MLOps Market Faces Strategic Crossroads: Edge Impulse Chooses to Expand Platform & Strategy and Is Rewarded with Developer Growth

Edge Impulse, one of the leading development platforms for Machine Learning (ML) at the edge, has recently gone through a major transformation targeting pricing model, developer capabilities, and partnership ecosystem expansion. These upgrades/expansions have been driven by: 1) customer demand—an evolving developer community and enterprise market that requires new tools to effectively support deployment of models into production; and 2) lead generation—supporting the development of new channels to customers.

• Pricing Model: Introduced a new, monthly subscription payment plan aimed at professional developers to create a bridge between community offering and enterprise plan. This enables developers to test and evaluate the platform to see if it will work for their enterprise without committing to long-term contracts.
• Developer Capabilities: Expanded edge Artificial Intelligence (AI) platform to support data processes and model monitoring, pushing the company toward an end-to-end edge Artificial Intelligence Operations (AIOps) platform.
  • Data Processes: Implemented dataset creation tools with aggregation, validation, preparation, and generation tools.
  • Model Monitoring: Enables developer model resolutions, performance monitoring, and development of Continuous Integration (CI)/Continuous Delivery (CD) with its dataset creation tools.
• Partnership Ecosystem: Added extensively to partner ecosystems with new strategic and technology partners enabling developers to build, optimize, and test AI models across competing hardware solutions prior to procurement (or even prior to hardware being available).
  • Intellectual Property (IP): Most recently, partnered with CEVA (but previously, Arm and Brainchip) expanding support to NeuPro-Nano Neural Processing Units (NPUs) to enable developers to rapidly develop and test applications before physical architecture is available.
  • Silicon: Partnerships across Microcontroller Units (MCUs), Central Processing Units (CPUs), NPUs, and Graphics Processing Units (GPUs); core strategic partnership with NVIDIA built around Omniverse, RTX technology, and TAO. Given NVIDIA’s position in the AI market, this partnership will offer a fruitful growth channel for Edge Impulse.
  • Original Equipment Manufacturers (OEMs): Among others, strategic partnership with Advantech to support ICAM-540 and MIC AI solutions for industrial AI use cases.
  • IoT Platforms: Integrated closely with edge AI Internet of things (IoT) monitoring and management tools (i.e., Amazon Web Services (AWS) and Zededa), which acts as a channel to market and reduces barriers for developers building on the Edge Impulse AI platform.
• New Technologies: Combining digital twins (through integration with NVIDIA Omniverse) and Generative Artificial Intelligence (Gen AI) (through integration with OpenAI), Edge Impulse enables the generation of synthetic data for industrial AI applications. This technology, alongside new data labeling and cleaning tools, lowers data-related costs and accelerates customer implementation of new applications, which were previously limited by data availability. In addition, Edge Impulse announced Yolo-Pro for testing and tuning with a specific subset of customers within the early access program, the first model trained for edge scenarios with extensive industrial datasets.

These new additions have already had a positive impact on Edge Impulse’s performance with continued growth in platform usage with 2.5X Year-over-Year (YoY) customer growth, and recently reaching a developer base of 160,000.

The AI/ML market over the last 2 to 4 years has been inundated with hundreds of startups. These players have brought deep and highly targeted expertise, developing products/solutions that support specific ML operations (e.g., synthetic data, optimization) or AI frameworks (e.g., predictive AI, computer vision, Gen AI). Edge Impulse is one of the first companies that is making a concerted effort to step away from this specialization with organic development of a whole suite of ML tools to support the entire developer lifecycle. However, this is just one approach to this steadily maturing market, other “startups” are at strategic crossroads and face the following choices:

• Bet on Themselves with Significant Internal Investment: Expensive and time consuming, but AI startups should look to expand their capabilities to start building end-to-end platforms that provide tools to support the AI lifecycle. Strategic      partnerships can complement internal capabilities, but are certainly not sufficient to alleviate all associated time/cost challenges.
• Present Themselves as a Mergers & Acquisitions (M&A) Opportunity: Hyperscalers, semiconductor vendors, and AI/data platforms are all looking to expand their AI software capabilities to drive differentiation and support customer engagements. This is unlikely to slow down, especially given their cash reserves, and valuations will only increase as AI software competition expands. The majority of software acquisitions originate as commercial partnerships, so this is a common channel that stakeholders should be cognizant of.
• Face the Inevitable Developer Drain: As enterprise AI deployments scale, specific tools will not scale effectively, as they often create friction between processes. Instead, developers will likely move toward an end-to-end AI platform which will more effectively scale with enterprise AI deployments. This transition will reduce developer engagement, and slow growth for tools. This can be solved with frictionless integration into end-to-end developer platforms—but of course, like anything, this is easier said than done, given the costs involved, talent shortages, and other factors.

Given the AI software market saturation, ABI Research expects that the market will go through a period of turbulence with developers shifting between solutions in favor of end-to-end platforms, and AI leaders looking to consolidate their position in the market.

Edge Impulse’s strategic partnerships, platform expansion, and new pricing model highlight one key need for edge AI developers: an end-to-end “frictionless” experience. This is not unique to edge AI developers and should be taken onboard by the entire AI supply chain. Although low/no-code tooling will reduce developer friction, this will not be sufficient to build differentiation from leading AI platforms. ABI Research recommends that software providers should look to implement and invest in the following:

• Integrations with Data, Cloud Platform Providers: Time-to-value for AI deployment is heavily impacted by the availability of data for training/tuning models. Building technology partnerships and integrations between AI development solutions and data platforms (like Snowflake) will break down the barrier for developers and help accelerate testing and development. This will also help developers ensure alignment with emerging data regulations, as it will enable them to leverage data governance tools embedded within data platforms.
• Compatibility and Support for Open-Source Tools/Frameworks: Developers have favorite tools and frameworks that they use to build AI applications. AI software platforms (providing optimization or similar Machine Learning Operations (MLOps) capabilities) should provide simple integrations and effective compatibility with open-source tools to ensure developers can utilize tools that they are familiar with for applications.
• Professional Services Supporting Brownfield Integration: One challenge facing enterprise AI implementation is brownfield deployment. Industrial, healthcare, retail, and many other sectors struggle to implement AI into their existing infrastructure. Developer platforms should build partnerships with System Integrators (SIs) and consultants to support implementation.
• Monitoring and Management Platform Integration: AI developer platforms are often limited to training, tuning, and optimization, and they lack the ability to support management of resources, for example, the IoT. Integrating with AI management and monitoring platforms (like Edge Impulse has done with Zededa and AWS) will provide an end-to-end frictionless experience from model development to scaled implementation and management. This can also help developers build CI/CD feedback loops.
• Low Barriers to Switching Hardware & Benchmarking Tools: Platforms should enable developers to build, optimize, and test AI models across hardware solutions to benchmark performance, cost, power consumption, and other factors. Providing the tools to effectively benchmark solutions will help drive developer engagement and give them choice over which hardware to implement.
• Model Optionality and Flexibility: As developer demand reduces, the cloud, vendor lock-in, and MLOps platforms must also reduce model lock-in. Increasingly, developers want to be able to seamlessly test & evaluate different AI models, and then quickly switch in and out new models with hyper-optimization.
• Providing Pre- and Post-Sales Support: Software tools and MLOps developer environments are useful, but enterprises struggle to leverage these to create tangible enterprise value. Building go-to-market differentiation requires software vendors to provide a level of development services using their customer experience; however, it is important that software vendors do not tread on the toes of their SI and service provider partners.