Qualcomm targets $15bn data centre business as Meta anchors AI expansion

Qualcomm Incorporated (NASDAQ: QCOM) has set a target of generating more than $15 billion in annual data centre revenue by fiscal 2029 as it attempts to reduce its long-standing dependence on smartphone chips. The semiconductor company has secured a multi-generation agreement to supply Meta Platforms with Dragonfly data centre processors, unveiled a wider portfolio of central processing units, AI accelerators and connectivity products, and agreed to acquire AI software company Modular Inc. in an all-stock transaction worth approximately $4 billion. Qualcomm has also doubled its fiscal 2029 target for revenue outside smartphones to $40 billion, placing data centres alongside automotive systems, industrial automation and edge computing as its principal growth engines. Qualcomm shares were trading around $212.87 on June 25, up approximately 7.8% during the session as investors responded to the expanded strategy. The market reaction reflects renewed confidence in Qualcomm’s diversification, but achieving the targets will require the company to win hyperscale customers in a market dominated by Nvidia Corporation, Broadcom Inc., Advanced Micro Devices, Inc. and internally designed cloud chips.

Why is Qualcomm placing a $15 billion data centre target at the centre of its diversification strategy?

Qualcomm’s new financial target transforms its data centre effort from an experimental diversification project into a material part of the company’s future earnings narrative. The company expects more than $5 billion of data centre revenue by fiscal 2027 and more than $15 billion by fiscal 2029, including central processing units, AI inference accelerators, custom silicon, connectivity products and related systems. Qualcomm is effectively telling investors that data centres could become a business large enough to reshape its revenue mix within three years.

The timing is driven partly by opportunity and partly by necessity. Qualcomm remains one of the world’s most influential mobile semiconductor companies, but handset demand is mature, replacement cycles can be volatile and large customers increasingly design more components internally. Apple Inc.’s continuing effort to reduce dependence on external modem technology illustrates why Qualcomm cannot rely indefinitely on smartphones to deliver premium growth.

Data centres provide a larger and faster-growing addressable market, particularly as artificial intelligence workloads increase demand for processors, memory bandwidth, networking and power-efficient infrastructure. Qualcomm estimates that data centre AI, automotive systems, industrial platforms, robotics and intelligent devices collectively represent a market opportunity of approximately $1.7 trillion by 2030. That figure is not a revenue forecast, but it explains why management is willing to expand research spending, complete acquisitions and accept the execution risk of entering markets where established suppliers already possess deep customer relationships.

The strategy also reflects a shift in artificial intelligence workloads from model training towards large-scale inference. Training has created extraordinary demand for Nvidia graphics processing units, but inference involves operating trained models continuously across cloud systems, enterprises and connected devices. Qualcomm believes its experience in low-power mobile computing can be adapted to reduce the electricity and operating costs associated with inference.

That argument is strategically plausible because power is becoming one of the principal constraints on AI infrastructure. However, superior performance per watt must be demonstrated through independently verifiable customer workloads rather than internal benchmark claims. Hyperscale customers evaluate total system cost, software compatibility, reliability, networking performance and developer support, not merely processor specifications.

How does the Meta Platforms agreement strengthen Qualcomm’s attempt to enter hyperscale computing?

The agreement with Meta Platforms gives Qualcomm something that many new semiconductor platforms lack: a large reference customer willing to plan deployments several years in advance. Qualcomm’s Dragonfly C1000 data centre processor is expected to enter production during the second half of 2028 and support Meta Platforms’ future server fleet under a multi-generation collaboration. The arrangement provides commercial validation for a processor platform that would otherwise face scepticism because Qualcomm has made earlier attempts to enter servers without establishing a durable market position.

Meta Platforms is an especially valuable customer because it operates enormous social media, advertising and artificial intelligence infrastructure. Its computing requirements span general-purpose server workloads, AI orchestration, storage, networking and custom accelerators. Securing design participation from Meta Platforms can help Qualcomm optimise its products for real operational requirements rather than developing a processor first and searching for a market later.

