From Wood Pulp to Cloud: The Data Center Cooled by the Baltic Sea

A bankruptcy notice in 2007. A €40 million gamble on a 50 year old paper mill in 2009. And today, the most energy efficient hyperscale data center on the planet, cooling servers with seawater through granite tunnels built when Dwight Eisenhower was president.

Most data center stories follow the same script: find cheap land, build a box, install chillers, plug in servers. Google's Hamina facility in Finland threw out that script entirely. When Joe Kava's team first walked through the abandoned Summa paper mill in early 2009, they weren't looking at real estate. They were looking at infrastructure that couldn't be replicated anywhere else on Earth.

The mill's underground tunnels, carved through solid granite and running a quarter mile to the Gulf of Finland, had been drawing massive volumes of seawater for steam generation since 1955. Stora Enso had spent decades perfecting the intake and outflow systems. Google saw something different: the world's largest natural chiller, operating at exactly the temperatures they needed, with zero electricity required.

Fifteen years later, that bet has become a €4.5 billion campus that maintains a power usage effectiveness of 1.09 while operating without a single mechanical chiller. The facility now heats 2,000 homes in Hamina with its waste heat, and Google calls it the blueprint for their next generation of thermal recovery projects worldwide.

But this isn't a story about perfect planning. It's about what happens when you buy industrial infrastructure that nobody wanted and figure out how to make it work for an entirely different purpose. Here's what we learned from the operators who built it, the mistakes they'd do differently, and why no other Google facility has tried to replicate it.

The Mill That Nobody Wanted

The Summa paper mill wasn't supposed to fail. When Alvar Aalto Foundation, Enso-Gutzeit (later Stora Enso) was building what they expected to be a multi generational asset. The location made sense: direct access to the Baltic for raw material shipping, abundant fresh water, and proximity to Finland's paper belt.

By 2006, the mill was producing 400,000 tonnes annually of newsprint, magazine paper, and book paper, generating €228 million in revenue. Not spectacular numbers, but solid industrial performance. Then the global financial crisis hit the print industry like a sledgehammer.

Stora Enso Mill Closure Announcement, along with mills at Kemijärvi and Norrsundet in Sweden. The company cited "dramatic cost increases" and the need to "safeguard long term profitability." Translation: newsprint was dying, and they needed to cut capacity fast.

The final paper rolls came off machines 2 and 3 on January 31, 2008. Four hundred twenty direct employees lost their jobs, plus another 170 service staff. For Hamina, a town of just over 20,000 people, it was an economic gut punch.

But the industrial infrastructure remained intact. The granite tunnels that had supplied seawater to the mill's steam systems since 1955 were still there, large enough to drive a tractor through. The electrical substations, the heavy foundations, the rail connections, the deep water port access. Most importantly, the environmental permits for large scale seawater intake and discharge had already been negotiated with Finnish regulators.

Ernst & Young was hired to find a buyer who would "establish economically viable businesses which will create new employment without competing with Stora Enso." For 14 months, the mill sat empty while they searched for someone who could see value in 166 hectares of industrial infrastructure that was built for an industry that no longer existed.

The €40 Million Bet That Changed Everything

Google Hamina Data Center and closed the deal in March for exactly €40 million. At the time, that seemed like a massive gamble. Google had never converted an industrial facility. They'd never operated seawater cooling. They were buying a paper mill in a country where they had no data center operations, in a town most of their engineers couldn't pronounce.

Joe Kava, then Google's Senior Director of Datacenter Construction and Operations, later admitted the internal reaction was mixed. You could build a perfectly functional data center on greenfield land for less money and with fewer unknowns. But Kava's team wasn't looking for perfectly functional. They were looking for something that could fundamentally change the economics of cooling.

The numbers that convinced them were surprisingly simple. A typical hyperscale data center uses 30 to 40% of its total electricity for cooling systems, mostly running chillers that fight the laws of thermodynamics. Every kilowatt of heat removed requires about 0.4 kilowatts of chiller electricity. At scale, that's tens of megawatts of pure overhead.

