As data-center developers frantically seek to secure power for their operations, one startup is proposing a novel solution: Build them into floating offshore wind turbines.

San Francisco–based offshore wind-power developer Aikido Technologies today announced its plans to start housing data centers in the underwater tanks that keep its turbine platforms afloat. The turbines will supply the power for the servers, and onboard batteries and grid connection will provide backup.

The company’s first prototype, a 100-kilowatt unit, is scheduled to launch in the North Sea off the coast of Norway by the end of this year. A 15-to-18-megawatt project off the coast of the United Kingdom may follow in 2028.

Aikido is one of several companies planning data centers in unusual places—underwater, on floating buoys, in coal mines and now on offshore wind turbines. The creativity stems from the forces of several trends: rapidly rising energy demand from data centers, the need for domestic renewable power production, and limited real estate.

The North Sea serves as an ideal first spot for floating, wind-powered data centers because European policymakers and companies are looking to regain domestic control over energy production. They’re also looking to host an AI economy on servers within the continent’s boundaries. Floating wind platforms keep the compute out of sight while tapping the stronger, more consistent air streams that blow over deep waters, where traditional, seabed-mounted turbine monopiles can’t go.

“A lot of energy in the clean-energy space is focused on powering AI data centers quickly, reliably, and cleanly in a way that does not upset neighbors and remains safe, fast, and cheap,” says Ramez Naam, an independent clean-energy investor who does not have a stake in Aikido. “Aikido has that, and a smart team,” he says.

Floating Wind-Power Designs Evolve

Aikido’s design builds on many iterations tested by the growing floating wind industry. When Norwegian energy giant Equinor finished construction on the world’s first floating wind farm in 2017, it kept the turbines upright with ballasted steel columns extending 78 meters into the water—a design called a spar platform. This gave it a dense mass like the keel of a boat. Since then, the floating wind industry has largely coalesced around a semisubmersible design based on oil and gas platforms. Semisubmersibles don’t go as deep as spar platforms; instead, they extend buoyancy horizontally. Anchors, chains, and ropes keep the platform floating within a certain radius.

Aikido is taking the semisubmersible approach. Its football-field-size platform holds the turbine in the center, and three legs extend tripod-like outward, like a Christmas-tree stand. At the end of each leg is a ballast that reaches 20 meters deep. This holds tanks largely filled with fresh water to maintain the platform’s buoyancy in the salty ocean.

The data centers will go in the upper part of each ballast tank. There’s room for a 3- to 4-MW data hall in each tank, giving the platform a combined compute of 10 to 12 MW. Below the data halls is an open chamber used as a safety barrier, and below that sit the freshwater tanks. The water is piped up to the data center for liquid cooling of the servers. The warmed water is then funneled back down the ballast into the tank. There, proximity to the cold ocean water cools it again as the heat is conducted out through the tank’s steel walls.

“We have this power from the wind. We have free cooling. We think we can be quite cost competitive compared to conventional data-center solutions,” says Aikido CEO Sam Kanner. “This crunch in the next five years is an opportunity for us to prove this out and supply AI compute where it’s needed.”

One challenge, he says, is that liquid cooling can’t cover all the data center’s needs. For example, heat generated from Ethernet switches that connect the GPUs can’t be liquid-cooled with commercially available technology. So Aikido installed an air-conditioning method for that.

Another challenge is the marine environment, which is “pretty brutal to engineer around because there’s the increased salinity, there’s debris, and there’s various kinds of corrosion and fouling of metal piping that you wouldn’t have in a freshwater environment,” says Daniel King, a research fellow at the Foundation for American Innovation in Washington who focuses on AI infrastructure.

Offshore Data Centers Face Challenges

Aikido’s plan avoids the prickly not-in-my-backyard complaints that are dogging both onshore wind and data-center projects. It might also circumvent some inquiries into water usage and power demand too, or so Aikido’s thinking goes.

But it might not be that easy. “Instinctively many people reach for offshore or even orbital outer-space data centers as a way to circumvent the typical burdens of environmental reviews,” says King. “But there could be more or additional requirements around discharging heat and the effects that has on marine life that are different from the considerations of a terrestrial data center. It’s unclear to me whether this actually makes life easier or harder for a developer.”

Prefabricated data halls could be installed quayside, followed by final electrical and plumbing connections to commission the data center.Aikido

Aikido’s “design choice to use the fresh water in the ballast as a working fluid is a novel one” that, thanks to the closed-loop system, may “alleviate some of the engineering problems you see when a really high temperature fluid is pumping its heat directly into a marine environment,” King says.