The agreement may also improve Qualcomm’s credibility with other hyperscalers. Large cloud and internet companies are generally reluctant to adopt unproven server processors because qualification costs are high and operational failures can affect thousands of systems. A successful deployment at Meta Platforms would provide evidence that the Dragonfly platform can operate reliably at scale.

However, production is not scheduled to begin until the second half of 2028, leaving a long interval between the announcement and meaningful revenue. Semiconductor roadmaps can change during that period, and Meta Platforms will continue working with other processor suppliers and developing internal hardware. Qualcomm must therefore meet power, performance, software and delivery milestones while competing architectures continue improving.

The agreement should not be interpreted as an exclusive commitment. Meta Platforms’ infrastructure strategy depends on supplier diversity and internal innovation, which means Qualcomm will need to win each generation on economics and technical performance. The partnership opens the door, but it does not remove the competitive auction taking place behind it.

Can Qualcomm’s Dragonfly portfolio compete with Nvidia, Broadcom and custom cloud silicon?

Qualcomm’s Dragonfly portfolio is broader than a single server processor. The company has introduced the Dragonfly C1000 central processing unit, the Dragonfly AI300 inference accelerator, High Bandwidth Compute technology, custom semiconductor offerings and connectivity products supporting speeds of 800 gigabits and 1.6 terabits. Qualcomm is presenting these components as a rack-scale platform designed to handle agentic artificial intelligence, general cloud computing and specialised customer workloads.

The C1000 is expected to use more than 250 customised processing cores and support frequencies above 5 gigahertz. Qualcomm is also targeting advanced connectivity through PCI Express Gen 7 and Compute Express Link systems, while supporting both air and liquid cooling. These capabilities are intended to address processor utilisation, memory movement and power consumption across large data centres.

The central challenge is that Qualcomm is entering several competitive markets simultaneously. Nvidia Corporation dominates AI accelerators and benefits from an extensive software platform. Broadcom Inc. has become a major partner for hyperscalers developing custom processors. Advanced Micro Devices, Inc. and Intel Corporation compete in server central processing units and accelerators, while Amazon.com, Alphabet Inc., Microsoft Corporation and Meta Platforms increasingly develop their own silicon.

Qualcomm’s strongest differentiation may be its ability to combine processors, connectivity, power-efficient design and custom engineering. The acquisition of Alphawave Semi added high-speed connectivity and chiplet capabilities, while the planned Modular Inc. acquisition is intended to strengthen the software layer. This creates the outline of a vertically coordinated platform rather than a collection of individual components.

Execution risk remains high because data centre customers do not purchase roadmaps. They purchase production-ready systems supported by mature software, validated suppliers and predictable manufacturing. Qualcomm must build an ecosystem of server manufacturers, networking providers, memory companies and software developers while maintaining an annual product cadence. Missing one generation could make ambitious 2029 revenue targets difficult to recover.

Why is Qualcomm paying nearly $4 billion for Modular Inc. rather than relying on its own AI software?

Qualcomm has agreed to acquire Modular Inc. through the issuance of as many as 19.2 million Qualcomm shares. Based on Qualcomm’s share price when the transaction was announced, the consideration was valued at approximately $3.9 billion to $4 billion. The acquisition is expected to close during the second half of 2026, subject to customary conditions.

Modular Inc. develops software intended to allow artificial intelligence models to operate across central processing units, graphics processing units, neural processing units and custom accelerators without requiring developers to rewrite applications for each architecture. Its platform addresses one of the most important barriers facing alternative AI hardware providers: software compatibility.

Nvidia Corporation’s competitive position is supported not only by semiconductor performance but also by CUDA, its software ecosystem used by developers to build and operate accelerated applications. Customers may be reluctant to move workloads to another processor when doing so requires extensive redevelopment, testing and optimisation. Qualcomm needs a credible software pathway that reduces those switching costs.

Modular Inc. therefore represents strategic infrastructure rather than a conventional software bolt-on. Qualcomm can use the platform to support its own Dragonfly products while preserving compatibility with other hardware. Maintaining that hardware-neutral positioning could attract developers who want flexibility and prevent Qualcomm’s platform from appearing like another closed ecosystem.