The Gulf of Finland offered something different: a thermal sink that naturally operated at exactly the temperatures Google needed, with unlimited capacity and zero electricity required. Winter water temperatures averaged 0.8°C. Even summer peaks rarely exceeded 17°C. Google Data Center Summit 2011 and run thermal modeling for wind patterns, tidal effects, and water density variations. The Gulf was more consistent than any mechanical system they could build.

The engineering challenge wasn't the temperatures. It was the salt. Seawater is incredibly corrosive to standard data center cooling equipment. Most operators won't even consider it because the maintenance costs overwhelm any energy savings. But Google's team designed around that constraint from day one.

Engineering for Eternity (Or At Least 20 Years)

The seawater cooling system at Hamina operates on a principle most data centers can't use: complete isolation between the natural cooling medium and anything that touches servers. Google built a two stage heat exchange architecture that keeps Baltic seawater entirely separate from the purified water systems that actually cool the servers.

Stage one uses standard server cooling with purified water in closed loops. Nothing exotic here, just highly efficient air handlers and liquid cooling blocks extracting heat from the server environment. That heated purified water then flows to stage two: custom built heat exchangers designed specifically for seawater exposure.

The seawater side heat exchangers use titanium plates throughout. Titanium costs roughly 10 times more than stainless steel, but it's effectively immune to seawater corrosion. All the seawater piping is fiberglass reinforced plastic, not metal. Hofmann Heat Exchanger Manufacturing, noting that "the pipe is the fiberglass reinforced pipe, the plate from the plate heat exchanger is made of titanium, which in order to resist the corrosion of seawater."

The seawater intake system uses the original 1950s granite tunnels, but with four separate stages of filtration that progress from coarse to microscopic. The first stage removes obvious debris, flotsam, and anything larger than a few centimeters. Stage two catches biological matter, seaweed, and smaller particles. Stages three and four handle microscopic filtration to prevent biofouling in the heat exchangers.

Here's where Google made a design choice that shows they were planning for decades of operation: they only use 15% of their seawater intake for actual heat exchange. The other 85% serves as a cooling medium for the return flow. After the warm water exits the heat exchangers, it's gravity fed to a separate tempering building where it mixes with the 85% of cold incoming seawater before being returned to the Gulf.

Kava was explicit about this design choice: Google Data Center Summit 2011 Google went beyond Finnish regulatory minimums because they were designing for permanent operations, not the minimum viable system.

The result is a cooling system that operates with zero compressors and zero refrigerants. The Gulf of Finland is literally Google's chiller, operating 24/7/365 at exactly the temperatures they need. When the Gulf freezes over every winter, it becomes an even more effective cooling medium.

What Actually Went Wrong (And What They'd Change)

Google doesn't publicize data center failures, but the Hamina project had challenges that anyone considering seawater cooling should understand. The biggest wasn't technical; it was regulatory and timeline.

The environmental permitting process for industrial seawater discharge in Finland is complex under EU environmental directives. Google needed approvals from the Regional State Administrative Agency of Southern Finland for both the intake modifications and the thermal discharge limits. Even though the paper mill had existing seawater permits, Google's use case was different enough to require new environmental impact assessments.

The first phase construction, which Google initially estimated at 18 months, actually took about 30 months from groundbreaking to full operations in September 2011. Some of that was the learning curve of converting industrial infrastructure, but permitting delays were significant. Helsinki Times Hamina Coverage, suggesting the complexity was higher than Google's standard data center builds.

The seawater filtration system required more maintenance than initially projected. The four stage setup works, but biological fouling in the Gulf varies seasonally and requires more frequent cleaning cycles than Google's models predicted. They've never disclosed specific maintenance costs, but industry experts estimate seawater cooling adds 15 to 25% to annual mechanical system maintenance compared to air cooled systems.

The bigger challenge was scalability within the same site. The original granite tunnels were sized for a paper mill's steam requirements, not a hyperscale data center's thermal load. Google has never disclosed the maximum cooling capacity of their seawater system, but industry analysis suggests they've had to supplement with air cooling for some of their later expansions. The slight uptick in PUE from 1.09 to 1.10 in their latest reporting might reflect this constraint.

