Taiga Cloud service powers future Generative AI evolution

Managing Director of Taiga Cloud (part of The Northern Data Group) Karl Havard showcases the company’s clean-energy-powered CSP offering for generative AI

All of this talk about technology evolution—artificial intelligence (AI) looming and companies building much more advanced solutions to natural digital transitions in the world—makes you wonder how companies can build the most intricate systems at such a fast rate. The development and operation of highly intelligent AI is reliant on significantly larger amounts of data, and the areas to process and store said data have their own operational challenges and carbon footprints to manage. For Taiga Cloud, It’s all in a day’s work to ensure that the latest digital projects are supported in a sustainable way: powered by clean energy and with minimal impact to the environment.

Europe’s largest provider of clean-energy-powered machine learning (ML) and artificial intelligence (AI) capacity was born out of German based Northern Data Group. Recognising the ever-growing demand for high-performance compute technology, the company foresaw the need for much more computing capacity to support Generative AI development and other associated technical revolutions.

Having joined Northern Data in March 2023, Karl Havard, now Managing Director for Taiga Cloud (formerly Northern Data’s Cloud Solutions business) shared Aroosh Thillainathan’s vision to support growing trends and leverage the best technologies of its partners.

“One thing that struck me when initially engaging with Aroosh and Northern Data Group was the vision and the opportunity to build a brand new type of cloud business specifically for Gen AI. Such opportunities don't come along that often; it's a hugely exciting one, and it made the decision to join Northern Data Group very easy,” says Havard.

Taiga Cloud is supported by its sister business Ardent Data Centers; a key strategic partner from a data-centre perspective, which was also born out of Northern Data to manage its own specific business area. Although Taiga operates worldwide, its heart lies within Europe, it is a European company, and has the ability to offer its customers true EU compliance and sovereignty.

“We do have global customers, but our heart is in Europe. We're building the largest estate of the latest Nvidia hardware, and this expansion will continue right through 2024, and beyond. The A6000 and A100s we already have, have been in place for two years, all now being used. For those that don’t know, the A100 is what ChatGPT was trained on,” says Havard as he demonstrates the significance of Taiga’s current footprint across Northern Europe.

“We invested an initial €400m in June 2023, and boosted this with a further €320m in early December 2023 to expand our GenAI estate. We will be close to 20,000 GPUs of the latest H100s in 2024, The H100 being the machine that everybody in the industry is trying to get their hands on. Our first installation will be live at the end of 2023.”

Building the compute power that GenAI developers need

While this is good news for the startups and AI-powered businesses in Europe, there’s a high level of demand for the GPUs that Taiga Cloud needs to support a growing number of GenAI clients. Luckily for the company, its Elite Partnership with NVIDIA, the manufacturer of this hardware; as well as other strategic alliances with Hewlett Packard Enterprises (HPE) and Giga Computing Technology, has helped ensure Taiga Cloud are the first to market to deliver this highly sought after technology in Europe.

Not only has this already built a concrete position for Taiga, it has also enabled the business to offer a tailored set of services to its clients in order for them to get the most out of the capacity that they adopt.

“One of our key values and beliefs is that, by building this in Europe, and on such a large scale, we have a duty to all types of organisations who want to use this technology to bring their best ideas to life. This includes the startup community, enterprises, research industries and even educational institutions. We are not going to sell out to one customer, despite having the opportunities to do so.” says Havard. But, what exactly does this mean? He explains further: “We want to democratise access to this computer power, which means we'll offer portions of our estate to various different organisations to allow them to train, tune and infer their large language models and to offer them the ability to roll-on and roll-off the platform to suit their needs..”

Not only will such a model enable access to more businesses as and when they require Taiga’s service, but also encourage higher utilisation to reduce the impact from an environmental perspective. If customers can “roll-on and roll-off” as and when they need, then no power goes to waste.

“The easy thing for us to do would be to sign massive three-year deals that utilise all of our estate capacity, but the reality is others are doing this in the marketplace,” says Havard. “Fundamental to Taiga Clouds vision is to bring a fresh perspective and democratise access to this technology which will make a long lasting and positive impact on society,”

“We do not want to constrain or restrict access to the computing power that others are desperate to get their hands on, due to the supply being limited for the foreseeable future. So we're very happy to enable our customers to take one-month or three-month rolling contracts, as well as longer ones that enable our customers innovation.”

