We’re halfway through the year, and it got me thinking about the State of Cloud and State of DevOps reports that the big players used to publish annually. They even got printed. But it’s 2025, and the pace of change has outstripped the cadence of those once-a-year snapshots. Trends now evolve in weeks, not years. By the time the ink dries on the report, it’s out of date.

With that in mind, I thought it fitting to reflect on the themes shaping cloud and AI this year in bite-sized pieces — the trends, opportunities and challenges that stand out most clearly for me and the tech leader community I’m part of.

AI is Solving the Bootstrap Problem, Not the Retention Problem

Over the past 6–12 months, it has become easier than ever to launch a SaaS or AI-powered product. Models, APIs, and integrations can be spun up and tested in hours. Just the other day, our team went from requirements gathering to a working prototype in less than 24 hours. Now comes the longer part: making it secure, integrated, and embedded in our processes.

I even noticed a startup launch its product recently — forgetting to strip out traces that it was built on Lovable before going public. That’s how fast things are moving.

While startups still face the challenge of limited runway, the capital stretch is less constrained than it used to be. Bootstrapping is no longer the biggest bottleneck. Instead, the real hurdle is retention. Getting early users is easy; keeping them long term is not. Many early adopters churn once the novelty fades, or when the subscription cost outweighs the perceived value.

This has sharpened the focus for venture capitalists, who now scrutinize retention and paid subscriber metrics more than ever before writing the next cheque.

Bots Starting to Flood the Web

If last year we started seeing AI integrated into products, this year we are starting to see companies preparing infrastructure for the side effects. One of the biggest? Traffic from bots and changing consumer preferences – driving traffic via virtual agents who can now complete whole tasks on behalf of the user.

Web applications must now be architected for an influx of automated traffic — both legitimate (agents, crawlers, integrations) and malicious (scraping, fraud, denial of service). Managing identity and access is becoming critical not only for protecting users and delivering great user experience to them, but also for distinguishing human activity from automated in the context of running cost of the app and the underlying infrastructure.

That’s why I expect identity and access management skills — combined with security and broader compliance expertise — to be in high demand. When I spoke recently with local students about “future careers,” I found myself likening it to becoming a modern-day good cop, safeguarding a digital estate. Will it get the children interested enough to take it up as a career? Who knows…

Observability for Virtual Agents

As more companies deploy virtual agents — customer support bots, AI copilots, workflow automators — the question becomes: are they performing well, and are they worth it? How much are they being used?

That is where Observability proves really handy. Leaders now need visibility into Performance (speed, accuracy, helpfulness of agents), Cost impact (are token charges or API calls ballooning unexpectedly?) Customer outcomes (is the agent improving satisfaction or adding friction?)

The organisations that thrive will be those who treat their AI agents with the same discipline as production infrastructure, including performance and user experience monitoring.

Partners Moving Upstream into Strategy

The technology menu is expanding faster than most enterprises can keep up. AI frameworks, cloud platforms, security add-ons, observability tools — every choice has downstream impacts on cost, compliance and overall complexity of the estate that determines how efficient it is and how much value can be derived from it.

This is pulling cloud partners and service providers further upstream. They are no longer just implementers but strategic advisors, helping enterprises cut through the noise and align technology choices with business goals.

Soft skills are becoming as important as technical depth. Yet finding engineers who excel at both is no small feat. It’s something I’ll be encouraging young people to think about as they consider their future careers: technical mastery plus the ability to communicate and collaborate will be game-changing.

Data Residency: USA and New Zealand in Focus

Data residency remains a board-level concern in many industries. In the US, state-level privacy laws and regulatory debates continue to shape where and how data can be processed, who can work with it and how they can access it.

In New Zealand, AWS and Microsoft are investing heavily in local infrastructure to meet rising demand for in-region compute and storage. For regulated sectors like healthcare, finance, and government, in-country residency is a key consideration. And it opens the door for innovation closer to home which is great.

Migration momentum is still strong globally, alongside the initial investment in local, sustainably powered datacenters. With energy supply and costs rising, hyperscalers will be expected to partner with governments and energy companies to build more sustainable energy generation capacity, and offset their environmental impact in the regions where they operate. This is something leaders like Satya Nadella are openly acknowledging and solving for. I have been surprised though, that sustainability hasn’t started to feature as high on enterprise priority list as I’d have expected…it certainly isn’t driving repatriation of workloads to more sustainable regions (yet!).

Looking Ahead

Cloud in 2025 is a paradox: adoption has never been easier, yet navigation has never been more complex. AI accelerates innovation while introducing new risks around cost, retention, and trust. Identity and observability are becoming frontline disciplines. Partners are being asked to step up into strategy.

For enterprises, the winning strategy this half of the year — and beyond — is practical governance paired with selective adoption. The key question isn’t “Can we do this?” but “What happens if we don’t — and how will we manage it once we do it?”