The acquisition price nevertheless creates a demanding return hurdle. Modular Inc. is a private company operating in a competitive and rapidly changing market, and Qualcomm is paying largely with its own stock before the target has become a mature revenue generator. The strategic logic depends on Modular Inc. accelerating customer adoption, developer engagement and time to market across multiple Qualcomm product families.

Integration will require careful handling. Developers may become sceptical if Qualcomm restricts Modular Inc.’s support for competing processors or prioritises its own silicon too aggressively. Qualcomm must extract strategic value without damaging the openness that makes Modular Inc. commercially attractive.

What role will Hugging Face play in Qualcomm’s developer and open-model strategy?

Qualcomm has expanded its relationship with Hugging Face to connect the model developer ecosystem with Dragonfly data centre infrastructure and Qualcomm-powered devices. The collaboration is expected to make Hugging Face workloads available on Qualcomm systems, simplify model deployment across devices and cloud environments, and support agents that choose where workloads should operate based on privacy, cost, latency and performance.

The relationship gives Qualcomm access to a large community of developers and millions of open artificial intelligence models. That matters because semiconductor adoption increasingly depends on how easily developers can move from experimentation to production. Hardware may offer attractive economics, but customers will avoid it if model deployment requires specialised tooling and lengthy optimisation.

Qualcomm’s edge presence could create a differentiated proposition. The company already supplies processors for smartphones, personal computers, automotive platforms and connected devices. Connecting those systems to Dragonfly data centres could allow applications to divide artificial intelligence tasks between local devices and cloud infrastructure.

Such hybrid orchestration could become important for industries that need lower latency, stronger privacy or reduced cloud expenditure. A device might process sensitive information locally while sending larger tasks to a data centre. Qualcomm’s ability to participate at both ends of that workflow distinguishes it from suppliers focused primarily on centralised computing.

The commercial risk is that developer partnerships often generate visibility before they produce substantial revenue. Qualcomm must convert software availability into cloud deployments, enterprise contracts and processor purchases. The number of supported models is less important than the number of production workloads generating recurring demand for Qualcomm infrastructure.

Does Qualcomm’s latest stock move show that investors believe the AI diversification story?

Qualcomm shares were trading around $212.87 on June 25, representing an intraday gain of approximately 7.8% after the company presented its expanded targets. The stock traded as high as approximately $224 during the session, reflecting an initially stronger response before some gains moderated. Qualcomm’s 52-week range stood between $121.99 and $259.92.

Despite the positive daily reaction, Qualcomm remained down approximately 5.9% compared with its June 18 closing price of $226.11. The shares were also about 14.5% below the May 26 close of $248.82, meaning the investor-day rebound recovered only part of the decline recorded over the preceding month.

This creates a more nuanced sentiment picture than the headline rally suggests. Investors appear willing to reward credible evidence of data centre customer demand, especially the Meta Platforms agreement and the fiscal 2029 revenue target. However, the stock’s recent weakness indicates concern about valuation, execution and the time required for new businesses to offset smartphone-related uncertainty.

Qualcomm’s current market valuation reflects neither complete disbelief nor unquestioning enthusiasm. The company trades with established semiconductor earnings, licensing cash flows and substantial diversification potential, but the largest new data centre products will not reach production until 2028. Investors are effectively being asked to value customer commitments and product roadmaps before significant revenue arrives.

The market reaction aligns with the strategic significance of the announcement, but it does not eliminate the burden of proof. Further rerating will probably depend on named customers, product sampling, manufacturing milestones, order visibility and evidence that Modular Inc. strengthens software adoption.

Can Qualcomm reach $40 billion in non-handset revenue without weakening capital discipline?

Qualcomm now expects revenue outside smartphones to exceed $40 billion in fiscal 2029, approximately double its earlier target. Data centres are expected to contribute more than $15 billion, while automotive revenue is targeted at approximately $10 billion. The remaining contribution would come from personal computers, industrial systems, networking, robotics, extended reality and other connected platforms.

The broader revenue mix would reduce Qualcomm’s exposure to smartphone cycles and customer concentration. It could also make earnings more durable because automotive and infrastructure programmes generally operate through longer product cycles than consumer devices.

