The Next Year in Manufacturing | 2026

Breaking the Bottleneck Newsletter

---

For AI in Manufacturing

I think we’re going to see a big uptick in applied AI in manufacturing. This will be driven by domain-specific models. Manufacturers have been asking operators to use off-the-shelf, enterprise AI that’s not built for the shop floor and are seeing limited success. In 2026, manufacturers will move to purpose-built models. And not just LLMs, but also other predictive models that are injected into existing workflows and bring the concepts of “agents” to life for manufacturing companies.

– Devin Bhushan, CEO of Squint

This year the wave of AI washing will continue, with countless undifferentiated companies slapping a chatbot on anything and everything. Despite that, industrial AI will be delivering real, measurable value in two main areas:

1) applied machine learning for optimizing quality, energy, and performance 2) leveraging Generative AI to more rapidly configure OT systems.

— Rick Bullotta, Advisor and Founder of Thingworx (Acquired by PTC)

AI will play a massive role in unleashing factory productivity in a few key ways: 1) production data collection through sensors and computer vision, fed to AI decision making tools & agents which can automate production decisions 2) AI-powered robots which are more intelligent, flexible, dexterous and easily re-purposable than traditional robots, accomplishing more work per unit of robot 3) AI tooling to modernize production changes (using AI to automate site evaluation, robot selection, simulation & validation, maintenance etc.)

– Saman Farid, CEO and Founder of Formic

As 2026 unfolds, agentic workflows will dominate the AI conversation, driven by their potential to automate manual work and administrative drag across manufacturing. At the same time, a harder truth will emerge: most companies are still stuck at the prototype stage, and the real value only materializes when AI can be deployed reliably at scale with precise, real-time operational context for operators, technicians, and engineers. This will expose a growing tension between technology leaders eager to move to the next wave of AI and operations teams focused on ROI, trust, and whether today’s solutions are truly ready for the factory floor.

— Howard Heppelmann, CEO, OpsMate AI

The rhetoric around China’s engineering human capital advantage over the US will grow louder, and large companies will turn to AI to fill the gap. Similar to healthcare 18 months ago, many companies will rush in to fill the gap with big promises and shallow products. The winners will be companies that build deep tool integrations, have an unfair data moat, and can get executive buy-in at their target customers.

— Nick Boesel, Product Lead at P-1 AI

Quick wins and use cases from Agentic Workflows. This is no longer the time to talk about chat bots, but how AI is moving into removing simple tasks that are repetitive and low value add to allow those that are doing this now to drive value add and game changing insights.

— Doug Bellin, Industry 4.0 Head, Amazon Web Services

Many of the new AI capabilities that we developed further in 2025 will come to fruition in 2026 and will make industrial and manufacturing operations far less dependent on large external engineering teams, and shift value from bespoke platforms to flexible, multimodal, and agentic systems that plants can adapt themselves. Multimodal models will automatically ingest diagrams, time series, and technical documentation to build usable knowledge graphs in hours rather than months, at a tiny fraction of current platform costs. Agentic, multi-agent orchestrators will enable plant-level retrieval‑augmented workflows that operators can reconfigure with natural language.

— Moe Tanabian, CEO of Intuigence AI

In 2026, industrial software vendors will double-down on their (you guessed it) “AI strategies and offerings.” What remains to be seen is where real value will be created for manufacturers beyond the hype. Specifically, WHERE in the tech stack will AI be leveraged to access all data sources and will vendors up, down and across the stack play nicely or try to create their own, proprietary solutions?”

— Brad Hafer, Advisor and Board Member, Corp Dev at Plex

The AI bubble will burst, but that’s good news. What remains will be the applications that actually matter—built by people who understand manufacturing deeply enough to close the loop between prediction and action, not just generate impressive dashboards or be another wrapper around one of the LLM providers.

— David Ariens, IT OT Insider & Board Member at Timeseer.AI

For decades, we’ve optimized for marginal gains through rigid, rule-based automation. In 2026, we are breaking through that ceiling by pivoting to agentic AI, moving from systems that merely flag problems to autonomous agents that reason, plan, and execute solutions.

This shift isn’t about reducing headcount, but about amplifying human potential, transforming operators into ‘super-users’ who are empowered by AI to solve more complex problems and drive higher value.

