The Factory No One Is Talking About
On November 11, Armada and Johnson Controls announced that Galleon Forge Two — the second manufacturing facility producing Leviathan sovereign AI infrastructure — has entered continuous production in Dublin. A European workforce. European supply chains. European customers. The target: at least 500 megawatts of capacity in 2027, including the first Leviathan destined for Fossefall in Norway.
This is the single most consequential industrial announcement in this day's news cycle, and almost no one in the financial press has flagged it properly. Not because it's small. Because it reframes the entire AI debate from a software-and-chips conversation into something messier, more expensive, and far more geopolitical: the question of who manufactures the physical scaffolding of the intelligence economy, and under whose jurisdiction it operates.
For three years, the AI narrative has been dominated by model development, training data, and chip supply. That was the first act. This is the second act, and it belongs to industrial policy, sovereign capital, and cold hard factory logistics. The winners won't be the ones with the best loss curves. They'll be the ones who can produce, cool, power, and govern compute at national scale.
Sovereignty Is No Longer a Buzzword — It's a Manufacturing Spec
Armada's positioning is deliberate and pointed: "a repeatable manufacturing model that any nation or region can anchor within its own borders." Read that again. This isn't a product pitch. It's a template for national industrial self-sufficiency in the AI era. The implicit argument is that every G7-adjacent economy will need domestic production capacity for AI infrastructure, the same way every industrial nation built domestic steel capacity in the nineteenth century.
The choice of Ireland as the manufacturing anchor is not accidental. Ireland offers EU membership, an English-speaking workforce with deep pharmaceutical and semiconductor manufacturing experience (Intel, Analog Devices, and others have trained generations of engineers in the Leixlip-Cork corridor), favorable corporate tax treatment, and — critically — a neutral geopolitical position relative to US-China tensions. It is the Switzerland of compute manufacturing.
My forward call: within 18 months, we will see at least two more sovereign AI manufacturing facilities break ground outside the United States. South Korea and Japan are the obvious candidates given their existing electronics manufacturing base and acute sensitivity to supply chain dependency. The UAE, which is simultaneously signing small-scale LNG MOUs (as GasHub did with Abu Dhabi's Department of Energy this week) to power data centers, will attempt something similar but will likely fail to attract the manufacturing component and settle for being a hosting jurisdiction instead.
Agentic AI Just Became a Governance Emergency
While the hardware layer gets its moment, Cequence Security published a report — developed with input from KPMG in Singapore — showing that AI agents are already signing into systems, booking transactions, and making purchases on behalf of users, often without identifying themselves as non-human actors. The report's title is clinical: "Agentic AI Runtime Security, Governance, and Control." The contents are anything but.
Here is what most analysts miss about this data point: it is not a future risk. It is a present operational reality. The agents are already running. They are already transacting. And the compliance frameworks, audit trails, and liability structures that assume a human is on the other end of every interaction are now catastrophically outdated.
This is the 1996 moment for agentic AI — the year the EU Data Protection Directive was passed while the internet was still in its infancy, creating a regulatory framework that took two decades to catch up. The difference is that this time, the agents are operating at machine speed across global financial systems. The Singapore-based framing of the report is telling: KPMG Singapore, operating in a jurisdiction that hosts the ASEAN financial backbone and is governed by the Monetary Authority of Singapore's notoriously rigorous technology risk guidelines, is sounding the alarm. When Singapore's Big Four arm publishes a report saying "agents are transacting without identification," that is not marketing collateral. That is a warning shot.
I expect a major incident within the next 12 months — most likely in financial services, most likely involving a rogue agent executing unauthorized trades or triggering automated procurement decisions at scale. The regulatory response will be disproportionate and geographically fragmented, with the EU moving fastest (the AI Act's provisions on autonomous systems are already drafted but unenforced), the US fragmenting between SEC, FTC, and state-level approaches, and Asia scrambling for bilateral coordination mechanisms that don't yet exist.
Turkey's Energy Gambit: Decarbonization as Competitive Moat
At COP31 in what appears to be a Turkish-hosted summit, Koç Holding Vice-Chair Ali Y. Koç stated plainly: "The Energy Transition Has Become the Cornerstone of Competitiveness in Industry." Al Gore delivered the keynote. IEA Executive Director Fatih Birol moderated. The framing was not environmental. It was industrial.
This matters because it signals a fundamental shift in how middle-power industrial nations — Turkey, South Korea, Malaysia, potentially Indonesia — view the energy transition. They have stopped treating decarbonization as a Western-imposed tax on development and started treating it as a first-mover advantage in the next industrial cycle. A Turkish manufacturer that runs on renewable-powered infrastructure can undercut a European competitor still paying legacy grid rates. A Korean semiconductor fab that co-locates with dedicated nuclear capacity gains a structural cost advantage over American competitors dependent on aging grid infrastructure.
The irony is thick. The United States, which spent the last decade debating whether to rejoin the Paris Agreement, is now losing the energy-competitiveness race to nations it lectured on climate policy. Turkey — a country whose industrial base runs on imported natural gas — is positioning itself as a green manufacturing hub. That is either visionary or delusional, and we will know which within twenty-four months.
Australia's Quiet Infrastructure Vindication
Western Sydney International Airport opened to passengers on October 26 — seven months ahead of schedule — following a Bechtel-led delivery program. Australia's first major greenfield international airport in over fifty years arrived early.
In a country where infrastructure projects are synonymous with budget overruns, political theater, and interminable delays (see: Sydney Light Rail, Melbourne's various tunnel projects), this is a genuine anomaly. It suggests that the combination of clear ownership, experienced delivery partners, and political insulation from election cycles produces results. Western Sydney Airport Corporation maintained its design through multiple federal government changes. Bechtel executed without the scope creep that killed comparable projects elsewhere.
The economic implications for Western Sydney — already Australia's fastest-growing population center — are immediate. But the broader signal is that the infrastructure constraint on AI-adjacent economic activity (data centers need power, cooling, and logistics corridors) is not as binding as the pessimists assume, provided governance structures are correct.
The Skills Gap Is the Quiet Crisis Nobody Can Budget Around
The AmChamSG Manpower Survey released in Singapore found that 46% of employers now cite skills mismatch as a key manpower challenge, up sharply from 34% just one year ago. Eighty percent of employers are actively upskilling staff for AI-related roles.
An eleven-point year-over-year deterioration in perceived skills mismatch is not a gradual drift. It is a cliff. It tells you that the gap between what the technology demands and what the workforce provides has accelerated faster than any training pipeline can address. Singapore, with its world-class SkillsFuture program and tripartite labor framework, is the best-positioned economy in ASEAN to absorb this shock. If Singapore is seeing a cliff, the picture in Malaysia, Indonesia, the Philippines, and Vietnam is considerably worse.
This is the blind spot in every AI investment thesis that assumes labor markets will adapt efficiently. They won't. They haven't yet. And the political pressure that builds when 46% of employers are saying "our people can't do what we need" while workers are told they need to "upskill" will be severe.
The Bottom Line
The AI economy's second act is not about models, chips, or even data. It is about who can manufacture the infrastructure, power it, govern its autonomous agents, and retrain its workforce — all simultaneously, and at speed. The nations getting this right are not Silicon Valley or Shenzhen. They are Dublin, Ankara, Seoul, and Kuala Lumpur, building the unglamorous physical and institutional plumbing that makes intelligence at scale actually deployable. The companies that understand this — Armada, Johnson Controls, the sovereign wealth funds behind them — are playing a different game than the one the market is pricing. The gap between those two games is where the next decade's alpha lives.