By Robbin Laird
The history of warfare is inseparable from the history of weather. At Normandy in June 1944, Dwight Eisenhower held four stars and command of the largest amphibious operation in history and deferred to a meteorologist. The 24-hour delay that enabled D-Day was not the product of a committee, a requirements process, or a procurement cycle. It was the product of timely, accurate environmental intelligence delivered by someone who understood the data.
That logic has not changed. What has changed is the technology available to deliver that intelligence and the institutional arrangements through which militaries and governments access it.
In a recent conversation with Teppo Kuisma, VP for Sales and Marketing at Vaisala Xweather, the implications for defense and security planning came into sharp focus.
From Instrument Maker to Intelligence Provider
Vaisala, the Finnish technology company behind Xweather, is celebrating its 90th year. Founded when its inventor repurposed an abandoned radiosonde and concluded he could build a better one, Vaisala grew quickly into the backbone of global meteorological infrastructure. Its first client was MIT. For decades it partnered with virtually every national meteorological organization in the world and, as Kuisma notes, those organizations themselves trace their origins to military necessity.
Xweather was established within Vaisala in 2022 as a distinct business area, separate from its weather instruments and industrial measurement segments. It currently serves approximately 2,500 subscription clients, 75 percent of them in the United States, with the remainder distributed across Europe and the broader global market. Its client base spans energy, road administration, insurance, and government, with the latter accounting for roughly 20 percent of its business. Among its government clients: the U.S. Navy, the U.S. Air Force, the FAA, and NOAA.
The AI Inflection in Weather Forecasting
The core of what Vaisala Xweather offers is not simply better weather data. It is a fundamentally different approach to generating forecasts. Traditional numerical weather prediction (NWP) works by modeling the atmosphere as a physical system on a grid.
The problem, as Kuisma explains, is structural: “You needed to recalculate the whole world to add observations. Every major model run — European Centre for Medium-Range Weather Forecasts (ECMWF), Global Forecast System (GFS is a U.S. system) — is six hours old when it’s published. It’s practically based on observations made 12 to 24 hours ago. So even when the latest model comes out, you’re about a day and a half late.”
Artificial intelligence and machine learning change this fundamentally. Rather than recalculating a physical model of the entire atmosphere to incorporate new observations, AI-driven systems can derive forecasts directly from current sensor inputs in near real time. “If the conditions are now this,” Kuisma describes, “what I’m seeing that means the next conditions in two hours are something like that.” The lag embedded in NWP is replaced by continuous, observation-driven inference.
Kuisma is candid about where this innovation is originating: “It’s not coming from governmental players. It’s coming from tech leaders and private companies, because that’s where the competence for running this new forecast technology resides.” Vaisala Xweather, with its roots in sensor manufacturing and observation networks, is positioned precisely at the intersection of those two capabilities, the physical observation infrastructure and the computational intelligence to make use of it in real time.
What distinguishes Xweather further is its hybrid model: in roughly 40 percent of its subscriptions, the company delivers not only the digital forecast but also the weather station that feeds local observations back into the system. Two proprietary global networks, a lightning detection network that identifies every storm forming worldwide, and an expanding atmospheric observation network, provide data inputs that no national meteorological service can replicate at comparable speed or coverage.
The Arctic Problem
Among the clearest strategic implications of Xweather’s approach is its relevance to the Arctic, a domain that is simultaneously opening to commercial and military activity and among the most poorly served by existing meteorological infrastructure. As Kuisma notes, the Arctic currently has the worst weather forecasting accuracy in the world because it has the fewest observations and because no one has invested in optimizing a forecasting model for the region. Storm prediction probabilities in the Arctic today hover around 60 percent, barely better than chance.
That is not an abstraction. Cruise ships are transiting Arctic waters with minimal search-and-rescue capacity nearby. Military planners and Coast Guard operators are extending operational reach into a domain they do not yet understand well enough to manage. The satellite coverage is limited. The observation networks are sparse. The forecasting models were not built for the region.
This is exactly the gap that Vaisala Xweather’s architecture is designed to address: Add proprietary observations at the edge, apply AI-driven modeling optimized for local conditions, and deliver real-time intelligence that narrows the uncertainty window from days to hours. “The technology gives us the opportunity to run a lot better,” Kuisma says, “in an area where you’re thinking about being operative, by doing the two things, adding observations and doing better modeling.”
