Filing: Alphabet's AI unit DeepMind lost £460.9M last year, down 2% on 2018, and Alphabet wrote off a further £1.1B in debt
Tim Bradshaw / Financial Times :
Context & Ripple Effects
This filing closes the loop on a run of DeepMind loss disclosures: the lab lost £123.5M on £40.2M of revenue in 2016, then $572M in 2018 while carrying more than $1.2B of debt due for repayment. The 2019 numbers — a £460.9M loss, essentially flat, plus Alphabet writing off £1.1B of debt — show the parent absorbing the lab's liabilities rather than calling them in.
The debt write-off is the telling move: it converts intercompany loans into equity on Alphabet's books, cleaning DeepMind's balance sheet a year before the lab's revenue tripled to £826M and it posted its first-ever profit in 2020.
First-order effects
- Alphabet takes a £1.1B write-off and continues to fund DeepMind's ~£460M annual burn, while the lab itself is relieved of debt that was due for repayment.
Second-order effects
- With the balance sheet cleaned, Alphabet's scrutiny shifts to commercialization — the path that produced the 2020 profit turn and, later, the 2022 employee-expense cuts of 39% when revenue fell.
Third-order effects
- The pattern that emerges across these filings — parent-funded losses, then integration and cost discipline — culminated in the Google Brain merger that delivered DeepMind's first meaningful operating profit of £136M in 2023, suggesting Alphabet treats research labs as subsidiaries to be absorbed into product organizations rather than permanent cost centers.
The trend: Big-tech research labs are moving from open-ended parent-funded losses to commercial integration, with balance-sheet cleanups like this write-off marking the pivot point.