— Praveen Rao, Industry Head (Manufacturing) at Google

---

For Manufacturing Data

For manufacturers looking to leverage AI for production projects, it’s vital to prioritize an industrial data strategy focused on making data accessible while standardizing and normalizing it for AI usage. But in reality, many organizations rely on brittle, non-scalable data architectures that increase the risk of AI hallucinations, leading users to distrust the information from AI agents. As Industrial AI adoption accelerates, manufacturers will need to formalize and improve how they integrate, curate, and deliver reliable, contextualized data through an Industrial DataOps platform. A strong industrial data strategy, powered by a DataOps architecture, will be the key differentiator for manufacturers in generating real business value from their AI tools.

— John Harrington, CPO of HighByte

Graph databases have been “the next big thing” for decades, Neo4j was founded in 2007, for example, but 2026 will be the year they land with impact in manufacturing. Context graphs that record plant decision histories (or traces), asset model graphs that replace fixed and brittle hierarchies, and graphs that assemble data sources for multi-source queries will become standard features in vendor offerings.

– Michael Risse, Advisor and Former Chief Strategy Officer at Seeq

In 2026, the shift in manufacturing won’t be about chasing the next technology wave—it will be about designing for resilience. That starts with fixing how data flows across IT, OT, and the supply chain, and putting a governance layer in place, so everyone is operating on the same operational truth. Once that foundation exists, AI becomes genuinely useful—not as a replacement for people, but as an assistant that helps engineers, operators, and planners make better decisions inside real constraints.

— Indranil Sircar, Global CTO of Manufacturing & Mobility, Microsoft

Deeper dive with integration of data from on-premise standalone systems, that are statements of record, to integrated information stores that allow new insights and actions.

— Doug Bellin, Industry 4.0 Head, Amazon Web Services

2026 is going to be the year when businesses will drive to bring manufacturing closer to consumption, and for this, one needs to automate more – simply because you don’t find enough labor, and furthermore, the playground with General purpose technologies( GPT) like AI & Cloud necessitates “ hyper automation” in Industrial manufacturing to disrupt. The production of the future won’t just be bigger or faster. It will be closer to demand, more adaptive, and even more autonomous.

— Rajiv Sivaraman, VP of Strategic Partnerships at Siemens

As industrial AI moves from experimentation to production, data quality has emerged as the primary constraint to scale. AI applications depend on reliable, well-understood data to operate safely and deliver value. Continuous validation of sensor and process data, early detection of anomalies, and transparency into data behavior are essential to reduce risk, accelerate deployment, and ensure AI systems perform as intended in real-world operations.

— Jane Arnold, COO at Aperio

Playtime’s over. After years of cloud and edge experiments, 2026 will force organizations to face the unsexy reality: managing large-scale industrial data platforms requires proper data governance, system management, and operational discipline—or they’ll collapse under their own complexity.

— David Ariens, IT OT Insider & Board Member at Timeseer.AI

In 2026, I’m most excited about seeing AI in manufacturing shift from optimization to orchestration. Flexible automation, real-time quality learning, and AI-driven design-for-manufacture will start to matter far more than standalone robots or better models. The real breakthroughs will come from factory systems that can learn, adapt, and compound experience on the production line.

— Glen Turley, Head of Quality Intelligence, Stealth Next-Gen Contract Manufacturing

In 2026, hyper-personalization moves from the consumer world to the factory floor, where AI agents with ‘long-term memory’ tailor insights to the specific context of every operator. Instead of generic alerts, workers receive personalized directives—like proactive machine adjustments based on local humidity or shift-specific productivity coaching. This creates a more intuitive, efficient workplace where the right data finds the right person at the exact moment of need.

— Praveen Rao, Industry Head (Manufacturing) at Google

---

For Factory Workflows and Orchestration

US manufacturers will get more automated, connected, and competitive in 2026. AI will automate more paperwork so manufacturers have more time to actually make parts, faster. This includes front and back office workflows like quoting and job tracking. Digital platforms will bring our hundreds of thousands of small manufacturers “online”, helping buyers discover those that fit their capability and capacity needs as well as reduce time and friction in working with them.

— Alex Huckstepp, Co-Founder, CCO, Uptool

The long-held dream of ‘made-to-order’ manufacturing finally becomes a reality in 2026 as agentic AI eliminates the massive capital drain of standing inventory. We are entering the era of the ‘self-healing’ supply chains, where AI agents autonomously detect disruptions and reroute logistics in real-time. By bridging the gap between the design studio and the shop floor, we turn supply chain volatility into a source of institutional strength and capital fluidity.

— Praveen Rao, Industry Head (Manufacturing) at Google

In 2026, factory orchestration will emerge as a distinct category, unifying automation, process control, and real-time observability into a single execution layer at the workcenter - finally closing the gap between plans and reality on the shop floor. Manufacturers will shift away from fragmented point solutions toward systems that eliminate wasted motion, enforce correct execution, and provide a live, shared view of how work actually runs.

