Beverley Eve

Apparently, enterprise AI has entered its "PIVOT" era 😅 Just came across some enterprise research that was published by OpenAI earlier this month that suggests businesses are moving from asking AI questions to giving it actual work to do A few findings that caught my attention: 1️⃣ Its most active 10% of enterprise customers now generate 8.3× more AI output per user than typical firms, up from 2.6× in Jan this year 2️⃣ Agent use is spreading well beyond engineering. Since Feb, weekly active enterprise Codex users grew 108× in legal, 41× in sales and recruiting, and 26× in marketing. 3️⃣ Access alone doesn’t appear to be enough... OpenAI points to shared workflows, employee learning, data infrastructure and governance as important ingredients for wider adoption. Same technology. Very different progress in turning it into repeatable ways of working. So yes… PIVOT! (But do make sure everyone knows which direction you’re going first ;) Dropping report link in the comments - worth a look!

Coverage alone doesn’t always mean readiness. A network can reach an area and still come under enormous pressure when demand surges, infrastructure is disrupted or operational conditions unexpectedly change. That’s why the question I now ask about connectivity isn’t simply how fast or how far it goes. It’s this: what will it enable people to keep doing when conditions become difficult? For public safety, that can mean maintaining communications and access to information when they matter most. It also means designing resilience before an incident occurs, not trying to add it in the middle of one. In my latest Tech on Tour newsletter, Beyond Coverage, I take a deeper look at why coverage and capacity are not the same, how growing network demands affect operational readiness, and why trusted teams need trusted systems. Read the full newsletter below. When you think about operational resilience, what conditions are you actually preparing for? T-Priority Partner

☁️ So… do we owe everyone an apology about 'the cloud'? For years, people in tech have patiently explained that it isn’t really in the sky. But now SpaceX is seriously exploring plans to build AI data centres in orbit. A recent Reuters article by Akash Sriram (link in comment) got me thinking about how quickly the demand for AI compute is making ideas that once sounded like science fiction 🤯 seem increasingly plausible. JPMorgan’s observation definitely makes it feel more tangible: “Beyond 2029, we expect SpaceX to pivot to orbital compute for incremental capacity additions, while continuing to operate and maintain its terrestrial compute clusters.” So this isn’t necessarily about replacing data centres on Earth, but rather adding capacity in space as demand continues to grow. The argument is that space could offer abundant solar energy and more room to scale without placing further pressure on power grids here on Earth - sounds great. The reality is a tad more complicated. Radiation, cooling, maintenance, latency and cost all present enormous challenges. So I guess the cloud isn’t heading into space tomorrow... BUT the fact this is now part of a serious infrastructure strategy feels pretty significant in and of itself. ❔ What do you think: a genuinely brilliant long-term answer to AI’s growing infrastructure demands, or has the race for compute officially left the planet? (haha sorry, couldn't resist!) Let me know your thoughts!

“A hyperscaler that designs its own accelerators is renting NVIDIA GPUs from a rocket company.” I think that sentence pretty neatly sums up where we are with AI infrastructure in 2026. 😅 The quote's from a Forbes piece by Ashish Bhatia that I just read on NVIDIA’s growing role beyond the chips themselves... and there’s a lot to unpack. For so long, the AI conversation has centred on models and GPUs. But increasingly, headlines seem to be shifting towards everything required to actually turn those GPUs into usable compute at scale: land, power and shell (LPS). The article makes an interesting point that we’re now seeing companies with plenty of compute but less immediate demand effectively supplying those with huge demand and insufficient capacity, or as Ashish puts it: "the compute-rich and adoption-poor are leasing capacity to the compute-starved and adoption-rich" Which suggests the scarce resource that really matters may not always be the model, or even the chip... rather, the megawatts needed to make all of it work. That also makes NVIDIA interesting to watch beyond the GPU itself. As AI adoption grows, the company increasingly seems to be thinking about the entire ecosystem required to get compute into the hands of organisations that need it. AI may look like a software revolution from the outside, but underneath its is becoming an enormous physical infrastructure story too. So where do you think the biggest constraint on AI growth will ultimately sit? Compute, power, capital, connectivity.. or something else? Let me know in the comments. https://lnkd.in/e5-vAdy2

❔ Should we be calling it 'Rogue AI'? I'm sure, like me, you've seen these words or similar increasingly pop up in headlines and tech news over the recent weeks. This write-up in Reuters from Greg Bensinger about AI 'Going Rogue' got me thinking: https://lnkd.in/eUMF3zd8 This point, for example... “AI is vanishingly unlikely to be conscious,” wrote Anil Seth, professor of cognitive and computational neuroscience at the University of Sussex, in a post on social media site Bluesky. “The anthropomorphic language we use (e.g., "going rogue") makes the challenge of alignment/control much harder than it already is.” S my question is - are we humanising AI to shirk responsibility? Or to grab attention? Either way, is it the right thing to do? ❔Let me know your thoughts in the comments.