If Kava's team were starting over today, they've indicated they'd focus more on the heat recovery opportunity from day one rather than treating it as an afterthought. The district heating partnership they launched in 2024 provides 80% of Hamina's residential heating needs and represents a new revenue stream that most data center operators ignore.

By the Numbers: When Engineering Excellence Pays Off

Fifteen years of operations data tells the story of whether Google's seawater gamble worked. The numbers are unambiguous: Hamina consistently operates at a power usage effectiveness of 1.09, matching Google's fleet wide average and significantly better than the industry average of 1.56.

For context, a PUE of 1.09 means that for every 100 watts of IT load, Google uses only 9 additional watts for cooling, power distribution, and facilities management. Industry average facilities use 56 watts of overhead for every 100 watts of IT load. At hyperscale, that difference is massive.

We track Google as the second largest data center operator in Finland with 193 MW of capacity across three facilities. The Hamina campus represents Google's largest single site investment in Europe, with cumulative spending reaching approximately €4.5 billion through seven expansion phases.

The economics work because Google eliminated the largest operational expense in data center cooling. Cornell University Lake-Source Cooling than when it used traditional chillers. Google has never published specific energy savings figures, but the absence of any mechanical refrigeration equipment implies similar performance.

The carbon impact is equally significant. Google achieved 97% carbon free energy at Hamina in 2024 and 98% in 2023, sourced through power purchase agreements with Swedish wind farms and three wind facilities near the complex. Combined with the elimination of chiller electricity consumption, Hamina operates with a carbon footprint that's effectively zero for cooling operations.

The Economic Transformation of Hamina

When Stora Enso closed the Summa mill in January 2008, Hamina lost its largest private employer and faced an uncertain economic future. Google's investment has transformed the town's employment and tax base in ways that most data center projects don't achieve.

Current permanent employment at the Hamina campus is approximately 400 people across full time and contractor roles: computer technicians, electrical and mechanical engineers, security, catering, and facilities management. The seventh expansion, expected to complete around August 2025, will add another 100 permanent jobs, bringing the total to about 500.

But the construction impact dwarfs the permanent employment. Each major expansion has generated 1,000 to 2,000 construction jobs. Deloitte Economic Impact Study during the two year construction period.

For a town of 20,000 people, Google's cumulative investment exceeding €4.5 billion is transformative. Business Finland Mayor Ilari Soosalu described Google as "an excellent example of a company with strong sustainable future orientation."

The property tax and building permit revenue alone has changed Hamina's municipal finances. Former Mayor Hannu Muhonen acknowledged that international companies "do not pay a high rate of corporate tax" but noted that property taxes and building permit fees provide substantial municipal revenue.

Heat Recovery: The Next Chapter

Google's newest innovation at Hamina represents a fundamental shift in how data centers interact with their surrounding communities. The heat recovery partnership launched in 2024 with Haminan Energia using waste heat from the data center operations.

The system uses a 7.5 MW heat pump supplied by Finnish company Nohewa, capable of producing 40 GWh per year of district heating. Construction began in November 2024 with expected completion by the end of 2025. When operational, it will heat approximately 2,000 households plus schools and public buildings throughout Hamina.

Google is providing the waste heat free of charge. Haminan Energia invested €5 million in the heat pump infrastructure. Haminan Energia CEO Statement

Jukka Vainonen, Google's Site Operations Manager at Hamina, framed the project as a global template: Google Hamina Heat Recovery Announcement

The district heating partnership represents a new revenue model for data center operators. Instead of treating waste heat as a disposal problem, operators can monetize it as a community utility. For Google, it strengthens their relationship with local government and provides a tangible community benefit beyond employment.

Why This Can't Be Replicated (And Why That Matters)

The most important lesson from Hamina is also the most frustrating: it can't be copied. Joe Kava was explicit about this at the 2011 Google Datacenter Summit in Zurich: Google Data Center Summit 2011

No other Google data center worldwide uses seawater cooling. Their other European facilities employ completely different approaches: Saint-Ghislain in Belgium uses canal water cooling, while Dublin relies on outside air cooling. Each design is optimized for local geographic and climatic constraints that don't exist elsewhere.