— David Caputo, Co-Founder of Harmoni

Over the next 12 months, I expect OEMs in aerospace, defense, EVs, and other critical sectors to increasingly adopt Factory-as-a-Service (FaaS) models as a compelling alternative to traditional contract manufacturing. This shift is driven by persistent reshoring pressures, supply chain vulnerabilities, and the demand for faster, more secure scaling without heavy capital expenditure.

This early exploration—primarily through pilots of dedicated “private cloud” production capacity managed by specialized operators—will gain traction as initial adopters showcase shorter lead times, improved predictability, and enhanced supply security.

Over the next decade, we will witness a major shift where the fragmented and transactional contractor model yields to FaaS as Europe’s prevailing paradigm. OEMs will transition from sourcing individual parts to owning dedicated production capacity operated by expert providers, enabling seamless scaling from prototypes to high volumes while reducing overhead and boosting innovation.

This evolution will be supercharged by plummeting software development costs and accelerating build speeds, allowing manufacturing businesses to create bespoke operating systems tailored to their unique workflows. As coding becomes cheaper and faster, the justification for generic SaaS subscriptions like traditional ERPs will weaken. Companies will favor unique, in-house solutions that deliver superior agility and control, ultimately fostering a more resilient and competitive manufacturing ecosystem.

— Tom Smith, Co-Founder of Stealth Startup

We will see plant floors increasingly ask for AI that works, solves real operational problems, and scales. We will also see AI insights delivered through an Operational 3D twin layer resonate more with plant floors, as they will help reduce cognitive load for real-time decision-making and set the foundation for further use of 3D twins in Manufacturing. Sight Machine, in partnership with Microsoft and NVIDIA, is well-positioned to deliver on these needs. –

— Sudhir Arni, SVP at Sight Machine

The number one trend I'm seeing in manufacturing is the loss of experts on the plant floor. For US Manufacturing to achieve its goals of onshoring and rebuilding its manufacturing base, it will need to invest in equipment modernization, increased automation, and AI adoption. For those companies who have been making incremental improvements to replace old equipment, who have invested in plant floor data infrastructure, and have a mindset of continuous improvement—you will be rewarded with an easier journey. For companies that have not done these things, the road ahead will be bumpy.

— Bryan Debois, Director of AI at Rovisys

---

For Physical AI

Humanoids represent a long-term bet on general-purpose robotics, not a near-term manufacturing breakthrough. In 2026, their influence will be cultural and technical (shaping roadmaps and talent), and funding spent on humanoids will develop a class of better actuators & sensors (training the supply chain). More task-specific robots however will continue to do the heavy lifting, high speed work & practical production activities.

– Saman Farid, CEO and Founder of Formic

The humanoid robot craze will fall woefully short in functionality, but we’ll finally see investment in end effectors/hands and new types of tactile AI models that will actually move things forward.

— Rick Bullotta, Advisor and Founder of Thingworx (Acquired by PTC)

Robotics and AI are at an inflection point. Computer vision’s impact is finally percolating across the industry—something that wasn’t true even a few years ago. With approved budgets now in place, particularly for action recognition in process quality monitoring, vendors with solid products should see accelerated pilot-to-deployment timelines and revenue growth.

The humanoid story is wildly different. There’s a stark divide in expectations between those who’ve rarely set foot in a plant and those who have—and the billions already spent have only amplified the hype. The reality: bits moving bits (LLMs) is one thing. Bits moving atoms in a controlled environment (today’s humanoids) is another. Bits moving atoms in the messy, unconstrained real world (tomorrow’s humanoids) is something else entirely. Despite the pronouncements, I suspect we’re at least a decade away from that last capability—much like fully self-driving cars.

Then there’s manipulation, the final frontier. As a friend aptly put it: ‘Trillions of dollars flow through our hands.’ Whether it’s a factory robot, a cobot, or a humanoid, they all need hands that work. I’m excited by pioneering work applying reinforcement learning to grasping, and paired with better electromechanical design, real progress is being made. But we still have a long way to go.

– Prasad Akella, CEO at Stealth & Multi-Time Manufacturing Founder

Early days in the PhysicalAI space. While I see this to be very early days, companies have to get ready for this with first unlocking the data sets mentioned before then the initial usage such as with Digital Twins to understand where robotics will quickly add value. There is a long way to go to be able to replace human beings with the current technology but it is moving quickly so be ready!

