AI’s office land grab: what it means for London’s secondary space

London’s office market is being reshaped by an unlikely new tenant class: artificial intelligence firms.

According to Knight Frank, AI firms leased 661,068 sq ft of London office space in the first half of 2026 alone, exceeding the 500,000 sq ft taken across the whole of 2025. Since the start of 2025, AI firms have committed to almost 1.2 million sq ft, including Anthropic’s 158,000 sq ft letting at One Triton Square in Euston and Databricks’ 137,311 sq ft lease at Network W1 in Fitzrovia for its new EMEA headquarters.

Why this is happening

Knight Frank’s Philip Hobley notes that AI companies are moving from flexible, early-stage space into substantial, permanent headquarters as they secure funding, grow revenues, and make long-term commitments to London. Demand is clustering around King’s Cross, Euston, and Fitzrovia, where firms can access transport, universities, research institutions, and the wider tech ecosystem, echoing the campus feel of Silicon Valley.

The advantage: a shot in the arm for prime stock

For landlords and investors, AI-driven demand has been a welcome development. This has been broadly positive news for landlords, and reassured office investors who feared AI-driven automation would eliminate jobs and reduce demand for desk space. As one developer put it, ‘Real estate is the most tangible measure of economic confidence, and AI is not London’s threat but a growth lever’. Vacant space left behind in the post-pandemic market is being absorbed, particularly in central London technology clusters such as King’s Cross. British Land, for example, refurbished a building returned by Meta and subsequently secured Anthropic as an occupier after demand from the life sciences sector softened.

The disadvantage: reinforcing the two-speed market

The catch is that AI occupiers overwhelmingly want the same thing as banks, law firms, and hedge funds: best-in-class buildings. They are competing for the same prime buildings as these established sectors, adding another layer of pressure to an already tight market. The result is a sharply bifurcated landscape. The gap between Grade A and Grade B performance widened through 2025, with premium buildings achieving record pricing while secondary stock saw rent declines of up to 19%, and Savills forecasts Grade B rental growth of -1.5% in 2026.

Secondary buildings – typically older, in non-prime locations, and often lacking modern climate control and energy efficiency – are being left behind rather than lifted by the AI wave. Some commentators suggest that while AI may boost demand within London’s technology clusters, it could simultaneously reduce overall office employment, further increasing vacancies in poorly located secondary stock.

A growing legal and regulatory divide

The challenge for owners of secondary stock is not merely a leasing problem but an increasingly legal and regulatory one. The same buildings being overlooked by AI occupiers are often those facing the greatest exposure to tightening environmental and sustainability requirements. With minimum energy efficiency standards expected to become more demanding over the coming years, landlords of older office assets may be forced to undertake significant capital expenditure to preserve lettability and value.

In practice, this means the premium commanded by Grade A space is no longer driven solely by occupier preference. Regulatory compliance is increasingly becoming a differentiator in its own right, creating a further divergence between buildings that can attract institutional capital and modern occupiers, and those that risk becoming stranded assets.

A second squeeze: Covid downsizers now scrambling for space

AI firms are not the only ones adding pressure to the market. Many businesses reduced their office footprints during the pandemic, expecting hybrid working to permanently reduce space requirements. As return-to-office (RTO) policies have strengthened, some are finding they no longer have enough space.

A Knight Frank survey found just 21% of occupiers are now seeking to reduce office space, the lowest level in seven years. Firms including JPMorgan, WPP, Amazon and Novo Nordisk have tightened attendance rules through 2025 and 2026, with HSBC being the clearest cautionary tale: having already announced a 40% reduction in its London office space by moving from Canary Wharf to a smaller building in St Paul’s, the bank has since realised it will lack sufficient space and is now assessing options, including retaining smaller offices elsewhere in the city.

This dynamic is compounding the squeeze that AI leasing is putting on prime stock. Around 30 large companies are each looking to lease more than 100,000 sq ft in London, and one developer has warned there simply will not be enough space for even half of them to move. For secondary buildings, this represents a genuine silver lining largely absent from the AI story: with new supply constrained, improving demand from returning occupiers could ultimately benefit second-hand office space, as firms priced or squeezed out of prime buildings are forced to compromise on quality just to secure enough desks. Corporate confidence is visible elsewhere too: Amazon is opening new offices in Shoreditch as part of a £40 billion UK investment, while a global bank has committed to a major new tower in Canary Wharf.

The catch is that this demand is not evenly distributed. Occupiers pursuing RTO policies are also concentrating on well-located, amenity-rich buildings that encourage employees back into the office. As Savills has noted, occupiers are becoming more selective rather than simply taking more space. Poorly located secondary offices are therefore unlikely to see a significant uplift.

Knock-on effects

There may nevertheless be some indirect benefits for higher-quality secondary stock. As prime space becomes scarcer and more expensive, smaller occupiers may increasingly consider refurbished secondary buildings. At the same time, tightening MEES requirements, including the anticipated 2031 deadline for larger properties to achieve a minimum EPC B rating, are driving refurbishment, redevelopment and alternative-use conversions, reducing the volume of weaker office stock.

The outlook

For prime London real estate, AI has arrived as a genuine new demand driver. But for secondary stock, the effect is closer to indifference than uplift: AI firms are pulling capital, talent, and construction investment toward a handful of quality clusters, leaving older, poorly located buildings to face obsolescence, conversion, or a longer wait for occupiers who can no longer afford – or find – anything better.