Offshore sites are also vulnerable to sabotage, King notes. Since Russia’s invasion of Ukraine, fleets of vessels directed by the Kremlin have reportedly started messing with offshore wind and communications infrastructure in northern Europe. Russian and Chinese boats have allegedly cut subsea cables in recent years.

But vandalism is a risk anywhere, including at conventional data centers, Aikido CEO Kanner notes. Unlike those on land, where the local police have jurisdiction, Aikido’s data centers would enjoy protection from national coast guards, which he suggests gives an added degree of security.

North Sea Hosts Clean Energy

Kanner first began thinking about offshore wind turbines as a place to build data centers after a chance phone call with a cryptocurrency billionaire. The financier wanted to know whether turbines in international waters could power servers generating digital tokens at a moment when crypto-mining faced increased scrutiny from regulators. The talks fizzled. But that encounter sparked Kanner’s curiosity about how to use power generated onboard floating turbines.

When ChatGPT emerged in 2022 and sparked a heated debate over how to power and cool such technology, the idea to put the data center in the floating turbine clicked for Kanner. The idea really congealed after he met with the chief executive of Portland, Ore.–based Panthalassa. The wave-energy company was proposing to enclose small, remote data centers in buoys attached to equipment that generates power from the surf. Panthalassa just completed its full-scale prototype tests off the coast of Washington state last summer.

At that point, Aikido had already designed a modular platform for floating wind turbines. Each platform consists of 13 major steel components that are snapped together with pin joints—like IKEA furniture. The platforms fold up in a flat configuration that takes up roughly half the space of other designs, allowing it to be transported by a wider range of ships, according to Aikido. From there, it was a matter of figuring out how to accommodate a data center in the unused space.

Aikido’s prototype will use a refurbished Vesta V-17 turbine. It will need onboard batteries for backup power and will also be connected to the grid for additional power during seasons with less wind. Aikido envisions eventually sprinkling its data centers among large arrays of offshore turbines to tap into that larger power infrastructure.

Between Russia’s threat to expand its war in Ukraine to EU countries and the Trump administration’s bid to pressure Denmark into ceding sovereignty of Greenland to Washington, Europe is scrambling to build up its own energy production and AI capabilities. The North Sea, increasingly, looks like a primary theater of that effort. In January, nearly a dozen European nations banded together in a pact to transform the North Sea into a “reservoir” of clean power from offshore wind.

Google’s newest AI model is here: Gemini 3.1 Flash-Lite, and the biggest improvements this time around come in cost and speed, especially for enterprises and developers seeking to leverage powerful reasoning and multimodal capabilities from the U.S. search and cloud giant.

Positioning it as the most cost-efficient and responsive model in the Gemini 3 series, Google is offering a solution built specifically for intelligence at scale.

This launch arrives just weeks after the February debut of its heavy-lifting sibling, Gemini 3.1 Pro, completing a tiered strategy that allows enterprises to scale intelligence across every layer of their infrastructure.

Technology: optimized for the “time to first token”

In the world of high-throughput AI, the metric that often dictates user experience isn’t just accuracy—it’s latency. For real-time customer support, live content moderation, or instant user interface generation, the “time to first answer token” is the primary indicator of whether an application feels like a tool or a teammate. If a model takes even two seconds to begin its response, the illusion of fluid interaction is broken.

Gemini 3.1 Flash-Lite is engineered specifically for this instant feel. According to internal benchmarks and third-party evaluations, Flash-Lite outperforms its predecessor, Gemini 2.5 Flash, with a 2.5X faster time to first token. Furthermore, it boasts a 45 percent increase in overall output speed — 363 tokens per second compared to 249.

This speed is achieved through what Koray Kavukcuoglu, VP of Research at Google DeepMind, describes in an X post as an unbelievable amount of complex engineering to make AI feel instantaneous.

Perhaps the most innovative technical addition is the introduction of thinking levels.

Standardized across both the Flash-Lite and Pro variants, this feature allows developers to modulate the model’s reasoning intensity dynamically. For a simple classification task or a high-volume sentiment analysis, the model can be dialed down for maximum speed and minimum cost.

Conversely, for complex code exploration, generating dashboards, or creating simulations, the thinking can be dialed up, allowing the model to perform deeper reasoning and logic before emitting its first response.

Product: benchmarking the lite-weight heavy hitter

While the “Lite” suffix often implies a significant sacrifice in capability, the performance data suggests a model that punches well into the territory of much larger systems. Gemini 3.1 Flash-Lite achieved an Elo score of 1432 on the Arena.ai Leaderboard, placing it in a competitive tier with models much larger in parameter count.

