Edge AI Vendors Look to Stand Out from the Crowd by Pivoting from Platform to “Turnkey” Solution

As the edge Artificial Intelligence (AI) market steadily matures, and leading-edge adopters are exhausted, vendors must look to generate demand for their products in the long tail of enterprise verticals. New target customers bring a significant change in requirements. The enterprise long tail is characterized by lower time to market expectations, lower internal AI expertise (especially at the edge), and tighter budgets for Proofs and Concept (PoCs) and implementation. This means that they demand more complete, targeted solutions, rather than edge AI platforms to support implementation. Changing demand means that edge AI chip vendors, software vendors (e.g., STMicroelectronics or Edge Impulse), and solution providers must adapt their commercial and product strategy to effectively align with enterprise pain points. Vendors must invest aggressively to make it cheaper and easier to build, deploy, and scale edge AI solutions. This means that basic model zoos, reference architectures, model playgrounds, and low/no-code tooling are no longer sufficient to differentiate and drive customer demand.

To align effectively with these new customer pain points, the market is transitioning toward vertical- and application-specific complete solutions. One key example of this strategy is Syntiant building on top of its strong product portfolio, which combines both hardware and software (capable of standalone deployment), with the recent acquisition of Knowles’ Consumer MEMS Microphone division. This acquisition aims to provide customers with an end-to-end edge audio AI solution with closely integrated sensors, processors, and software. This should simplify product development and accelerate time-to-value, while remaining customizable to fit a range of applications. Syntiant is not the only edge AI chip vendor looking to lower barriers to deployment for its customers:

• Hailo has built vertical-specific solutions targeting high-value use cases, especially within retail and industrial.
• Microchip has built partnerships with software vendors (independent software vendors and system designers) to implement a framework to support application deployment for different personas with various edge AI competency.

Chip vendors are not the only players building a more targeted product strategy. Edge AI software vendors are also looking to accelerate their Go-to-Market (GTM) strategies in the customer long tail by blending their traditional platform approach with customized solutions. SensiML and Ultralytics are both leveraging their internal expertise to enable their custom development strategies. Although valuable, supporting application-specific or custom GTM strategies requires much more skills  and is time-intensive, making them far less scalable with lower profit margins compared to the platform approach that has worked to date.

Changing customer demand means that edge hardware and software vendors must reposition themselves as solution providers. To effectively manage this transition, vendors must target their product investment to solve customer pain points. Below, ABI Research provides a breakdown of the key pain points that most verticals are facing when deploying edge AI:

• Data Availability: One of the biggest challenges for uncarpeted verticals that “should” be deploying edge AI use cases (like anomaly detection, product tracking, predictive maintenance, and environment sensing) is that they do not have enough data to effectively train and deploy models with a high degree of accuracy. For example, take a manufacturer that wants to deploy computer vision to check for screw corrosion within its product line. They will not have usable data to effectively train a model to support this case.
• Time to Market Considerations: Bringing a new edge application into production can take from months to years. This has a massive impact on Return on Investment (ROI) and is slowing/stopping use case deployment. One of the critical bottlenecks is quality testing prior to deployment. Customers do not want to put high-risk use cases into production without thorough testing and evaluation.
• Scaling Across Use Cases: Customers that have already deployed one edge AI use case struggle to effectively scale beyond adding new use cases, as the hardware/software does not effectively scale upward, bringing in management/operational challenges, especially if customers are using two different solutions from different vendors.

Building a turnkey solution that effectively addresses each of these pain points will be challenging, but will bring sufficient rewards and enable vendors to craft their GTM strategies to effectively differentiate their solutions within this saturated market space.

Changing customer preferences and requirements, alongside a saturated/competitive market space (with Original Equipment Manufacturer (OEM) contracts having, on average, between 5 and 10 edge AI hardware bidders), means that vendors must build a solution-focused product strategy. Below, we recommend several areas where edge AI vendors could invest to support a differentiated, “turnkey” and pain point-focused product strategy:

• Target Specific Verticals Across Numerous Use Cases: Manufacturing, healthcare, retail, and other similar verticals still have plenty of addressable use cases for edge AI. ABI Research recommends that vendors target 2 to 3 verticals, and provide “turnkey” solutions for different use cases. This will solve enterprise implementation challenges, but also help enterprises scale across different valuable applications.
• Aggregate Customer Data for Training: Long-tail enterprises lack access to well-structured data to support various use cases. Edge AI vendors should look to drive their “solution” strategy by aggregating customer training data to support training new use cases. Aggregating and anonymizing data will significantly reduce time, effort, and cost of deployment for customers, especially for more challenging (but valuable) use cases. However, within certain regions, this may be challenging, especially in the European Union (EU)/United Kingdom (UK), given General Data Protection Regulation (GDPR) regulation.
• Accelerate Application Quality Testing: As one of the critical bottlenecks for customers, edge AI vendors must look to build new testing and validation solutions. Working with partners like TenXer Labs may be the answer, as they can support efficient Integrated Circuit (IC) exploration and evaluation, providing a robust and efficient validation platform.
• Expand Partnerships with Verticalized System Designers & Integrators: Most vendors have limited partner ecosystem, mostly focusing on Independent Software Vendors (ISVs) or individual implementation partners. As vendors transition to customized solutions that are highly targeted to specific verticals, edge AI vendors should look to develop system designer and integrator partnership with vertical-specific players like Honeywell, which can effectively scale use cases within target customers.
• Invest in Software to Reduce Reliance on Third-Party Partners: Although hardware is vital, a key part of Syntiant’s growth story has hinged on its investment in “agnostic” software. It has developed a standalone business through its software development, which is capable of supporting competing edge AI solutions. This approach should be replicated by others seeking to: 1) build differentiation by reducing reliance on third-party software vendors like Edge Impulse, which has successfully partnered across the ecosystem; 2) develop a new commercial model that monetizes software; and 3) support OEM engagements by building differentiated hardware solutions that align with upcoming software developments, e.g., new model types. This will be expensive and challenging, but may be necessary to compete within this saturated market space.

Although less scalable, changing customer demand means that it is critical that edge AI hardware and software vendors invest effectively to develop “complete” and turnkey solutions that target vertical-specific applications. Players like Syntiant that have built effective software solutions, alongside turnkey sensor and hardware, will be best positioned to drive commercial success in the long tail of the edge AI market.