Enterprises are expected to spend US$220 billion on Generative Artificial Intelligence (Gen AI) by 2030, up from US$95 billion in 2026. Amid the AI market's ascendance, technology vendors face tougher decisions about which opportunities deserve investment, where customer demand originates, and which parts of the value chain still offer room for meaningful differentiation.
Here are seven market signals technology vendors should consider when deciding where to place their next AI bets.
1. AI Inference Is Becoming the Next Infrastructure Battleground
The Market Signal: More and more AI models are moving from development into real-world production. This will increasingly shift infrastructure demand from training toward inference. In this light, ABI Research forecasts AI inference workloads to overtake training by 2033, with inference power capacity growing from 2 Gigawatts (GW) in 2026 to 46 GW by 2035. Code generation is and will remain the top use case, reaching nearly 24 GW by 2035.
What to Watch:
- Agentic AI will increase inference demand through repeated model calls.
- Open-source models are giving enterprises greater deployment flexibility.
- Performance, multi-vendor hardware, and optimization will become important areas for differentiation.
The Business Decision: Determine where your portfolio can differentiate as AI infrastructure spending expands from model training toward running AI at scale. ABI Research Principal Analyst Larbi Belkhit notes that “AI inference providers must demonstrate business value beyond fast model serving alone. He continues, “Enterprise buyers will increasingly seek vendors that support multiple hardware environments, fit naturally into their preferred deployment models, and cover the broader model lifecycle without adding operational complexity.”
2. Open Source AI Models Are Changing Where Vendors Can Differentiate
The Market Signal: Open-source models accounted for 35% of global enterprise Generative AI (Gen AI) spending in 2025, according to ABI Research data. This year, we anticipate that figure to jump to 55%, quickly leapfrogging proprietary models. Key investment drivers for enterprises include greater cost efficiency, scalability, and deployment flexibility.
What to Watch:
- Greater model choice could make differentiation at the model layer more difficult.
- Infrastructure performance, optimization, orchestration, and integration provide other opportunities to stand apart.
- Enterprises increasingly expect AI platforms to accommodate different models, hardware environments (beyond NVIDIA alone), and deployment strategies.
The Business Decision: Reassess which parts of your AI portfolio should remain proprietary and where openness can expand your addressable market.
3. AI Claws Could Be the Next Step Beyond AI Assistants
The Market Signal: AI agents are evolving from tools that complete individual tasks into a new class called AI claws. AI claws are long-running autonomous agents with persistent memory, system access, and the ability to operate continuously. This could reposition enterprise AI as a means to introduce shared digital coworkers that manage ongoing workflows. Key innovators in the space include Anthropic, Microsoft, Amazon Web Services (AWS), Perplexity, NVIDIA, Intel, and Nous Research.
What to Watch:
- Enterprise adoption of AI agents to expand beyond individual productivity use cases toward shared agents and workflows.
- Hybrid local and cloud execution will create new competition around inference and orchestration.
- Open agent ecosystems will create opportunities for infrastructure and compute providers to build developer adoption.
The Business Decision: Belkhit states that the practical short-term strategy is “to offer open blueprints, runtimes, sandboxes, and optimization support so developers can build claws suited to their own enterprise needs on top of their compute stack.”
4. Industrial Buyers Are Paying for Workflow Improvements Before Autonomy
The Market Signal: Manufacturers are assessing where industrial AI can improve operations without introducing unnecessary risk. Near-term deployments are centered on worker assistance, troubleshooting, coding support, operator guidance, error detection, knowledge capture, and other low-risk applications.
What to Watch:
- AI is transitioning from standalone copilots toward deeper integration within existing workflows.
- Human oversight is important for customers, particularly in environments where reliability and repeatability are critical.
- In Manufacturing Execution Systems (MESs), Agentic AI already generates roughly 3X as much revenue as Gen AI despite its more controlled use.
The Business Decision: Prioritize measurable workflow improvements and operational fit before pursuing ambitious autonomy claims. Manufacturers remain skeptical of trusting AI for total control over the factory floor.
