Google is doubling down on building out its AI business in the U.K. On Monday morning in London, the CEO of Google DeepMind Demis Hassabis and Google Cloud CEO Thomas Kurian appeared alongside customers BT and WPP to spell out some of its plans.
The company said it would expand UK data residency to include Agentspace so that AI agents for enterprises built on Google infrastructure can be hosted locally — a key detail for organizations that are wary of hosting data outside of their own purview.
Alongside this, Google is introducing more financial incentives for AI startups to work with Google awarding those joining its new UK accelerator up to £280,000 in Google Cloud credits; plus expanded AI skills training.
Additionally, Google used the event at its DeepMind offices to announce that Chirp 3, the company’s audio generation model that was developed there, would be added to its Vertex AI developer platform. You can read more about that here.
“Agentic” has become the code-word for how enterprises will practically start adopting AI — the pitch being that AI agents can be built both to help people do their work faster, and to interface better with customers.
Agentspace is Google’s platform for building these assistants for work. One of the most notable features in it is NotebookLM for enterprises — a service that can ingest large amounts of information and then summarise it, now set up for use in large business environments. Other features of Agentspace include multimodal search, and, of course, the building of AI agents using generative AI.
Google launched Agentspace as a beta in December 2024, while Google announced data residency for the UK in October 2024, allowing for private and public organizations to store data at-rest, train AI and run inference on Gemini 1.5 Flash within the U.K. Today’s news brings Agentspace into the UK data residency region.
The idea here is to lay the groundwork to get more businesses working with Google (instead of its competitors) to build their future AI services, but also to try to bridge some of the trust issues that have arisen among organizations over how their proprietary data is handled in the building of AI and other services in the cloud. Data is the “new oil” and so it remains a precious commodity.
“We know from our research that a significant percentage of organizations across Europe are still very nervous about using AI in the public cloud,” said IDC analyst Mick Heys. “They want to deploy AI, and they’re happy to do experimenting in the cloud, but when it comes to actual deployment at scale, they want to do that in dedicated infrastructure or some kind of co-location environment, something that is really managed much more closely by them. That’s partly because they’re just nervous about data security, privacy and sovereignty issues. So those things are still very live.”
“They will have full control to keep the data where they need it,” Kurian said at the event today.
The two companies that joined Hassabis and Kurian on stage today are longtime partners in AI services. Both BT and WPP have inked development and data partnerships with Google Cloud and are early adopters of pilot programs, such as newer releases of Imagen, Veo and Gemini.
“We are quietly reinventing all our operations,” said BT CEO Allison Kirby of how it’s using AI. “Operationally, there is just a huge potential for us.” Some of these have a very immediate customer face, for example in trying to help detect phone scams and to improve customer service agents. Back in 2023, it said it would be axing up to 55,000 jobs with one-third to be replaced by AI.
Google is on a development tear at the moment with its AI business. Last week saw the launch of a host of new Gemini developments, specifically Gemini 2.0, which starts to work around multimodal generation and understanding in real-time, using word prompts to generate images, a new robotics model and advances with its lightweight Gemma model.
Separately, the U.K. government is making a huge push to promote more AI development, both within its ranks and more widely as an industry. At the same time, European businesses are pushing for less reliance on Big Tech, in favor of homegrown businesses and services.
The U.K. government has laid out plans, and is pressing individual divisions, to demonstrate how it could adopt more generative AI services aimed at speeding up paperwork and building services across data that had previously been siloed by function and department.
The government’s dogfooding is part of a bigger strategy: its hope is that AI really will prove to be a big economic wave, and it wants to make sure the U.K. can catch it. To show that the U.K. is “open for business” for AI, it’s made commitments to AI regional zones that will include data center capacity as well as regulatory changes to smooth the way to working with more data, among other measures.
“These models are global and used everywhere, and we need to set an international standard for this,” Hassabis said of the move to relax how IP is handled in AI environments in response to a question today about how the U.K. is looking to change rules around how AI companies can use intellectual property to train models, one of the more controversial topics around how AI is being ushered into use in the U.K.
Interestingly, the two main firms the government has named so far that it is working with are OpenAI and Anthropic, two rivals to Google in the area of generative AI services. Google’s announcements are late to the party but could pave the way for more collaboration on the government front going forward.
