Last week it was reported that Intel may be moving away from its stewardship of open source projects, with one Intel exec quoted as saying the company is rethinking its strategy regarding open source software.

On a recent stop of the Intel Tech Tour, Kevork Kechichian, EVP and GM of Intel's Data Center Group, acknowledged Intel having “the largest footprint on open source out there from an infrastructure standpoint," while adding the chip giant needs “to find a balance where we use that as an advantage to Intel and not let everyone else take it and run with it.”

The news may not have come as a total surprise, considering Intel’s ongoing pivot under new CEO Lip-Bu Tan to reaffirm its foundry credentials, an endeavor which has seen the dismantling of its Network and Edge Group (NEX), the shelving of proposed factory plans in Europe, the closure of its automotive business, and multiple layoffs at the American giant.

The exec later clarified to The Register that the new strategy doesn’t necessarily mean Intel is closing the door completely on open source, but that the company is instead working out how to make the most of its investment in open source contributions for itself, as opposed to the fruits of its labor being of a greater benefit to the ecosystem.

An Intel spokesperson would go on to reaffirm the company was “deeply committed to open source” while honing its “focus on where and how we contribute – ensuring our efforts not only reinforce the communities we’ve supported for decades but also highlight the unique strengths of Intel."

That contribution has seen various use cases over the years, beginning with contributions in the early 1990s to what is today known as the Unified Extensible Firmware Interface (UEFI), the bridge between a computer’s hardware and its operating system (OS). Via the Intel Open Source Technology Center (OTC), the firm would later focus on the core of a Linux OS, the Linux kernel, becoming the single largest corporate contributor to the software over time.

Intel also opened up its Intel Math Kernel Library (MKL), optimized, high-performance mathematical functions for developers to use in their applications, under the name of OneMKL. While open source in nature, the OneMKL’s underlying performance libraries remain Intel's proprietary technology and have included a CPUID check, which would impose the most basic function when used in a non-Intel CPU setup, regardless of any parameters in place. This could be a bellwether, perhaps, of Intel’s strategy going forward.

An open and heavy Intel

More recent years saw Intel lean in on HPC-heavy use cases with its open source contributions, such as its quantum software development kit (SDK). Released in 2022, the kit allows developers to interface with Intel’s quantum computing stack to program new quantum algorithms for executing qubits in simulation and on real quantum hardware in the future.

A few years prior, Intel, Alibaba, Cisco, Dell EMC, Meta (then Facebook), Google, Hewlett Packard Enterprise (HPE), Huawei, and Microsoft formed the Compute Express Link (CXL) consortium, revolving around the high-speed, low-latency interconnect which removes the bottlenecks in computation-intensive workloads such as AI training.

These initiatives were from the last days of Pat Gelsinger’s CEO tenure, a time where Intel was arguably expanding its software prowess as opposed to fine tuning its foundry, as is currently happening under Tan.

Initiatives such as Tiber Developer Cloud came from this period, based on an open-source software foundation with oneAPI, to allow small- and mid-sized businesses (SMBs) to train models without increasing costs and enable larger companies to perform bursting techniques which support peak loads.

Intel and RAN

It would be remiss to not mention Intel’s established footprint in open radio access network (RAN) technology with its FlexRAN software, which analysts such as Téral Research have claimed dominates the virtualized radio networks (vRAN) and open vRAN markets.

Introduced more than a decade ago, FlexRAN is Intel’s reference vRAN software platform, designed to help operators build and run 4G or 5G base stations on standard Intel servers instead of relying on proprietary hardware.

FlexRAN, used by the likes of Dish Network, Rakuten Symphony, and Telefónica, implements open interfaces so it can interoperate with other open RAN components from different vendors. But the platform is not fully open source as core components, such as lower-level signal-processing elements, are provided as prebuilt and proprietary Intel software, similar to elements in its OneMKL provision. Ultimately, the offering is dependent on Intel's hardware and instruction sets, driving up demand in Intel’s Xeon processors.

This may provide another steer for Intel’s next steps in its open source patronage. But a more timely example may come in the form of OpenVINO, with "VINO" not standing for the Italian word for wine, but for visual inference and neural network optimization.

Will OpenVINO age like a fine wine?

OpenVINO allows devs to optimize applications for low power consumption, low latency. and deployment on existing hardware at the edge, enabling standard hardware that is already deployed at the edge to run AI applications without expensive and drastic overhauls.

The open-source kit was used to drive a new modular, edge-native software platform from Intel released last year, while the OpenVINO Model Server is designed for cloud and server deployments, compliant with hyperscaler cloud storage.

For certain use cases of OpenVINO, such as redistributing the entire toolkit with hardware, the Intel Distribution includes a separate, more restrictive license agreement for the end user.

With the increasing demand for AI inference encroaching further on the edge, OpenVINO may prove to be a versatile and leading tool for Intel with its ability to accelerate deep learning models across Intel's diverse hardware ecosystem, meaning it may age like a fine wine as the current age of AI accelerates ever further across networks.

While Intel may no longer be heading toward AI on the software front, it is likely closing the loop on open source to lead clientele towards its AI-centric chips, with AI-centric software as the bait.