Taiga Cloud is very humble with its hot-commodity GPUs, which aids its engagement, specifically with startups, giving them the service they need on a flexible basis. A statement that defines it all comes from Havard who says the business’s existence is to help organisations or all kinds to bring their best ideas to life. In mentioning this, he goes on to praise some of the great feats to have come from partnering with its services.

“We're really pleased that Poolside AI decided to work with us and use our GenAI Cloud. They have a great proposition, which is to unlock humanity’s potential by using GenAI to simplify software development. They’re creating a world where you don’t necessarily have to be a coder to actually build software,” Havard explains.

“Soon you will be able to use Poolside to write prompts that will make software development more accessible. This is a great partnership as we share similar values, and we’re inviting more organisations like them to partner with us.”

This is great indeed and also shows the commitment that Taiga Cloud is making to wider industries. As an AI development enabler, the company’s service can be linked to groundbreaking innovation and potential solutions.

A sustainable approach at the heart of data-centre development

The company doesn’t falter when it comes to sustainability. If anything, from discussing this with Havard, it seems as though the company has it figured out. With the target being net-zero emissions, the company is actively sourcing 100% renewable energy for its operations, and when it comes to cooling, direct-to-chip will be the standard that Taiga is setting—not to mention getting creative with the positioning of its sites; on the edge of the Arctic Circle and many metres below sea-level.

One of its data centres is located in a disused mine, which provides the perfect environment for natural cooling and also eliminates the concerns about space. From an energy perspective, though, Taiga sources its energy from renewable means, citing hydroelectric power as a key component of its Scandinavia-based operations.

“We will always, wherever we build or whoever we partner with on the data-centre front, use 100% carbon-neutral energy to power and cool the sites,” Havard says. “As well as having these clean sources of energy, we also use it in the most efficient ways”

“Our power usage effectiveness (PUE) numbers will always be 1.2 or lower”—the perfect score being one. “In fact, Boden in Sweden is 1.05 and Lefdal in Norway is 1.15.”

He continues: “If you can get parity, it means all the energy is powering and cooling the IT equipment and nothing else. We will always want to be as efficient as we possibly can be and we’re evaluating all sorts of different powering and cooling technologies; we know we can do more.”

Setting it in stone, the lifelong sustainability goal is clear for Taiga Cloud—reduce emissions to zero and only consume the energy that it needs. With support from the likes of NVIDIA, HPE and Giga Computing Technology, the team is capable of accessing the right equipment to help achieve this goal. Moreover, partnering with a provider like Bulk is a foundational move that supports the sustainable day-to-day running of its data centres.

“As well as the low PUEs, the clean energy aspects, and the location, Bulk also offers some really strong elements around tier-three capability. Customers that have moved through their evolution of generative AI and moving to interference, will need absolute tier-three capability,” says Havard.

“There are various elements to it, including very high-levels of security and resilience, as well as only allowing a maximum of 1.6 hours per annum downtime. So we absolutely need that type of partner. Bulk offers that.”

Leveraging the partnership with Bulk also enables lower latency, achieving 10 milliseconds round trip time, or less, which some of our customers insist on.

“Our growing estate is spread across our data centres and is configured in identical islands, securely segmented for each of our different customers' needs. We have built them to the highest spec, and to be clear the H100s are the SXM version, in islands of 2,032 GPUs, with BlueField DPU management and 3.2TB InfiniBand interconnectivity. Basically, super efficient to provide the desired outcomes, faster.“

In the foreseeable future, Taiga’s partner network and devotion to new innovations in GenAI will see the company through its continuous development to provide startups and technology majors the computing power to deliver on their goals. The partner network the company is building will provide additional ML & AI services and software to its customers and accelerate the achievement of their desired objectives.

“The next 12 to 18 months are all about effective execution,” Havard explains. “As the H100s come in, we’ll be making sure they are deployed on time, clients are onboarded, and that they’re working at their absolute best.

“On top of that, we will provide a high level of managed services and expertise to support our customers as the market matures.”