In its latest move to boost the scalability and performance of high-performance workloads, Microsoft has acquired data processing unit (DPU) startup Fungible. Microsoft, which announced the acquisition on Monday, said Fungible’s engineers would join Microsoft’s datacenter infrastructure engineering team.
Former Juniper CEO Pradeep Sindhu and one-time Apple SVP for engineering Bertrand Serlet founded Fungible in 2016. While neither company disclosed the terms of the deal, reports last month indicated that Microsoft had acquired Fungible for $190 million. Fungible has reportedly raised over $300 million in venture funding from SoftBank’s Vision Fund and Northwest Ventures.
Expect demand for DPUs, microprocessors that run in servers and data center infrastructure, to accelerate with the rise of datacentric applications. DPUs, also known as “Smart NICs,” are designed to process data significantly faster than CPUs.
According to a forecast by Allied Market Research, the DPU market in 2021 was $554 million. At a projected 27% compounded annual growth rate, look for the market to reach $5.5 billion by 2031. Besides its veteran founders, Fungible tapped former Broadcom and Marvell exec Eric Hayes as CEO in 2021. Hayes brought on former NetApp and Rackspace executive Toby Owen as VP of product.
Last year, Fungible launched its FunPX partner program, which included the rollout of a partner portal with deal registration. Fungible added regional channel account teams and started offering demand-generation assets.
Why Microsoft Acquired Fungible
It remains to be seen whether Microsoft will keep the partner program intact.
“Today’s announcement further signals Microsoft’s commitment to long-term differentiated investments in our data center infrastructure, which enhances our broad range of technologies and offerings including offloading, improving latency, increasing data center server density, optimizing energy efficiency and reducing costs,” said Girish Bablani, Microsoft’s corporate VP for Azure Core.
The Fungible deal follows last month’s Microsoft acquisition of Lumenisity, which provides hollow-core fiber (HCF) technology. According to Bablani, Lumenisity offers “innovative and industry-leading HCF product can enable fast, reliable and secure networking for global, enterprise and large-scale organizations.”
Lumenisity will let Microsoft “optimize its global cloud infrastructure and serve Microsoft’s Cloud Platform and Services customers with strict latency and security requirements,” Bablani added.
Expect the addition of Fungible to extend that strategy.
“Microsoft’s acquisition of two component-level technologies signals a focus on improving the underlying compute, storage, and networking functionality in Azure,” wrote Mary Jander, a senior analyst at Futuriom.
Fungible started shipping its first DPU chips in 2020. The company claims its DPUs can disaggregate and pool hundreds, or even thousands, of compute and storage servers. Fungible’s F1 DPU is an 800 Gbps processor that the company designed for high-performance storage analytics and security appliances. The Fungible S1 DPU is a 200 Gbps processor for standard compute servers running in host sites. It offloads all infrastructure services from the CPU.
Since then, Fungible has launched appliances powered by its DPUs. Among them include the Fungible Storage Cluster, which the company upgraded last year. Fungible last year also rolled out GPU-Connect, designed to connect any GPU to any server linked via an Ethernet network.
Nevertheless, Fungible faces much larger rivals making DPUs, including Nvidia, which offers its BlueField DPUs. Last year, AMD jumped into the DPU race, acquiring Pensando for $1.9 billion. VMware last year optimized its new vSphere 8 for DPUs, while Dell, HPE and Lenovo are planning to offer servers with DPUs.
Jander noted that all major chip suppliers have relationships with AWS, Microsoft and Google Cloud. Similarly, she underscored that “AWS has its own silicon and custom Intel processors for streamlining basic cloud service functions.” For its part, she noted that Google Cloud is using Arm’s Ampere Altra chips to augment its infrastructure. Therefore, Microsoft was compelled to move similarly.
For the major cloud providers, “the advantage of owning the underlying technology,” Jander emphasized, is that it “ensures supply, reduces wait times, and enables more competitive technology — at least in theory.”
This article first appeared on our sister site Channel Futures.