Public cloud providers get into the chip market

The world of public cloud computing is at an inflection point. Cloud providers face significant challenges amid a global surge in demand for GPUs, driven by artificial intelligence and other data-intensive workloads. There is no easy path, given the shortages and the rising demand. So, what can be done? When in doubt, build your own.

The GPU shortage, compounded by pandemic-era supply chain disruptions and increased competition for high-end chips, has forced public cloud giants to turn inward. Companies like Microsoft, AWS, and Google are no longer settling for off-the-shelf hardware; instead, they are leading the charge in custom chip development. This shift, initially a response to immediate shortages, has the potential to reshape the cloud industry and the broader CPU and GPU markets. As cloud providers dive deeper into chip innovation, their advances will accelerate competition and disrupt long-standing industry dynamics.

The rise of custom silicon

Public cloud providers have always been innovators. Take Microsoft, for example. During its annual Ignite conference, the company revealed two new chips that promise to push the boundaries of what’s possible on the Azure platform. The Azure Boost DPU (data processing unit) is designed for optimized data handling to support AI workloads more efficiently. At the same time, the Azure Integrated HSM (hardware security module) enhances security for encryption and key management processes. Both aim to address the cloud ecosystem’s specific challenges—performance optimization and security—while relying less on traditional GPU and CPU supply chains.

Microsoft isn’t the only one making waves. AWS has already established itself with custom chips such as Trainium for machine learning training, Inferentia for inference workloads, and its Nitro system for advanced virtualization and security. Google brought its Tensor processing units (TPUs) to market years ago as a custom solution for machine learning tasks. Combined, these innovations are not just filling the gap left by conventional GPUs, they’re redefining how we think about workloads at scale.

Other industry players are following suit. Nvidia, best known for its GPUs, has introduced Bluefield chips, and AMD is leaning into its Pensando portfolio. The result is an ecosystem increasingly reliant on custom accelerators tuned to specific tasks—a far cry from the days when a handful of brands dominated the chip market.

Fixing GPU shortages and more

The overarching motivation for investing in custom silicon is clear: Traditional GPUs, while powerful, are often too power-hungry, expensive, and general-purpose to handle the nuanced demands of modern cloud computing. With more demand than supply of Nvidia’s GPUs, for instance, alternative solutions like custom-designed chips offer more control over price-performance ratios, energy efficiency, and cooling requirements.

Furthermore, custom chips allow for the creation of highly specialized systems that outperform general-purpose processors for specific workloads. Security chips such as Microsoft’s HSM, AWS’s Nitro, and Google’s Titan, for example, demonstrate how tailored hardware can solve niche problems, such as reducing latency in encryption tasks or validating system health from the hardware level. By moving security into dedicated silicon, these providers are enhancing scalability, cutting costs, and improving customer trust—three imperatives in today’s competitive cloud market.

This is also a workaround for the geopolitical issues many industries are facing. Custom chips allow public cloud providers to sidestep many of the production roadblocks facing traditional CPU and GPU vendors. Existing supply chains are grappling with American trade disputes with China, the fact that most semiconductor production is concentrated in Asia, and tariffs on high-tech goods. By taking their silicon fabrication needs into their own hands and innovating with partners, cloud providers gain an edge.

This will have profound consequences for the larger CPU and GPU industries. First and foremost, direct competition will increase. Intel, AMD, and Nvidia had a virtual monopoly on the chip market for years. But as public cloud providers expand their custom silicon offerings, an entirely new tier of competition emerges. Cloud providers are essentially becoming chipmakers.

Indirect competition is also ramping up. Competition from cloud providers pressures traditional chip companies to innovate faster as they defend their market share. If AWS or Google can provide next-gen computing power without relying on Nvidia, how long will other enterprises continue to pay premium prices for legacy hardware? Likewise, the hyperscalers’ expertise in specialized silicon design could lead to a broader trend: enterprises building tailored hardware in collaboration with semiconductor manufacturers.

It doesn’t stop there. Custom silicon may drive innovation outside the cloud industry entirely. Consider the potential for these chips to impact automotive automation, robotics, or even personal devices. Furthermore, because custom accelerators prioritize efficiency and specialization, they could help mitigate the environmental costs associated with high-end GPU manufacturing and operations, creating a net benefit for industries beyond tech.

Expanding the silicon race

In-house chip development comes with its own set of challenges. Designing, testing, and producing semiconductors is no small feat. Cloud providers are entering a capital-intensive, high-stakes arena typically dominated by firms like TSMC and Samsung. To succeed, they must partner closely with manufacturers and chip architects to ensure their specialized designs can be mass-produced quickly and affordably.

More importantly, regulatory scrutiny may soon become a concern. As hyperscalers acquire more control over their hardware ecosystems, questions of anti-competitive behavior will naturally arise. How these companies address these concerns without stifling innovation will be pivotal to the future of custom chips.

The growing investment in custom chips represents a radical shift for both the cloud and semiconductor industries. In many ways, it signals the beginning of an era when hyperscalers dictate cloud computing innovations and fundamentally reshape underlying hardware ecosystems. This transition won’t just enhance cloud platforms with better performance, security, and scalability—it could also open doors for indirect innovation, new partnerships across industries, and solutions to persistent trade hurdles.