What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Microsoft announced on January 9, 2023, that it had acquired Fungible, a company developing data processing units (DPUs) and composable datacenter infrastructure. The deal was aimed at improving the systems behind Azure—not launching a new app or a Fungible-branded product customers could buy. Microsoft said Fungible’s team would join its datacenter infrastructure engineering organization to work on DPU solutions, networking and hardware systems. The intended gains included lower latency, more server capacity per facility, better energy efficiency and lower operating costs; Microsoft did not publish customer benchmarks or a rollout schedule.
What Microsoft acquired
Fungible developed low-power DPUs and software for composable, disaggregated datacenter systems. Microsoft described the acquisition as a way to bring that technology and engineering expertise into its own datacenter infrastructure work. Its announcement said the Fungible team would join Microsoft’s datacenter infrastructure engineering teams and contribute to DPU solutions, networking innovation and hardware-system advances. Microsoft’s announcement did not describe a customer-facing service or a product launch.
That distinction matters: the strategic asset was infrastructure capability and talent. Microsoft’s aim was to shape the equipment and systems underneath cloud services, where changes can affect many servers, rather than to sell Fungible software as a standalone offering.
What a DPU does—and how it differs from a CPU or GPU
A DPU is a specialized processor for infrastructure work that moves data through a server and datacenter. Depending on the design, it may handle network packet processing, storage services, encryption, security functions, virtualization-related operations and other data movement. It does not replace a CPU: it takes on selected tasks so the CPU can focus more on operating systems and applications.
#1 Best Overall
| Processor | Primary role |
|---|---|
| CPU | General-purpose computation, including operating-system and application work. |
| GPU | Highly parallel computation, commonly used for graphics and AI workloads. |
| DPU | Selected infrastructure and data-movement tasks, especially networking and storage. |
The labels DPU, IPU, SmartNIC and infrastructure accelerator overlap across the industry, but they are not exact synonyms in every vendor’s terminology. Capabilities vary by product; a conventional SmartNIC may target a narrower set of network offloads than a full DPU platform.
How offloading can improve datacenter efficiency
In a conventional server, application work shares CPU cycles, memory bandwidth and I/O capacity with infrastructure functions such as networking, storage, security and data movement. When those functions are substantial, they compete with the work customers are paying to run. A DPU can move some of that processing onto specialized hardware and software.
- Free host resources: Offloaded infrastructure tasks can leave more CPU capacity for applications and cloud services.
- Reduce contention: Specialized processing may help with network or storage bottlenecks and improve latency for workloads that depend on those paths.
- Change fleet economics: If the offload is worthwhile, a provider may serve more work with a given server or facility footprint, or use power more effectively.
Microsoft cited offloading, lower latency, increased server density, energy-efficiency gains and cost reduction as intended benefits of Fungible’s technology. These are strategic goals, not published results from a measured deployment. A DPU also consumes power and adds hardware and operational costs. Results depend on the workload, software integration, utilization, and whether the resources saved exceed those costs. An application that is primarily compute-bound, or a lightly loaded server with little network and storage activity, may gain little.
At hyperscale, even a modest improvement per server could matter if it applies across a large fleet. That is why the relevant measures are not just peak throughput on one device, but useful work per unit of power, server utilization, rack density and the complexity required to operate the system.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Rank #2
What composable infrastructure adds
Composable infrastructure pools resources such as compute, storage, networking and acceleration so they can be allocated more flexibly instead of being permanently tied to one physical server. Disaggregation can make it easier to scale a particular resource or match infrastructure to a workload, potentially reducing underused capacity. Microsoft characterized Fungible’s approach as supporting high-performance, scalable, disaggregated datacenter infrastructure with reliability and security.
The flexibility has a cost. Separate resource pools and specialized devices require orchestration, observability, firmware, drivers and integration across servers, switches, storage and cloud-management systems. They can introduce new failure domains and make troubleshooting more difficult. Security isolation depends on the quality and maintenance of the entire device software stack; it is not guaranteed simply by adding a DPU.