Key benchmark results highlight its specialized strengths across diverse cognitive domains:

• Scientific knowledge: 86.9 percent on GPQA Diamond.

• Multimodal understanding: 76.8 percent on MMMU-Pro.

• Multilingual Q&A: 88.9 percent on MMMLU.

• Parametric knowledge: 43.3 percent on SimpleQA Verified.

• Abstract reasoning: 16.0 percent on Humanity’s Last Exam (full set)

The model is particularly adept at structured output compliance—a critical requirement for enterprise developers who need AI to generate valid JSON, SQL, or UI code that won’t break downstream systems.

In benchmarks like LiveCodeBench, Flash-Lite scored a 72.0 percent, outperforming several rivals in its weight class, including GPT-5 mini, which scored 80.4 percent on a different subset but lagged significantly in speed and cost efficiency.

Furthermore, its performance on CharXiv Reasoning (73.2 percent) and Video-MMMU (84.8 percent) demonstrates that its multimodal capabilities are robust enough for complex chart synthesis and knowledge acquisition from video.

The intelligence hierarchy: Flash-Lite vs. 3.1 Pro

To understand Flash-Lite’s place in the market, one must look at it alongside Gemini 3.1 Pro, which Google released in mid-February 2026 to retake the AI crown. While Flash-Lite is the reflexes of the Gemini system, 3.1 Pro is undoubtedly the brain.

The primary differentiator is the depth of cognitive processing. Gemini 3.1 Pro was engineered to double the reasoning performance of the previous generation, achieving a verified score of 77.1 percent on ARC-AGI-2—a benchmark designed to test a model’s ability to solve entirely new logic patterns it has not encountered during training.

While Flash-Lite holds its own in scientific knowledge at 86.9 percent, the Pro model pushes that boundary to a staggering 94.3 percent, making it the superior choice for deep research and high-stakes synthesis. The application focus also differs significantly based on these reasoning gaps.

Gemini 3.1 Pro is capable of vibe-coding—generating animated SVGs and complex 3D simulations directly from text prompts. For example, in one demonstration, Pro coded a complex 3D starling murmuration that users could manipulate via hand-tracking. It can even reason through abstract literary themes, such as translating the atmospheric tone of Emily Brontë’s Wuthering Heights into a functional web design.

Gemini 3.1 Flash-Lite, conversely, is the workhorse for high-volume execution. It handles the millions of daily tasks—translation, tagging, and moderation—that require consistent, repeatable results without the massive compute overhead of a reasoning-heavy model.

It fills a wireframe with hundreds of products instantly or orchestrates intent routing with 94 percent accuracy, as reported by early testers.

1/8th the cost of the flagship Gemini 3.1 Pro model (and cheaper than its predecessor, Flash-Lite 2.5)

For enterprise technical decision-makers, the most compelling part of the Gemini 3.1 series is the reasoning-to-dollar ratio.

Google has priced Gemini 3.1 Flash-Lite at $0.25 per 1 million input tokens and $1.50 per 1 million output tokens.

This pricing makes it significantly more affordable than competitors like Claude 4.5 Haiku, which is priced at $1.00 per 1 million input and $5.00 per 1 million output tokens.

Even compared to Gemini 2.5 Flash, which cost $0.30 per 1 million input, Flash-Lite offers a cost reduction alongside its performance gains.

When contrasted with Gemini 3.1 Pro—which maintains a price of $2.00 per million input tokens for prompts up to 200k—the strategic advantage of the dual-model approach becomes clear. In high-context usage (above 200,000 tokens per interaction), Flash-Lite is actually between 12x and 16x cheaper.

By using a cascading architecture, an enterprise can use 3.1 Pro for the initial complex planning, architectural design, and deep logic, then hand off high-frequency, repetitive execution to Flash-Lite at one-eighth of the cost.

This shift effectively moves AI from an expensive experimental cost center to a utility-grade resource that can be run over every log file, email, and customer chat without exhausting the cloud budget.

Community and developer reactions

Early feedback from Google’s partner network suggests that the 3.1 series is successfully filling a critical gap in the market for reliable autonomy.

Andrew Carr, Chief Scientist at Cartwheel, has tested both models and noted their unique strengths. Regarding 3.1 Pro, he highlighted its substantially improved understanding of 3D transformations, which resolved long-standing rotation order bugs in animation pipelines.

However, he found Flash-Lite to be a different kind of unlock for the business: “3.1 Flash-Lite is a remarkably competent model. It is lightning fast, but still somehow finds a way to follow all instructions… The intelligence to speed ratio is unparalleled in any other model”.

