AI investment in Ireland in 2026 is led by a handful of American AI labs choosing Dublin as their EU base. OpenAI announced 200 more roles and an applied AI lab on 12 October, taking its Irish commitment to 450 jobs, as The Irish Times reported. Anthropic announced 200 roles in March. The bigger signal is the type of hire. OpenAI's new roles include engineers who work directly with customers on AI applications, which is implementation work that large consultancies and systems integrators have treated as their own. For Irish business, these investments bring talent, demand for advisers, and a widening gap between the firms that use AI well and the firms that don't.

Who has announced what, and when?

The headlines blur four quite different announcements made since October 2025. Two come from AI labs, one from an enterprise software company and one from an Irish-founded technology services firm, and only one of the four was announced in the weeks around the summit. Put side by side, they add up to about 850 announced roles, with very different shapes:

  • OpenAI (12 Oct 2026, 200 roles, 450 committed in total): Applied AI lab at its EU headquarters; customer-facing engineers, privacy engineers, sales and user operations
  • Anthropic (March 2026, 200 roles): Engineering, sales, finance, legal and compliance, operations; office six times larger
  • Workday (15 Oct 2025, 200 roles): AI Centre of Excellence; €175 million over three years
  • Version 1 (25 Mar 2026, 250 roles): Irish-founded; new headquarters and AI Studio at Four Park Place

Each one tells you something different. Workday's announcement is now a year into a three-year, €175 million plan for 200 specialised AI roles, mostly research and engineering for its own products. Anthropic tied its expansion to "strong enterprise demand across Europe", pointing to Europe, Middle East and Africa (EMEA) revenue growth of 11 times year on year and a move to a 21,000 square-foot office. Version 1 is the outlier: founded in Ireland in 1996, with more than 3,700 staff and revenue above €400 million, it is an Enterprise Ireland client, not foreign direct investment. It belongs in the table because it is building the same kind of client-facing AI studio the labs are. Read together, the four announcements show the market moving from building AI to deploying it inside other people's businesses, and Dublin is where much of that European deployment work will be staffed.

What is an applied AI lab, and why should advisers care?

An applied AI lab is the model vendor sending its own engineers to work inside client organisations. OpenAI's Dublin roles include engineers who work with customers on AI applications, privacy engineers who build safeguards and data-protection tools, and sales and user operations staff. Sarah Friar, OpenAI's chief financial officer, told The Irish Times the lab is meant to "help Irish businesses grow" and to support founders turning ideas into companies. The company that makes the model now also offers to help you put it into production. That is the work large consultancies, systems integrators and specialist AI firms have been billing for since generative AI arrived: scoping use cases, integrating models with core systems, building safeguards and getting staff to use what was built. The vendor version arrives with deep product knowledge, early access to new features and an obvious commercial interest in the result.

A model vendor earns more when its models are used more, and usage depends on deployment, and returns are slow to arrive: the William Fry Technology Report 2026 found only 11% of larger Irish businesses reporting a positive return on AI so far. Embedding engineers with customers shortens that path. For advisers, the question is where an independent firm still earns its fee when the vendor is offering engineers directly. The answer is in the parts a vendor engineer can't credibly own. That means neutrality across models and providers, which we covered in how to avoid AI vendor lock-in. It means mapping a system to the right regulator, which Part 1 showed can mean three regulators for one firm. And it means redesigning the process around the tool, plus the change management that decides whether staff use it. A vendor's engineer will make its model work. They are not paid to tell you a competitor's model would do the job for half the cost.

What work do these setups create for Irish firms and advisers?

Each expansion generates a familiar stack of professional work: tax and Pillar Two structuring, which Part 2 covered in what Budget 2027 changed for AI investment, plus employment law and people advisory for hundreds of hires, property for new headquarters such as OpenAI's move to the Tropical Fruit Warehouse in December, and data-protection work for every new privacy engineering team. The wider supply chain is large but uneven. The Department of Enterprise's Annual Business Survey of Economic Impact (ABSEI) for 2024 found that foreign-owned client companies spent €43.3 billion in the Irish economy, €12.2 billion of it on services bought in Ireland. Yet that spend equalled only 9% of their sales, and 5% in information and communications, against 55% for Irish-owned firms. Technology multinationals buy remarkably little locally, so Irish suppliers have to compete hard for that 5%.

