Have old smartphones become a hidden treasure?

Recently, with Google's support, the University of California, San Diego (UCSD) planned to construct a distributed computing platform using 2,000 decommissioned Pixel smartphones. Their goal was to create a large-scale cloud server from these second-hand devices, maximizing their computing potential while maintaining a focus on low carbon emissions and environmental sustainability.

Upon hearing this news, our initial reaction was: Are chips and computing power really that scarce? At the same time, many are undoubtedly curious: How can old smartphones be transformed into server equipment?

Unearthing Value from Second-Hand Phones: Chips and Storage as Precious Resources

According to a report by The Register, Jennifer Switzer, a former doctoral student at UCSD, collaborated with Google to repurpose 2,000 Pixel Fold smartphones donated by Google into distributed servers. The research team initially attempted to test a large number of second-hand phones together but soon realized that clustering too many batteries posed a fire hazard to the data center.

To address this, Switzer's team modified the phones by removing their batteries, casings, and components such as cameras and communication modules. Essentially, when using second-hand phones to build servers, the focus is on the motherboard and its processor, storage, and other essential parts. In addition to simplifying the hardware, the native Android system was replaced with a more efficient Linux system.

Next, these phones were grouped into computing clusters of 25-50 units each, with multiple clusters forming the final large-scale server. So, how do these numerous phones connect and communicate? The phones' native cellular networks and Wi-Fi were insufficient for this task, as connecting thousands of devices would overwhelm network signals. Researchers resolved this by using PCB boards equipped with wired network ports, along with a unified power supply to ensure stable operation and connectivity of multiple devices.

At this point, many may wonder: Can compact, power-limited smartphone SoCs handle cloud server tasks? After all, in most people's minds, servers are massive machines housed in large data centers.

In reality, smartphone computing power is not as feeble as assumed. Google's Pixel Fold, released in 2023, had a lackluster market performance and several drawbacks: it was expensive, had wide bezels, and noticeable creases. The chip used in this phone is Google's self-developed Tensor G2, with performance roughly between the Snapdragon 888 and Snapdragon 8 Gen1, making it relatively outdated for 2023.

(Image Source: Google)

However, the smartphone industry has become fiercely competitive in recent years, leading to rapid chip evolution. Chips deemed underwhelming by ordinary users are highly sought after in the server domain. Compared to mobile platforms, servers are less sensitive to chip energy consumption and heat dissipation. Once the Pixel Fold's motherboard is stripped of its casing and connected to a power source, issues of energy consumption and heat are effectively resolved.

Moreover, the Tensor G2 chip includes a Cortex-X1 super core and multiple A78 cores, outperforming many entry-level VPSs offered by cloud service providers. Critically, Google's chip also integrates 12GB of memory, along with 256GB or 512GB of flash storage on the motherboard, significantly reducing storage costs.

Additionally, the Tensor G2 was designed with AI applications in mind and integrates a TPU for edge computing, making it suitable for running small-scale local models.

Of course, using a single Pixel Fold to build a server remains unrealistic, but combining 2,000 phones results in substantial aggregate computing power. According to researchers, even a cluster of 20 phones can now support the workload of 75 students submitting assignments online.

Can Second-Hand Smartphones Alleviate AI's Computing Power Anxiety?

Frankly, expecting clusters built from second-hand smartphones to train large models with hundreds of billions of parameters is overly optimistic. However, if we shift our focus from centralized cloud supercomputing centers to decentralized edge computing, a new landscape emerges.

In Leitech's view, these micro cloud factories composed of decommissioned smartphones do not represent a downgrade in computing power; instead, they align perfectly with two core demands of future AI development: low power consumption and distributed low-latency computing.

First, they address the increasingly severe issue of high AI energy consumption. While the explosion of AI large models has undoubtedly boosted productivity, it has also led to a dramatic surge in energy consumption. Traditional centralized data centers require vast amounts of electricity for cooling and power supply to maintain the operation of massive computing clusters.

Smartphone SoC chips, on the other hand, have always prioritized energy efficiency as a core metric. Chips like the Tensor G2, which come with built-in TPU computing power, consume far less power for pure computing tasks than traditional x86 server processors once stripped of power-hungry components like screens and modems. Combining thousands of such devices not only results in extremely low carbon emissions and environmental friendliness but also distributes the enormous demand for computing power.

(Image Source: Google)

Second, they align well with the physical distribution characteristics of edge computing. As AI agents evolve and end-side application scenarios become more complex, future AI computing will no longer involve uploading all data to physically distant cloud data centers. Instead, it will favor immediate processing at the edge, closer to the user.

Decommissioned smartphone clusters are compact and flexible in deployment, no longer requiring the stringent physical space demands of traditional data centers. They can be deployed in micro-nodes within communities, campuses, or enterprise interiors. This proximity significantly reduces network latency in data transmission, making it ideal for AI inference, local model scheduling, or automated workflows that require real-time responsiveness.

Finally, this represents an attempt to address computing power costs and supply chain anxieties. Currently, supply chain prices for storage and chips fluctuate frequently, keeping hardware costs high. Meanwhile, mountains of obsolete smartphones worldwide not only waste resources but also contribute to electronic waste pollution.

Disassembling and repurposing decommissioned smartphones as components of edge computing effectively transforms former electronic waste into low-carbon cloud computing power nodes. This undoubtedly offers a lower-cost, more sustainable solution to alleviate global AI computing power anxieties.

However, while the prospects of this micro cloud factory model are enticing, its shortcomings are also apparent.

On one hand, the reliability and lifespan of smartphone SoCs and storage are inferior to those of traditional server components. The flash storage and chips in smartphones are designed for everyday use by ordinary consumers, not for the 24/7 high-intensity operation required by enterprise-grade products. Since smartphone storage chips and processors are directly soldered onto the motherboard, a failure typically means the entire node is rendered useless.

(Pixel Fold motherboard, Image Source: iFixit)

On the other hand, computing clusters composed of old smartphones face maintenance challenges. Maintaining several standard rack-mounted servers is vastly different from maintaining 2,000 exposed, piecemeal smartphone motherboards. The vast number of micro-nodes means hardware failure rates are amplified indefinitely. Frequent system crashes would require operators to expend significant effort on physical troubleshooting and motherboard replacements.

In fact, the idea of using clusters of old smartphones to build servers existed even before the AI era but was abandoned due to an unfavorable cost-benefit ratio. Today, this approach is being revisited for the reasons mentioned at the outset: storage and chip costs are skyrocketing, and computing power has become scarce. Building servers through conventional means now costs far more than before.

At the same time, due to intense competition in the smartphone industry over the past few years, the number of outdated models phased out is enormous, providing a relatively inexpensive source of materials. Reusing obsolete models is akin to mining for gold in electronic waste.

In Conclusion

Google's collaboration with UCSD is less of a computing power revolution and more of a geek experiment to address current computing power anxieties.

Amid soaring storage prices and an undersupply of AI computing power, people have become accustomed to focusing on top-tier GPUs costing tens of thousands of dollars, overlooking the vast amounts of idle mobile computing power. While constrained by factors like flash storage lifespan, these makeshift micro cloud factories assembled from second-hand smartphones cannot replace traditional data centers. Nevertheless, they provide a highly imaginative and practical case study for edge computing.

Perhaps in the near future, second-hand tablets, PCs, game consoles, NAS devices, and all other computing-capable equipment beyond smartphones will also be repurposed, leading to a restructuring of the second-hand industry chain.

Google AI AI Server Computing Power Storage

Source: Leitech

Images in this article come from: 123RF Royalty-Free Image Library Source: Leitech