Artificial intelligence companies may be preparing to enter the smartphones market, challenging established manufacturers with devices designed around powerful, always-available AI assistants. Qualcomm CEO Cristiano Amon has revealed that several leading AI companies are discussing smartphone development with the chipmaker, with ambitions to run AI models containing at least 100 billion parameters directly on a phone by 2028.

Amon shared the details during a conversation with technology journalist Alex Heath on the Sources podcast, recorded around Qualcomm’s Snapdragon Summit. He did not identify the companies involved, leaving open the possibility that established AI developers could eventually compete with Apple, Samsung and Google in the hardware market.

The development points to a potential shift in the AI industry. Instead of relying primarily on applications installed on existing smartphones, AI companies could build devices around their own models, software and assistants.

However, the ambition comes with significant technical challenges. Running a model of this size continuously on a battery-powered device would require substantial computing power, memory, and energy efficiency. Neither the identities of the prospective manufacturers nor the launch dates of their devices have been confirmed.

What Did Qualcomm CEO Cristiano Amon Reveal About AI Smartphones?

Amon said some of the companies at the forefront of AI development were talking to Qualcomm about smartphone capabilities. He indicated that these discussions included a requirement to run a model with at least 100 billion parameters on a phone by 2028.

“I have to be careful not to disclose names,” Amon said, describing the conversations with AI companies.

His comments suggest that at least some AI developers are considering a move into hardware, rather than limiting their businesses to AI models, cloud services and software applications.

Qualcomm’s role is significant because its processors power smartphones from many of the world’s leading manufacturers. The company develops mobile chipsets that combine computing, graphics, connectivity, and AI processing, making it a potential technology partner for companies looking to build AI-focused phones.

Still, discussions with Qualcomm do not establish that a commercial product is guaranteed. Amon’s comments describe ambitions communicated by prospective customers, not a confirmed product announcement from a named AI company.

Why Would AI Companies Build Their Own Phones?

Control over hardware could allow AI developers to design a smartphone specifically for their software rather than adapting their models to the limitations of existing devices.

A purpose-built AI phone could potentially integrate an assistant more deeply into the operating system, allowing it to coordinate tasks across applications, understand user preferences, and respond to context without requiring a separate app for every interaction.

For example, a user might ask an AI assistant to organize a trip. Instead of merely suggesting destinations, the assistant could potentially compare travel details, manage calendar entries, organize reservations and prepare an itinerary using authorized tools.

Such capabilities would require appropriate permissions, reliable integrations, and safeguards. Nevertheless, a device designed around these tasks could offer a different experience from a conventional smartphone with an AI chatbot added to its existing software.

What Is a 100-Billion-Parameter AI Model?

Parameters are numerical values that an AI model learns during training. They help determine how the model processes information and generates responses.

Parameter count is one indicator of a model’s scale, but it does not independently determine intelligence, accuracy, or performance. Architecture, training data, optimization and the task being performed also matter.

A 100-billion-parameter model would represent a substantial computational workload for a smartphone, particularly if the model had to remain available continuously rather than being loaded only when needed.

For comparison, Qualcomm announced the Snapdragon 8 Elite Extreme Gen 6 in September 2026, describing support for on-device models with more than 30 billion parameters using a mixture-of-experts architecture. Amon’s stated 100-billion-parameter target would therefore push beyond the capabilities highlighted in Qualcomm’s latest flagship mobile platform announcement.

How Does Local AI Differ From Cloud-Based AI?

Most people encounter powerful AI systems through services that perform at least some of their computing on remote servers. A phone sends a request to a cloud service, the server processes it, and the result returns to the device.

Local AI works differently. The model runs directly on the phone’s hardware, allowing supported tasks to be processed without sending every request to a remote server.

The two approaches offer different advantages.

• Privacy: Local processing can reduce the need to transmit sensitive information to external servers, although the result depends on how the application handles data.
• Availability: On-device models can perform supported tasks without a reliable internet connection.
• Response times: Local processing can reduce network delays, although performance depends on the model and device.
• Operating costs: Running tasks locally may reduce cloud-inference expenses, but it increases the demands placed on the phone’s hardware and battery.
• Capability: Cloud systems can use larger computing resources, while phones face strict limits on memory, power consumption, and heat.

A future AI smartphone could combine both approaches, using local models for personal context and routine tasks while relying on cloud systems for more demanding requests.

Can a Smartphone Really Run a 100-Billion-Parameter Model?

The goal is technically ambitious because model size affects memory requirements, processing speed, energy consumption, and heat generation.

Even when a model is compressed using quantization, it needs sufficient memory to store its parameters and additional working space to process requests. A 100-billion-parameter model stored at four bits per parameter would require approximately 50 gigabytes just for the parameter weights, before accounting for other memory requirements.

That is a substantial amount for a smartphone, where memory must also support the operating system, applications, and other processes.

