Intel Announces Bold Software Strategies In Artificial Intelligence (AI)

At Intel’s Innovation 2022, the company’s message on AI took a very interesting turn. Intel was very vocal about its determination to overhaul its AI software strategies. Intel’s new AI strategy, announced at the event, revolves around three key pillars:

• Openness: Intel continues to invest in software solutions that are hardware agnostic. OpenVINO, Intel’s machine vision deep-learning inference framework, will support even more edge AI inference chipsets. Through oneAPI, Intel intends to offer the most open cross-architecture programming model for AI chipsets for all major vendors. The company even announced an overhaul of the governance structure for oneAPI. Through these efforts, Intel hopes that developers can appreciate the openness and the ease of development of using Intel’s AI software tools, leading them to migrate to Intel from more popular and established competitor solutions.
• Trust: Intel has time and again highlighted its expertise in hardware security. The aim is to reassure developers and ecosystem partners that Intel will pull out all stops to ensure the safety and security of innovative ideas, mainly through hardware-based memory encryption under its Software Guard Extensions solution. This solution isolates specific application code and data in memory, preventing unauthorized access and deterring malicious behavior by larger companies that wish to prey on ideas from smaller startups and independent developers.
• Choice: Intel wants to position itself as the most open enabler for AI developers. It already has several open software solutions. Still, the company intends to provide AI developers with even more choices in the future through a wide range of industrial partnerships and minimal, if not zero, proprietary solutions that cause vendor and developer lock-in. To assist with this, Intel has acquired Granulate, a data center workload optimization startup, and cnvrg.io, a heterogeneous computing platform for data science and AI workloads. Both are offering hardware-agnostic solutions across multiple AI hardware vendors.

Intel has been facing several headwinds in AI. First, the company failed to integrate and sustain many of its prized AI acquisitions, such as Nervana and Movidius. Second, the company has not been competitive in many AI verticals. It is entirely left out of consumer device AI. In automotive, while the company is doing well with Mobileye, most of the attention still goes to other rivals that position themselves as thought leaders in other verticals, including healthcare, robotics, metaverse, and enterprise automation. Third, more and more cloud AI giants are exploring alternatives to the x86 Central Processing Unit (CPU) architecture and are introducing their own Arm-based CPU and AI processors. Lastly, the AI hardware startup scene is extremely crowded with new approaches to AI computing.

These combined factors mean it will take significant effort from Intel to be recognized as a thought leader in AI. In response, Intel has decided to lean heavily into software. The company is squarely focused on leveraging its software expertise by identifying areas for monetization. Intel’s Developer Cloud is a good head start as it provides access to Intel products from a span of a few months up to a full year ahead of the product, allowing developers to create and optimize their AI solutions across various hardware configurations. In addition, Intel has also introduced Geti, a low-to-zero code development platform for AI-based computer vision applications. The solution works with deep-learning models in several formats, including Google’s open source TensorFlow framework and open source OpenVINO toolkit files for Intel CPUs, Graphic Processing Units (GPUs), and Vision Processing Units (VPUs). Aside from Intel Developer Cloud and Geti, Intel has plans to further monetize more of its software efforts, likely through close partnerships with ecosystem partners and system integrators.

No doubt software monetization requires a significant shift in Intel’s business model. Many Intel executives are still talking about their AI strategies at a high level, emphasizing that their plan for 2022–23 is ready to execute, although they have not gone deeper into specifics. ABI Research estimates that there is close to US$500 billion of total addressable market for AI chipset, software, and services in 2027, and Intel needs to act fast to capture these market opportunities.

In the end, AI market leadership still relies on competitive hardware and software. By launching new AI inference chipsets—namely, Flex GPU and 12th generation CPU for Internet of Things (IoT) devices to complement its data center CPU and GPU and Habana Labs accelerators—the company finally has a complete lineup of AI hardware that is allowing it to go head-to-head against major rivals. At the same time, Intel appears to have identified the right software strategy. The company’s open AI software tools and services will enable developers to simplify and accelerate AI application development while finding the right hardware solutions to maximize the performance of these applications. Alongside a customer-first mentality and culture, Intel should have the right ingredients to be competitive for AI thought leadership.