Why the deal mattered to Azure
Cloud providers design infrastructure for economics and performance across a fleet. Owning more of the hardware-and-software design can give Microsoft more control over networking and storage architecture, reduce reliance on standard server designs and tailor systems to Azure workloads. Potential benefits could reach customers indirectly through service performance, capacity or cost, but the acquisition announcement did not promise a selectable Azure feature or quantify a customer saving.
The purchase also fits a wider industry shift toward custom infrastructure and hardware-software co-design: specialized CPUs and accelerators, networking systems, and offload processors can be tuned to the needs of large cloud and AI deployments. Microsoft’s FY2024 filing describes datacenter growth as dependent on infrastructure inputs including servers, networking equipment, accelerators, power and land. That broader investment context is documented in its FY2024 Form 10-K. It does not, however, establish that Fungible technology became the basis of any named later Microsoft chip or Azure product.
What Microsoft announced—and what remains unconfirmed
Microsoft’s January 2023 statement established the acquisition, Fungible’s technology focus, the team’s intended integration and the areas where the team would work. It did not provide an immediate customer migration path, public performance benchmark, product availability date or named Azure service. Microsoft’s acquisition-history page continues to list Fungible with the date January 9, 2023.
The public sources cited here do not establish whether Fungible products were discontinued, absorbed or reworked internally, nor do they document a public customer rollout attributable specifically to the acquired technology. That limits what can responsibly be said about the deal’s realized customer impact: the strategic intent is clear, but its specific public outcomes are not.
Was the purchase price $190 million?
Microsoft did not disclose financial terms. Contemporary coverage by Data Center Dynamics put the deal at roughly $190 million, but that figure is reported rather than an officially confirmed purchase price. It should not be treated as a Microsoft-reported amount or as evidence, by itself, about Fungible’s technical value or the deal’s success.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How Fungible fit a competitive market
Fungible entered a market where established suppliers offer overlapping approaches to infrastructure offload. Product names alone do not establish equivalence: buyers would need to compare supported tasks, software integration and total operating costs for their own environment.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11| Platform or approach | Positioning | Official information |
|---|---|---|
| NVIDIA BlueField | DPU platform for infrastructure acceleration, with networking, security and storage capabilities. | NVIDIA BlueField |
| AMD Pensando | Distributed-services and infrastructure-processing technology in AMD’s networking portfolio. | AMD Pensando |
| Intel IPU | Infrastructure Processing Unit family aimed at cloud and service-provider infrastructure offload. | Intel IPU |
| Conventional SmartNICs | Can suit narrower, targeted offload tasks where a more expansive DPU architecture is unnecessary. | Capabilities depend on the particular product. |
A practical evaluation should cover offload capability, network performance and latency, storage acceleration, isolation, software development tools, cloud or server-platform integration, support lifecycle, portability and total cost of ownership—including power and operational complexity. No superiority claim follows from this comparison alone.
What the acquisition means for customers and infrastructure buyers
For Azure customers, the expected path was indirect: if Microsoft integrated the technology successfully, benefits could appear within cloud infrastructure without a customer purchasing a Fungible device. The announcement itself did not establish that such benefits reached customers or quantify them.
For organizations deciding whether to deploy DPUs or IPUs themselves, the key question is whether infrastructure processing is a meaningful bottleneck and whether the value of offload justifies added hardware and software operations. High-throughput storage, network-heavy multi-tenant platforms, encryption, software-defined storage and large distributed systems are plausible candidates. CPU-bound applications, low-utilization servers or systems constrained by memory or accelerator availability may not benefit materially.
Quick Recap
- Measure CPU time spent on infrastructure services, along with network and storage utilization.
- Check compatibility with the required firmware, drivers, orchestration and observability tools.
- Estimate power, support and operational costs as well as any capacity or latency gains.
- Consider a narrower SmartNIC or a managed cloud service if a full DPU platform adds complexity without enough measurable value.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