— Doug Bellin, Industry 4.0 Head, Amazon Web Services

World models will become the new focus for those working on digital twins. Don’t expect a production-level solution until 2027, but we will see significant investment this year.

— Mark Burhop, CTO of Sciath aiM

---

For Product Development and Simulation

By 2026, manufacturing will shift from static planning artifacts to living, continuously updated production orchestration that change in real time as engineering, supply, and demand evolve.

As product complexity and change velocity rise, companies that still rely on manually managed data will see mounting scrap, delays, and quality escapes, while context-aware, computable production systems will stabilize ramps and compress lead times.

The competitive advantage will no longer come from better CAD or ERP, but from owning the layer that keeps the factory’s build definition continuously correct under change.

– Filip Aronshtein, CEO of Dirac

The success of software development tools like Speckit, OpenSpec, and others will result in similar tools for physical product development in 2026. The processes we use to develop new products and the documentation that comes from it will be created, co-created, checked, and verified by AI. This will result in higher quality, better adherence to process requirements, and a plan that is constructed in half the time with half the people.

— Mark Burhop, CTO of Sciath aiM

2026 will be the first year that sees agents running engineering workflows for the development, delivery, and servicing of production-grade machines. Enterprise appetite has converged with technological acumen such that the demand for constantly executing and autonomous work will explode and eventually become industry standard.

– Kiegan Lenihan, CEO of xNilio

I see Jensen trying to replicate the CUDA playbook in robotics with Omniverse. But here’s the challenge: game simulators get away with nearly correct physics—they don’t need perfection. The factory floor demands exactly that: perfection. It’s unclear how much end-to-end models can circumvent the laws of robotics to generate truly realistic simulations. The gap matters.

– Prasad Akella, CEO at Stealth & Multi-Time Manufacturing Founder

In 2026, agentic systems will finally be accurate enough to allow reliable deployment into real-world hardware design-to-production and post-production manufacturing systems

– Rui Aguiar, CEO of Cosmon

By 2026, physics won’t be something you “run.” It will be something you query. Zero-shot, deterministic foundation models will reason over new designs continuously, at production scale. Tools that require setup, tuning, or human-mediated workflows won’t survive contact with real automation.

– Hardik Kabaria, CEO of Vinci4D

This year, we will stop worrying about models and focus on training data. Expect specialized vertical training sets to create the first true generative AI mini-apps, while the more general CAD models and the large LLMs like Claude and Gemini remain interesting tech demos. My friends at Mecado, Hanomi, and nTop are leading this space.

– Blake Courter, Founder of Gradient Control Laboratories

AI is no longer a slide-deck concept; it’s finally capable of delivering real value on the factory floor. But manufacturing has a reckoning ahead: decades of legacy infrastructure are now a liability, not an inconvenience. Without upgrading the foundation, the AI layer simply doesn’t matter. The old barriers to entry are collapsing. Companies like Infinitform, Encube, and a growing wave of startups are offering capabilities once locked behind expensive, slow-moving engineering giants at a fraction of the cost and at startup speed. The incumbents won’t out-innovate this shift; they’ll acquire it.

– Milan Kocic, Head of Innovation and Partnerships at SmartSkin, Former Hexagon Sixth Sense Lead

---

For OT Cybersecurity

AI is not blocked by a lack of data. It is blocked by change risk. In 2026, the winners will run their factories the way great software teams run production: every change versioned and reviewed, deployments reproducible and tested before they touch the line, blast radius kept small, and recovery measured in minutes. When change is safe and reversible, AI stops being a risky experiment and becomes a repeatable rollout.

— Josh Ross, CEO of IronLoop

The exponentially larger Operational Technology (OT) attack surface, driven by the need to digitally transform and rapidly optimize business operations, is making autonomous, AI-driven defense non-negotiable table stakes in 2026.

Making industrial cyber-physical networks more resilient requires a multifaceted approach to securing IT and OT systems across hybrid on-premises and multicloud networks; companies must invest in new security paradigms leveraging advanced prompt engineering and AI guardrails. Crucially, AI agents are deployed as autonomous or semi-autonomous software entities that play a pivotal role in securing this expansive cyber-physical environment by actively defending data across its three states—at rest, in transit, and in use—by continuously auditing configurations, monitoring industrial protocols for anomalies, and enforcing runtime AI guardrails to ensure the integrity, security, safety, and resilience of the underlying industrial operations.