For consumer-facing applications, the low latency of Flash-Lite has been the key to market expansion.

Kolby Nottingham, Head of AI at Latitude, shared that the model achieved a 20 percent higher success rate and 60 percent faster inference times compared to their previous model, enabling sophisticated storytelling to a much wider audience than would have otherwise been possible.

Reliability in data tagging has also emerged as a standout feature. Bianca Rangecroft, CEO of Whering, reported that by integrating 3.1 Flash-Lite into their classification pipeline, they achieved 100 percent consistency in item tagging, providing a highly reliable foundation for their label assignment and increasing confidence in structured outputs.

Kaan Ortabas, Co-Founder of HubX, noted that as a root orchestration engine, Flash-Lite delivered sub-10 second completions with near-instant streaming and 97 percent structured output compliance.

On the flagship side, Vladislav Tankov, Director of AI at JetBrains, noted a 15 percent quality improvement in the Pro model, emphasizing that it is stronger, faster, and more efficient, requiring fewer output tokens to achieve its goals.

Licensing and enterprise availability

Both Gemini 3.1 Flash-Lite and Pro are offered through Google AI Studio and Vertex AI. As proprietary models, they follow a standard commercial software-as-a-service model rather than an open-source license.

Operating through Vertex AI provides grounded reasoning within a secure perimeter, ensuring that high-volume workloads—like those being run by Databricks to achieve best-in-class results on the OfficeQA benchmark—remain protected by enterprise-grade security and data residency guarantees.

However, they also are limited in terms of customizability and require persistent internet connectivity, as opposed to purely open source rivals like the powerful new Qwen3.5 series released by Alibaba over the last few weeks.

The current preview status for Flash-Lite allows Google to refine safety and performance based on real-world developer feedback before general availability.

For developers already building via the Gemini API, the transition to 3.1 Pro and Flash-Lite represents a direct performance upgrade at the same or lower price points, effectively lowering the barrier to entry for complex agentic workflows.

The verdict: the new standard for utility AI

The release of Gemini 3.1 Flash-Lite represents the final piece of a strategic pivot for Google. While the industry has been obsessed with state-of-the-art reasoning for the most complex problems, the vast majority of enterprise work consists of high-volume, repetitive, but high-precision tasks.

By providing both the brain in Gemini 3.1 Pro and the reflexes in Gemini 3.1 Flash-Lite, Google is signaling that the next phase of the AI race will be won by models that can think through a problem, but also execute that solution at scale.

For the CTO or technical lead deciding which model to bake into their 2026 product roadmap, the Gemini 3.1 series offers a compelling argument: you no longer have to pay a reasoning tax to get reliable, instantaneous results. As Flash-Lite rolls out in preview today, the message to the developer community is clear: the barrier to intelligence at scale hasn’t just been lowered—it’s been dismantled.

“5G” is an umbrella term that encompasses the current fifth-generation cellular wireless network technologies. All the major carriers and phones support 5G connections, which can offer faster data speeds than older technologies such as 4G LTE or 3G.

Essentially there are three types of 5G: Millimeter-wave (mmWave), which can be fast but has limited range; low-band 5G, which has slower speeds but works on a broader range; and midband, which is a balance between the two that’s faster than low-band but also covers a larger range than millimeter-wave. Midband also incorporates C-band, a batch of spectrum auctioned off by the Federal Communications Commission in 2021.

Your phone’s 5G connection depends on which type blankets the area you’re in, as well as other factors, such as population density and infrastructure. For instance, mmWave is super fast, but its signals can be thwarted by buildings — or glass, or leaves — or whether you’re inside a structure.

When your device is connected to a 5G network, it can show up as several variations such as 5G, 5G Plus, 5G UW or others, depending on the carrier. Here’s a list of icons you see at the top of your phone for the major services:

AT&T: 5GE (which isn’t actually 5G, but rather a sly marketing name for 4G LTE), 5G (low band), 5G Plus (mmWave, midband)

Verizon: 5G (low band, also called “Nationwide 5G”), 5G UW/5G UWB (midband and mmWave, also called “5G Ultra Wideband”)

T-Mobile: 5G (low band), 5G UC (midband and mmWave, also called “Ultra Capacity 5G”)

There’s also 5G Reduced Capacity (5G RedCap), which is a lower-power, smaller-capacity branch of 5G used by devices such as smartwatches and portable health devices; the Apple Watch Ultra 3, for example, connects via 5G RedCap.

Just around the corner is 5G Advanced, promising much faster speeds due to carrier aggregation, or combining multiple spectrums.