Several engineering approaches could help address the challenge:

• More efficient model architectures: Techniques such as mixture-of-experts models can activate only part of a model for a given task, potentially reducing the computation required for individual requests.
• Quantization: Representing model parameters with fewer bits can reduce memory use and improve efficiency, sometimes with trade-offs in accuracy.
• Specialized AI processors: Neural processing units can accelerate supported AI operations more efficiently than relying exclusively on a general-purpose processor.
• Memory improvements: Faster, higher-capacity memory could help devices accommodate larger models.
• Hybrid processing: Phones could handle some tasks locally while sending more demanding workloads to cloud servers.

These approaches do not eliminate the underlying constraints. A model that fits into memory may still run too slowly, generate too much heat, or drain the battery too quickly for everyday use.

The distinction between running a model occasionally and keeping it available continuously is particularly important. A phone designed for always-on AI would need to manage background activity and power consumption without compromising ordinary smartphone functions.

Consequently, 2028 should be understood as a target identified by Amon, not a guarantee that every technical challenge will be resolved by then.

Could OpenAI or Other AI Companies Launch Their Own Phones?

Amon did not name the companies involved in the Qualcomm discussions. It would therefore be premature to attribute the reported smartphone plans to OpenAI, Google, Anthropic or any other specific developer without additional confirmation.

Speculation about AI companies entering the hardware market has nevertheless been growing.

OpenAI, for example, has attracted attention over its interest in developing new consumer hardware. Reports and industry discussions have raised questions about whether future AI devices might move beyond conventional smartphones, but those discussions should not be confused with confirmation of a specific OpenAI-branded phone.

The concept of an “anti-iPhone” has also surfaced in speculation about devices designed around AI rather than traditional app-based interactions. Such a device might prioritize an assistant that completes tasks across services instead of requiring users to open individual applications.

Whether consumers would embrace this approach remains uncertain. Smartphones are deeply established, and users already rely on them for messaging, payments, photography, entertainment, navigation, and work.

An AI company entering this market would need to offer more than a capable chatbot. It would also need dependable hardware, a polished operating system, strong security, competitive battery life, and access to the services people use every day.

Why Qualcomm Is Betting on On-Device AI

Qualcomm’s latest announcements show that the company is positioning its mobile processors for a future in which AI agents operate more closely with the phone’s operating system and applications.

At its September 2026 Snapdragon Summit, Qualcomm introduced the Snapdragon 8 Elite Extreme Gen 6 and Snapdragon 8 Elite Gen 6 platforms. The company highlighted advances in AI processing, power efficiency, and personalized on-device experiences. Its flagship Extreme platform was also promoted for running sophisticated models directly on the device.

Qualcomm has also emphasized the role of its Sensing Hub in supporting low-power, context-aware AI functions. Such technology could help devices respond to information from sensors and user activity while limiting the need for continuous cloud communication.

The potential demand from AI companies could create a new opportunity for Qualcomm. Rather than supplying processors only to established smartphone brands, the company could help AI developers build devices around their own software ecosystems.

For Qualcomm, however, the 100-billion-parameter ambition represents a future engineering target, not a capability that should automatically be attributed to its currently announced chips.

Will AI Smartphones Replace Traditional Smartphones?

Amon does not expect smartphones to disappear. He has argued that the phone remains one of the most important platforms for developing and delivering technology to consumers.

That position makes sense even as AI interfaces evolve. Smartphones already contain personal information, communication tools, location services and applications that an AI assistant could use to complete tasks.

A device that combines these capabilities with a powerful local model could make AI more useful by giving it access to relevant context, subject to user permissions.

For instance, an assistant could help prioritize messages, prepare a meeting summary, identify a suitable time for an appointment or organize information from multiple applications. The aim would be to reduce the number of manual steps required to complete everyday tasks.

However, the success of such devices would depend on whether they can perform those tasks accurately, safely, and consistently. Users would also need meaningful control over which applications the AI can access and what actions it can take without approval.

The most likely direction is not necessarily the disappearance of conventional smartphones, but a gradual change in how people interact with them.

What Happens Next in the AI Smartphone Race?

The next phase will depend on whether the industry’s ambitions translate into commercially viable devices.

Three developments will be worth watching:

1. Official product announcements: Whether any AI company publicly confirms a smartphone project, identifies its hardware partners or provides a launch timeline.
2. Advances in mobile hardware: Whether processors, memory systems and battery technology can support increasingly large AI models without unacceptable compromises.
3. Practical AI experiences: Whether AI-first devices can reliably coordinate tasks across applications while protecting personal data and giving users control.

Qualcomm’s discussions with AI developers suggest that the smartphone is becoming an important battleground in the competition to deliver more capable AI assistants. But a request for 100-billion-parameter models is not the same as proof that those models can run efficiently on a mass-market phone.

If the technical challenges can be overcome, AI companies could eventually compete with established manufacturers by building devices in which the assistant, rather than the app ecosystem, is the primary interface.

For now, the most significant takeaway is that AI developers appear to be looking beyond software. Their next competitive advantage may depend not only on how powerful their models are, but also on the hardware through which people use them.