5. Physical AI Is Opening a Much Larger Automation Opportunity
The Market Signal: Robotics Foundation Models (RFMs) enable robots to perform more sophisticated tasks. Traditionally, robotic usage was confined to narrow, pre-programmed workflows. Advancements in Physical AI models are making it possible for robots to manipulate deformable objects. That means robots are more effective at handling objects such as clothing, wires, plastics, and food. ABI Research expects another 30% of existing manufacturing workflows to become automatable, helping the wider RFM-enabled robotics ecosystem become a US$150 billion opportunity by 2036.
What to Watch:
- Manufacturing and warehousing offer the clearest immediate commercial opening.
- Collaborative Robots (cobots) currently represent a lower-friction marketing/sales path than humanoids.
- The opportunity extends beyond robot manufacturers into cloud and edge infrastructure, model training, deployment software, lifecycle management, and specialized silicon.
- China should be treated as both a volume engine and a long-term exclusion threat, necessitating prioritization of Western/allied partners.
The Business Decision: Map your existing capabilities against the broader Physical AI ecosystem rather than assuming robotics hardware vendors will capture most of the opportunity. Principal Analyst George Chowdhury tells us that Physical AI solutions that leverage foundation models require compute in three places: on-device Systems-on-Chip (SoCs), a local server, and a cloud service.
6. Distributed AI Is Creating New Infrastructure Battlegrounds
The Market Signal: Low-latency AI workloads could give Mobile Network Operators (MNOs) a more prominent role in the AI economy. Mission- and safety-critical applications such as robotics, machine vision, and autonomous driving will require inference close to where data are generated. As a result, operators have an opportunity to turn existing network infrastructure into a distributed computing layer.
What to Watch:
- Central offices, network facilities, and other edge-capable locations could support distributed AI infrastructure.
- Mobile Network Operators (MNOs) will participate as connectivity providers, AI landlords, infrastructure providers, or application enablers.
- Waiting for 6G risks allowing hyperscalers, neoclouds, and other infrastructure providers to establish the economics of distributed AI first.
The Business Decision: Identify which infrastructure assets and capabilities can justify premium AI services before the competitive landscape solidifies. Senior Research Director Dimitris Mavrakis explains it plainly: “The telecoms AI race will not be decided solely by who talked about AI first. It will hinge more on who still has the right physical footprint, who can expose useful assets to AI workloads, and who can turn network presence into a commercially relevant computing layer.”
7. Trust Is Becoming Part of the AI Product
The Market Signal: AI is creating a new security problem as agents gain greater access to enterprise data and systems. Rather than inventing an entirely new security stack, vendors are innovating proven technologies—encryption, identity controls, Trusted Execution Environments (TEEs), and Confidential Computing (CC). These solutions are key to governing unpredictable AI behavior and protecting sensitive data.
What to Watch:
- Agentic AI is creating new requirements around identity, access, attestation, and policy enforcement.
- Data sovereignty is more challenging to facilitate when agents can access information across jurisdictions or models process sensitive data elsewhere.
- Full automation will not make sense in many scenarios. High-risk security functions (e.g., certificate issuance and revocation) will continue to require human oversight to avoid false positives and false negatives.
The Business Decision: Look at where your existing security capabilities can solve emerging AI problems before chasing entirely new product categories. As ABI Research Senior Analyst Aisling Dawson explains, “While AI offers its own benefits to support and boost data sovereignty, vendors must also ensure that guardrails inhibit models or agents from breaching the sovereignty they are employed to protect.”
Place Your AI Bets Where Customer Value Is Moving
AI is transforming nearly every industry, underpinning numerous product roadmaps across global technology markets. As a vendor aiming to capitalize, selecting the right investment opportunities is paramount to commercial success. This means identifying where customer demand is rising, understanding how technical capabilities are advancing, and determining where product differentiation can be achieved. The winners will not necessarily be the companies making the largest AI bets, but those strategically placing them in the most value-adding spaces.
Download ABI Research's The Applied AI Playbook: Where Adoption Is Real and What Comes Next for Technology Leaders for deeper market forecasts, technology analysis, and strategic recommendations across the evolving AI landscape.
Ryan Martin
Ben Weaver