The specific conditions that enabled Hamina are essentially unique: pre-existing granite tunnels to deep water, industrial scale seawater permits already negotiated, year round cold water temperatures, low salinity (3.95‰ versus 35‰ ocean average), and available skilled industrial workforce after a mill closure. Most coastal markets lack at least three of these prerequisites.

The permitting alone would be prohibitive for a greenfield seawater cooling project. Google inherited environmental approvals that Stora Enso had spent decades negotiating with Finnish regulators. Starting from scratch, the environmental impact assessments, marine ecology studies, and regulatory approval process would likely take 3 to 5 years in most European jurisdictions.

But the transferable lesson isn't the seawater cooling. It's the heat recovery model and the approach to community integration. Google's statement that Hamina is "only the start" for heat recovery projects suggests they've learned something about data center economics that goes beyond cooling efficiency.

The bigger implication is that hyperscale data center design is moving toward site-specific optimization rather than standardized templates. Google's willingness to invest €4.5 billion in a single facility conversion suggests they see value in engineering for local conditions rather than deploying identical facilities globally.

Lessons for the Industry

Hamina proves that data center cooling doesn't have to be a solved problem. The industry default of building chillers and fighting thermodynamics exists because most operators optimize for speed and certainty rather than long term operational excellence. Google's approach required patience, capital, and willingness to solve problems that most operators would avoid.

The key lessons for other operators aren't about seawater specifically, but about the value of local resource optimization. Every data center site has unique thermal, geographic, or infrastructure characteristics that standard designs ignore. Google succeeded because they started with local constraints and opportunities rather than forcing a standard template onto the site.

The heat recovery opportunity is particularly transferable. Most data centers generate enormous quantities of waste heat that operators treat as a disposal problem rather than a potential revenue stream. Google's partnership with Haminan Energia shows that municipal utilities and district heating operators are willing partners for waste heat monetization.

The regulatory lesson is equally important. Environmental permitting for innovative cooling approaches takes longer and costs more than operators typically budget. Google's advantage was inheriting permits rather than negotiating them from scratch. Operators considering non-standard cooling should add 12 to 24 months to their development timelines for regulatory approval.

The employment impact also matters for community relations. Google's approach of hiring locally and providing substantial construction employment has generated sustained political support that most data center operators don't achieve. Mayor Soosalu's description of Google as having "strong sustainable future orientation" reflects community buy in that goes beyond standard corporate tax incentives.

The View from 2026

Fifteen years after Google's initial €40 million gamble, Hamina has become the template for how hyperscale operators can integrate with local communities and resources. The seawater cooling system operates as designed with industry leading efficiency. The heat recovery partnership will be operational within months. Employment has grown to 500 permanent positions with thousands of construction jobs across multiple expansions.

But the most significant achievement might be what Google proved about long term thinking in data center development. The infrastructure investments they made in 2009 for titanium heat exchangers, granite tunnel modifications, and voluntary tempering systems are still paying dividends in 2026. The facility that seemed expensive and complex compared to standard data center construction has delivered 15 years of operational excellence that no greenfield facility can match.

The slight uptick in PUE from 1.09 to 1.10 in Google's latest reporting suggests the facility is approaching some operational limits, possibly related to the maximum thermal capacity of the seawater system or the power density requirements of AI workloads. Google hasn't disclosed whether they've added supplemental cooling for recent expansions, but the physics of the original granite tunnels suggests there are constraints on further scaling within the same cooling architecture.

The seventh expansion, expected to complete around August 2025, will be the test of whether Google's seawater cooling approach can handle another major capacity increase or whether they'll need to supplement with conventional systems. Either way, Hamina has already proved that hyperscale data centers can operate with zero mechanical refrigeration while delivering industry leading efficiency and community benefits that most operators never achieve.

For the rest of the industry, Hamina remains an outlier rather than a template. But it proves what's possible when operators start with local resources and community integration rather than standardized engineering solutions. Google turned a bankrupt paper mill into a €4.5 billion facility that heats a town while cooling the cloud. That's not replicable, but it's a blueprint for thinking differently about what data centers can be.