Physical infrastructure sets limits on what else arrives. Under the Commission for Regulation of Utilities' policy for large energy users, new data centres above 10 megavolt-amperes (MVA) must supply their own new generation or storage, connected to the electricity market, and meet 80% of their annual demand from additional Irish renewables within six years of switching on, as William Fry summarised in December 2025. That makes large new compute projects slower and more expensive to land in Ireland. The 2026 announcements are about people, not power: engineers, sales teams, legal and compliance staff. For Irish firms, the most direct effect is competition for the same talent. Every engineer who joins an applied AI lab to deploy models for clients is one fewer available to an Irish insurer, bank or manufacturer trying to build the same capability in-house.

Is it a jobs boom or a reshuffle?

The jobs data doesn't support the word boom. The Central Statistics Office (CSO) Labour Force Survey for the first quarter of 2026 found that employment in information and communication fell by 20,300 people, or 10.7%, year on year. That was the largest decrease of any sector, driven mainly by computer programming and consultancy. That quarter was also the first published under a new industry classification, which makes some comparisons less clean, but the size of the fall is hard to explain by reclassification alone. The roughly 850 roles in the table above arrive in a sector that has been shedding jobs faster than it adds them. The roles being created are also different in kind from the roles being lost, which matters for anyone planning a workforce: AI deployment, customer engineering and compliance are rising, while general software development and services are shrinking.

The longer-term exposure is wider than tech. A joint report by the Economic and Social Research Institute (ESRI) and the Department of Finance estimated that around 7% of current jobs could be displaced by AI in the short to medium run under its central scenario. For a large Irish employer, that changes what the AI investment wave means in practice. The announcements bring a small number of highly paid, specialised roles into Dublin. The broader effect lands inside every organisation adopting AI, as tasks shift and some roles shrink. The firms that plan for that shift, with redeployment paths and training funded partly by the €150 million skills package in Budget 2027, will keep the knowledge their people hold. The firms that treat AI purely as a headcount saving will lose that knowledge, and may discover later that it was what made the process work.

Why do the gains stay inside the multinationals?

Because they adopt AI faster and start from a higher base. The CSO's productivity release for 2024–2025 found AI adoption among foreign-owned enterprises in Ireland rose to 43.6% in 2025, against 16.3% for domestically owned firms. The same release puts labour productivity at €507.30 per hour in sectors dominated by foreign-owned companies, against €65.80 per hour in the domestic sector. That comparison is between sectors, not individual firms, and the top figure reflects multinational profit accounting as well as output per worker. Even allowing for that, the gap is wide. The firms most likely to use AI are already the most productive, and the AI labs arriving in Dublin are selling first to firms like them. Without a deliberate effort on the Irish side, these investments widen the adoption gap rather than close it.

The gap closes through capability, and capability can be bought, borrowed or built. Irish firms can buy it from the vendors' own applied AI teams and accept the lock-in risk that comes with them. They can borrow it from advisers who work across models and know the Irish regulatory map. Or they can build it in-house, which is slowest but lasts longest. Government is paying for part of the third route. The summit's renamed Grow Digital & AI and Digital & AI for Business schemes cover firms with up to 50 staff, and the €150 million training package covers workforce skills more broadly. For larger Irish firms, the most practical lever may be the supply chain itself. That €12.2 billion of services spending goes to suppliers who can meet multinational standards, and a supplier that can show AI-enabled delivery has a stronger case for winning a share of it.

Where this leaves Irish business and its advisers

From the operator side, the gap is deployment capability, not access to models. Every Irish firm can buy the same models OpenAI and Anthropic sell to the multinationals. What separates them is the people and process work to get those models into daily operations: choosing the right use cases, integrating with systems that were never designed for AI, setting up the controls a regulator will ask about, and bringing staff along. We've spent our careers on the commercial side of regulated industries, where our own operating background taught us that the product rarely decides the outcome on its own and the rollout usually does. That is the real question behind the 16.3% and 43.6% adoption figures. The labs arriving in Dublin are betting they can close that gap for clients themselves, and they will close it first for the firms already ahead.

This series started with who regulates AI in Ireland, moved to how the tax system treats it, and ends with who is building the capability to deploy it. The three answers point the same way: the regulators expect firms to know their own AI estate, the tax code leaves adoption costs to operating budgets, and the vendors now sell deployment directly. If you're working through what that means for your organisation, start with who enforces the EU AI Act in Ireland, then use how to choose an AI model to keep your options open. Each article stands on its own, but together they make one argument: Irish firms will be judged on what they can deploy and show evidence for. If you want an independent view on where AI fits in your operations and who should build it, book a 30-minute working session.