However, diversification is not automatically value creation. Qualcomm must fund research, software, acquisitions, customer support and product launches across several industries with different sales models. Automotive programmes can take years to reach production, data centres require intensive qualification, and industrial markets are fragmented.

The Modular Inc. transaction also increases share-based dilution. Issuing up to 19.2 million shares represents less than 2% of Qualcomm’s outstanding equity, but the economic cost remains meaningful. Management must show that acquired software and talent produce returns that exceed the value transferred to Modular Inc. shareholders.

Capital discipline will also be tested by competitive pressure. Qualcomm may need to price early data centre products aggressively to win reference customers, reducing initial margins. It could also face higher development costs if hyperscalers require customised processors and software.

The strongest version of the strategy would use Qualcomm’s reusable intellectual property across multiple markets. Processing cores, neural engines, connectivity and software could support smartphones, computers, vehicles, industrial devices and servers, spreading development costs across a larger revenue base. The weaker version would create separate, expensive product organisations that struggle to achieve scale.

What could prevent Qualcomm from reaching its fiscal 2029 data centre revenue target?

The first major risk is timing. Qualcomm’s Dragonfly C1000 platform is expected to reach production during the second half of 2028, leaving a relatively narrow period to generate substantial revenue before the end of fiscal 2029. Any manufacturing, qualification or customer deployment delay could materially affect the target.

The second risk is software adoption. Modular Inc. and Hugging Face can improve accessibility, but they must support the tools, frameworks and performance optimisation required by large customers. Developers will not abandon established systems merely because Qualcomm offers an open alternative.

The third risk is competition. Nvidia Corporation continues expanding beyond graphics processing units into central processing units, networking and rack-scale systems. Broadcom Inc. and Marvell Technology, Inc. are deeply involved in custom silicon. Advanced Micro Devices, Inc. is strengthening its accelerator and server processor portfolio, while hyperscalers are investing in internal chip teams.

The fourth risk is customer concentration. Securing Meta Platforms is valuable, but dependence on a few hyperscalers could give those customers significant pricing and design influence. Qualcomm needs multiple customers across cloud computing, sovereign AI, enterprise infrastructure and regional data centre operators.

The fifth risk is artificial intelligence capital expenditure. Current spending is extraordinary, but customers will eventually demand measurable returns from AI infrastructure. A slowdown in hyperscaler expenditure could reduce demand precisely when Qualcomm’s products reach volume production.

Qualcomm’s target is achievable only if the company enters the market with substantial committed orders rather than hoping that future demand absorbs available products. The Meta Platforms agreement and additional unnamed customer commitments provide an important beginning, but investors will need greater visibility as 2028 approaches.

What are the key takeaways from Qualcomm’s $15 billion data centre and AI expansion strategy?

• Qualcomm’s target of more than $15 billion in fiscal 2029 data centre revenue represents a fundamental shift beyond smartphone semiconductors.
• The Meta Platforms agreement gives the Dragonfly C1000 processor an important hyperscale reference customer before production begins in 2028.
• The Modular Inc. acquisition addresses the software compatibility problem that often prevents alternative AI processors from challenging Nvidia Corporation.
• Qualcomm’s expanded relationship with Hugging Face could improve developer access and support hybrid AI workloads across devices and data centres.
• The Dragonfly portfolio combines processors, inference accelerators, custom silicon, memory technology and connectivity rather than relying on one product.
• Qualcomm’s power-efficient computing experience is strategically relevant as electricity becomes a larger constraint on AI infrastructure growth.
• The $40 billion non-handset revenue target could reduce smartphone dependence, but it requires simultaneous execution across data centres, automotive systems and industrial markets.
• Qualcomm shares reacted positively on June 25, although five-day and one-month performance remained negative after recent volatility.
• Meaningful data centre revenue remains several years away, making product delivery and customer qualification more important than the initial stock rally.
• Qualcomm can become a credible AI infrastructure challenger, but displacing established hardware and software ecosystems will require sustained investment and flawless execution.