The week in London technology, read and picked by hand. No algorithm, nothing in here because someone paid for it.
Tuesday · · 6 minute read
Turba Labs raises $52M for AI compute optimization
On 6 October 2026, Turba Labs announced it secured $52 million in funding. The company focuses on optimising AI compute performance.
Turba Labs aims to double the world's compute capacity without building new data centres. The company addresses the challenge of compute scarcity and complexity, stating that AI infrastructure has outgrown its current management tools. Their GPU performance platform operates across the full AI infrastructure lifecycle, specifically targeting the hardware beneath AI models. This differs from other tools that typically optimise only the model itself. Turba Labs' software works to increase output per GPU and watt of power, lower cost per unit of compute, and improve real-time predictability. The company believes that the true measure of compute lies in its delivered capability, not just its availability.
Scarlet, a five-year-old UK startup that reviews and certifies software and AI medical devices for regulatory approval, has raised a $50 million Series B funding round. Headline led the investment, with Notion Capital also participating alongside previous investors Atomico, Creandum, Kindred Capital, and Entrepreneurs First.
On October 9, 2026, London-based Sava Technologies announced a €32.08 million Series B funding round. This brings the company's total funding to €60.6 million. The round was led by Ascensia, with participation from Balderton Capital, Norrsken VC, Simplyhealth Ventures, JamJar Investments, and Pentland Ventures. This capital will fund manufacturing scale-up and commercial readiness for its wearable biosensor platform.
Sava develops a next-generation glucose microsensor that detects molecules painlessly beneath the skin. The company states its microsensors are about ten times shorter than the filaments in traditional continuous glucose monitors. Sava has also partnered with Ascensia for commercialisation and distribution, initially in Europe. Renato Circi, co-founder and Co-CEO of Sava, said, "For decades, continuous monitoring has been held back by the same legacy technology. This is about to change." The company plans a pivotal clinical study this year and targets CE approval for non-adjunctive use, which would allow the device to inform insulin dosing decisions.
Reported by EU-Startups, via StrictlyVC, 2026-10-13
Salience Labs Inc. raised $125 million in Series B funding on October 14 2026. The investment round was led by Bullhound Capital. New investors included Intel Capital, AMD Ventures, SBI Group, Hudson River Trading, and the British Business Bank. Existing backers also participated. The company develops specialised all-optical circuit switching technology for AI data centres.
This technology aims to resolve the strain on current data centre infrastructure. Most data centres rely on electronic switches and repeated optical-to-electrical-to-optical signal conversions. These systems struggle to meet the throughput demands of AI computing clusters, leading to latency bottlenecks and higher power consumption. Salience Labs proposes eliminating electrical signals entirely, transmitting data over optical connections for improved efficiency. Vaysh Kewada, co-founder and Chief Executive, stated, "Optical circuit switching will play a critical role in AI data center networking."
The company recently demonstrated a 64-port all-optical photonic switch at the ECOC 2026 conference in September. The new funding will be directed towards volume manufacturing, geographic market expansion, and the research and development of higher-density switches.
Reported by SiliconANGLE, via StrictlyVC, 2026-10-15
Microsoft reportedly pulls out of UK data center deal
Microsoft was reportedly poised to become the anchor tenant at a data centre site located in Loughton, Essex. This reported withdrawal is a setback for Nvidia-backed Nscale, which had filed for a US IPO last month.
Reported by Opening Bell Daily, 2026-10-19
Neural networks may be cleaner than previously thought
Interpretability researcher Nick Cammarata at Goodfire published a paper finding that individual neurons in modern AI models represent coherent concepts more often than previously thought. Preliminary work on Qwen3-32B suggests 90% of neuron poles have one primary function.
HMV's Oxford Street flagship store is closing again after its landlord doubled the rent. This decision has rendered the venue "commercially unviable". The store at 363 Oxford Street had only reopened in 2023, returning to its original location where it was founded in 1921. Its return was seen as a victory against the proliferation of "shady shops" in central London, following a period when the site was occupied by Candy World.
The landlord, identified as Glory Step Investment, was found to own the property as of January this year. HMV managing director Phil Halliday described the decision to increase rent as "frustrating". Glory Step Investment, once registered in the British Virgin Islands, is now listed as being based in Kowloon, Hong Kong. The company acquired the lease for the Oxford Street property in 2008 for £40m.
On Oct 14, 2026, City of London Police introduced a new enforcement approach at Bank Junction. The initiative targets cyclists who ignore red lights. Instead of an immediate fixed penalty notice, offending riders are now offered the opportunity to watch a short, hard-hitting road safety film. This new method provides an alternative for addressing traffic signal violations by cyclists in the area.
This approach gives riders a direct choice when stopped for disregarding red lights. They can either view the film or receive the standard immediate fine. The police were present at Bank Junction to implement this new system, which focuses on tackling cyclists who ignore red lights.
Meta releases advanced AI models Muse Wonder and Muse Wonder Full
Meta has released Muse Wonder and Muse Wonder Full, its most advanced AI models. Muse Wonder Full performs just under Claude Opus 5.5 and around GPT-6 Astra levels on benchmarks. It also introduces Active Memory and Active Focus features. Muse Wonder is priced at $4 per 1M input tokens and $20 per 1M output tokens.
Reported by MTS, 2026-10-12
OpenAI UK staff received average £2.3m in 2025
OpenAI UK paid its 110 London staff an average of £2.3 million each in 2025. This compensation comprised £794,000 in cash and approximately £1.5 million in share-based awards. The company reported this against £293 million of revenue, with £167.8 million in total share-based compensation.
McKinsey authors report that automation may require over 70 percent of US workers to reinvent their work as the mix of tasks changes. About 25 percent of workers could see more than 30 percent of their task time reshuffled or shifted to new activities.
Reported by mckinsey.com, via McKinsey, 2026-10-17
On May 10, 2024, Ilya Strebulaev and Alex Dang released their new book, *The Venture Mindset: How to Make Smarter Bets and Achieve Extraordinary Growth*. The book identifies nine principles that define a specific way of seeing the world. Strebulaev is a professor of finance at Stanford Graduate School of Business. Dang is a tech executive and innovation advisor who previously worked at Amazon and AWS.
This venture mindset contrasts with traditional business approaches that prioritise caution and consensus. It embraces uncertainty and accepts failure as an option. Strebulaev and Dang argue that this mentality has applications far beyond venture capital. Ilya Strebulaev states, "The VC mindset is becoming more and more useful and likely will dominate, or should dominate, a number of industries for a while." The authors suggest that industries undergoing dramatic changes can benefit from adopting these principles.
I have said it for sometime now, Compute and power availability (you can't have one without the other) are more likely to set the winners of AI than frontier chips.
While countries spend Millions on sovereign AI the reality is the infrastructure won't allow it ever to compete.
The AI players may change in the coming years but the technology is going nowhere.