The strategic demand signal is real and growing. Chinese investment in Russian Far Eastern infrastructure and interest in Arctic shipping routes, combined with American icebreaker acquisition and renewed Nordic security cooperation, means that operational activity in the High North is going to increase across military and civilian domains. The meteorological infrastructure to support that activity is currently inadequate. Building it through a government procurement process with its inherent lag means arriving late to a domain that is already becoming contested.
The Service Model Question: Entrepreneurial Government vs. Bureaucratic Government
What Vaisala Xweather offers governments is not outsourcing in the traditional sense. Kuisma is careful about this distinction. Classical outsourcing is the logic of efficiency or finding someone cheaper to do what you would otherwise do yourself. What he describes is something different: “You’re actually sourcing from someone who can do it better.”
This is not a cost-reduction argument. It is a capability argument.
The subscription model Xweather uses has structural implications that go beyond procurement mechanics. Because the company retains ownership of its capability and is paid for performance rather than for a product, it is free to improve that capability continuously without renegotiating a contract.
Xweather allows customers to better budget themselves while receiving increasing value from forecasts meaning a government client receives a better forecast next year than it received this year not because a new requirements document was written and a new contract awarded, but because the company’s incentive is to keep its forecasts ahead of what clients could obtain elsewhere.
This creates two distinct camps within defense establishments, as Kuisma has observed in his dealings with naval and air force clients. One camp seeks to incorporate the new capability into existing requirements processes, to treat it as another input to be managed through familiar acquisition pathways. The other camp, typically those operationally responsible for the domains in question, wants to move at the pace of the technology. “They’re willing to keep the bar high and keep increasing it,” Kuisma says, “and actually turn to companies to say, ‘Help us and take us to the next level.’”
The distinction is not trivial. It is the difference between a government that manages a product and a government that manages the evolution of a capability. The former is the default institutional mode. The latter is the condition under which a defense establishment can actually keep pace with commercial technology moving at private sector speed.
Environmental Intelligence as a Kill Web Input
The framing of weather intelligence as a critical input to distributed force operations is not an abstraction. For a Marine formation operating in the Pacific littoral — moving through a sequence of dispersed positions, timing its presence in each to avoid detection and exploit operational windows — meteorological accuracy at the operational scale matters enormously. Not global forecast accuracy. Not regional averages. Local, time-specific, operationally relevant weather data for a defined area over a defined period.
Xweather’s architecture is sized for exactly this. Its hyperlocal, real-time forecasting model can be applied to a specific operational area, not just to say what the weather will be globally, but to provide decision-relevant environmental intelligence for the area in which a force intends to operate for the next 96 to 120 hours. Whether a position at point X, Y, or Z offers the better meteorological window for a planned operation is exactly the kind of question the system is designed to answer.
Integrated into a broader sensor and intelligence architecture including the autonomous maritime systems and distributed sensor networks that are increasingly central to U.S. Indo-Pacific operational concept, this kind of environmental intelligence becomes a foundational input rather than a support service. Domain awareness in contested maritime and polar environments requires environmental awareness. The kill web concept depends on knowing not just where adversaries are, but what the operating environment allows and constrains for all parties.
The D-Day Reminder
Teppo Kuisma’s summary of the argument is concise: “Nothing has changed since D-Day.”
The requirement for accurate, timely environmental intelligence as a precondition for effective military operations is as old as organized warfare. What has changed is the technological basis on which that intelligence can be generated, the speed at which it can be delivered, and the institutional model through which defense establishments can access it.
Vaisala Xweather does not present itself as a defense company. It is a data and intelligence services business with a subscription model, proprietary observation networks, AI-driven forecasting, and a client base that includes some of the world’s most operationally demanding users. What it represents, from a strategic perspective, is the leading edge of a shift in which the capability to understand and predict the operating environment is increasingly generated outside government and available to those governments willing to adopt the procurement and management models that can take advantage of it.
Whether defense establishments will move fast enough to capture that advantage or whether institutional inertia will leave them relying on models that are a day and a half old when the operational window is measured in hours is the real question.