $40 billion non-handset revenue target and adjusted EPS above $18 by that year.

Qualcomm shares rose 13.3% to 15% in after-hours trading, then gave most of it back. By the close on June 25, the stock sat at $204.90, down 7.66% from June 22 and off 9.38% over the trailing week. The catch: the chips powering that $15 billion don't exist in shipping form yet. Qualcomm's data center CPUs come online in 2028, with custom AI inference silicon still being designed. Investors are underwriting a roadmap.

Qualcomm's actual current business tells a different story. Total Q2 FY26 revenue was $10.599 billion, down 3.46% year over year, with handsets accounting for $6.024 billion of that and falling 13% under memory supply constraints and Chinese OEM weakness. Data center revenue from the recently closed Alphawave Semi acquisition isn't separately reported yet. So Amon is telling shareholders that a segment producing roughly nothing in fiscal 2026 will produce around 34% of fiscal 2025's entire revenue base three years from now.

Qualcomm also announced a $3.92 billion stock acquisition of AI software firm Modular and a multi-year CPU supply agreement with Meta Platforms. The Dragonfly C1000 is set for 2028 production, with a Dragonfly AI300 inference chip teased for later. Per the Q2 FY26 filing, the hyperscaler engagement is "on track for initial shipments later this calendar year"—language that has not yet upgraded to "shipping."

The competitive picture is sobering. Bloomberg Tech's Ed Ludlow noted that "NVIDIA absolutely dominates the market and whatever's left, AMD kind of comes up and gets the rest," with NVIDIA holding what Bloomberg defines as a technical monopoly. NVIDIA holds more than 70% market share, and both it and AMD ship new architectures annually while Qualcomm's parts remain years away. To put it in perspective, NVIDIA generated $75.246 billion in data center revenue in a single quarter in Q1 FY27—a 92% year-over-year increase. Qualcomm's entire 2029 data center target would equal just one quarter at NVIDIA's current run rate.

Advanced Micro Devices is cementing the silver medal slot with the Meta-anchored MI450 deployment. Ludlow conceded the technical case for Qualcomm, calling its parts "highly performing, very efficient chips on a dollar per token basis or a dollar per kilowatt basis." Efficiency narratives gain traction when the cost of power, not silicon, becomes the limiting factor.

That same afternoon, as Amon presented his vision, Micron Technology reported the present. Fiscal Q3 revenue of $41.46 billion beat the $35.25 billion consensus by 17.60%, with non-GAAP EPS at $25.11 versus $20.28 expected. GAAP gross margin hit 84.6%, up from 37.7% a year earlier. Per CEO Sanjay Mehrotra, the result demonstrates "the strategic value of memory in the AI era," backed by newly signed multi-year Strategic Customer Agreements locking in HBM4 demand.

Morningstar lifted its fair value estimate to $200 from $155, conceding that Qualcomm "is not expected to unseat Nvidia's dominance in AI" while still capturing meaningful share. Over the next four quarters, watch for three things: whether the unnamed hyperscaler customer gets identified, whether Meta's CPU order converts into a deposit, and whether the Modular software stack starts producing benchmarks.

Read the original
Qualcomm promises $15 billion data-center AI… · Slicast