— Praveen Rao, Industry Head (Manufacturing) at Google

On Macro

As geopolitical tensions continue to rise between the US and Europe, China, Canada, and Mexico, the US will lean deeper into localization, requiring an increasing amount of American content to qualify goods as “Made in the US”. Infrastructure investments will reach all-time highs, spurred by desire to access hard-to-reach natural resources; not just in places like Venezuela and Greenland, but also in the US. Companies building this type of infrastructure will receive heavy government support by way of direct investments and exclusivity agreements. AI will play a supporting role, but infrastructure will be king.

— Will Drewery, CEO at Diagon

U.S. manufacturing accounted for nearly 19% of total industrial real estate demand in 2025, the highest share in more than a decade. In 2026, we expect demand to remain elevated as AI infrastructure, energy, defense, critical minerals, and pharmaceutical manufacturing scale rapidly. The key challenge won’t be demand, it will be the availability of power, entitled land, and specialized facilities to keep pace with this wave of industrial growth.

– Greg Matter (Vice Chairman) and David Sesi (MD) of Advanced Manufacturing at JLL

2026 is the year mid-caps begin to disrupt their industry. This GenAI tech cycle uniquely favors smaller firms over larger, more complex incumbents. Smaller firms are faster to redesign their processes and embed AI tools in a new business-as-usual. If they can stay ahead long enough, they will reshape the market.

— Daan Kakebeeke, Associate Partner Industry 4.0 at Bain

Fundamentally, most American factories are bottlenecks by available “hands” to do work, demographic trends are increasing this pressure. Robots seem to be the only likely solution to enable American factors to run more hours, increase throughput & capacity.

– Saman Farid, CEO and Founder of Formic

I expect to see a good bit of M&A in the industrial space this year, but unfortunately many of these deals will be fire sales of startups who overcapitalized and under delivered.

— Rick Bullotta, Advisor and Founder of Thingworx (Acquired by PTC)

I’m optimistic that the M&A momentum we saw building in late 2025 will accelerate through 2026, especially for companies at the intersection of software and physical AI, such as robotics, industrial automation, and smart manufacturing. Many large strategics have fallen behind on forming and executing on an AI strategy and are increasingly looking to M&A as a path to catch up. The companies that have survived the last few years with strong unit economics, defensible IP, and proven AI deployments are now well-positioned as these buyers deploy capital to close their capability gaps.

— Josh Ewing, Founding Partner, Onward Advisors

At least one GPU-enabled surrogate modeling company is acquired by an incumbent.

— Nick Boesel, Product Lead at P-1 AI

For the past 15 years big data has been the consistent theme anchoring the conversation with the same “data is the new oil” slide. This will shift in 2026 where it’s no long erdata but data centers as the new oil. More startups focused on data center optimization focused on efficient operations, improved pricing, and streamlined consumption will be funded than any other vertical or topic.

– Michael Risse, Advisor and Former Chief Strategy Officer at Seeq

The US and the Trump administration want to reshore manufacturing. Back in grad school, I could bike to several machine shops right here in Silicon Valley. Seriously talented machinists not only built what I dreamed up—they improved my designs. Today, I need to drive to select pockets of the Valley to find that same talent.

If this is any indicator, do we really have the talent pool to stand up massive plants and rebuild manufacturing? Revitalizing it will take decades: growing the skills, rebuilding supply networks, making the economics work. Does America have the stomach for that? Or will this be another Foxconn-Wisconsin?

– Prasad Akella, CEO at Stealth & Multi-Time Manufacturing Founder

Expect an acceleration of capital into industrial tech, driven less by organic transformation and more by consolidation. Big players will buy what small, nimble teams can build faster than corporate roadmaps ever could. In manufacturing, data and automation are oxygen, and physical AI will soon dictate everything from design to production to supply chains, assuming the infrastructure doesn’t choke it first.

Politics may try to slow this down, but technology historically ignores borders and narratives. If the U.S. remains distracted by internal battles and invented enemies, China won’t wait. It will quietly extend its lead by building, deploying, and scaling while others debate.

– Milan Kocic, Head of Innovation and Partnerships at SmartSkin, Former Hexagon Sixth Sense Lead

---

For the Workforce

The factory worker of the future will no longer be measured by analog tools of the past but by their mastery of the latest digital stack, giving way to the rise of the ‘technocrat’ workforce. These workers will serve as conductors of a digital orchestra, prompting models to synthesize complex maintenance data into instant fixes while supervising autonomous workflows for machine calibration and energy optimization.

This transition elevates the workforce into high-value roles focused on governance, risk, and control, where humans audit AI logic to ensure all autonomous actions remain within strict financial and safety guardrails.”

— Praveen Rao, Industry Head (Manufacturing) at Google