Infrastructure Resilience Lessons After Cloud Migration

When a council moves its records to new software, things can go wrong fast.

Three UK councils have run into trouble after shifting planning and land search work to Arcus Global’s software service.

People have felt the pain in real life: house sales have fallen through, land searches have piled up, and one council even approved a 5G mast by mistake.

Havant Borough Council, on the south coast of England, moved to Arcus Global in 2024 after a hosting deal with Capita was set to end.

After the move, residents complained about bad planning data, missing files, and lost comments on planning requests.

A council paper said the team faced data gaps, weak reports, and rough system controls, and the land charge’s part needed a major fix and a fresh data move.

Bracknell Forest Council also had a rough start, with delays affecting about 500 property deals because land searches backed up.

The council said it had 695 open search requests in mid-April, with 480 personal searches and 215 official ones.

Both the council and Arcus Global said the service had serious data problems and apologized for the mess, which hit local buyers and sellers hard.

Back in 2024, Haringey Council in London approved a 5G mast at Hornsey Cricket Club by mistake after an IT glitch in its planning team.

The council later said it took the blame, said sorry, and said the issue had been fixed.

The bigger message is clear: when public systems move too fast or carry old data into new tools, small IT faults can cause very real harm.

Kenya’s halted Microsoft data center plan says a lot about the gap between big tech plans and hard limits on power.

President William Ruto said Kenya has paused the $1 billion Microsoft-G42 data center because the country does not have enough electricity for it.

He said the site would have used about a third of Kenya’s total power supply, which would have put too much strain on the grid.

The project was meant to sit about 100 kilometers northwest of Nairobi and use geothermal power.

It was also meant to support cloud services through Microsoft’s Azure platform for both companies and government users.

A concept note was sent to the National Treasury, but it did not get approval, so the plan stopped moving forward.

The setback shows a bigger problem across Africa: demand for AI and cloud growth is rising faster than the power and other systems needed to support it.

Kenya still wants to raise national power capacity to 10,000 megawatts by 2030, and it is seeking billions of dollars to help fund energy and infrastructure work.

Even with this pause, data center work is still moving ahead in Kenya and elsewhere, with new projects from Airtel Africa’s Nxtra and more Microsoft spending in South Africa.

The main lesson is simple: big digital plans need strong power behind them, or they stall.

Cloud reliability has significantly declined, with 2025 recording over 48,000 outages across major providers like AWS, Azure, and Google Cloud. The core issue isn't just the frequency of these failures, but their intensity. Because businesses have moved from simple migration to deep, concentrated infrastructure dependence, a single bad configuration change can now ripple globally within minutes. These outages are no longer isolated incidents; they have become systemic single points of failure for thousands of downstream services.

Modern software relies on complex supply chains where applications depend on services that depend on other clouds. This tight coupling creates a "blast radius" that leads to massive financial disruption, often costing enterprises billions in lost revenue and SLA payouts. Many organizations are discovering too late that being a "customer of a customer" of a major provider still leaves them vulnerable. When the top of the food chain fails, the entire interconnected web of mobile apps, websites, and data systems halts, sometimes for days.

Enterprises can no longer assume that cloud providers will solve uptime issues on their behalf. To survive, companies must pivot toward active resilience planning, investing in redundant systems and transactional processing that can route around failures. As providers potentially become complacent or attempt to economize on operational talent, the responsibility for maintaining uptime shifts to the user. Building a resilient infrastructure is now a commercial necessity to protect against the inevitable dips in the economy and stock market caused by these widespread disruptions.

Anthropic has a fresh way to deal with the heavy demand on Claude, and it comes from SpaceX’s giant Colossus 1 data center in Memphis.

To keep up with fast‑growing use of Claude Code and its other AI tools, Anthropic signed a deal with SpaceX to use more than 300 megawatts of compute, spread across over 220,000 Nvidia GPUs in Colossus 1.

This extra power should let Anthropic raise limits and cut wait times for Claude Pro and Claude Max subscribers, who have often hit capacity walls during 2026.

Anthropic, SpaceX, and xAI

SpaceX, which also owns the AI firm xAI, gains a strong new customer for its large GPU cluster at Colossus 1.

For SpaceX and xAI, selling access to this supercomputer helps fund their own work on Grok and other models while keeping the hardware busy.

For Anthropic, the deal is mainly about removing compute limits so Claude can serve more developers and teams who rely on Claude Code and other Claude tools in their daily work.

Bigger plans on the horizon

Both firms also talked about interest in future “orbital” AI compute, where large data centers could one day sit in space and draw power and cooling in new ways.

Those long‑term ideas are still early and may not happen soon, but they show how far companies are willing to go to get enough compute for advanced AI systems.

Why this matters for Claude users

In the near term, the effect is simple: Anthropic gets access to one of the largest GPU clusters in the world, and Claude users should feel fewer slowdowns and fewer “at capacity” messages.

Developers using Claude Code, and customers paying for Claude Pro and Claude Max, are likely to see smoother service and room for new features that need heavy compute.

In short, Anthropic’s deal with SpaceX’s Colossus 1 is about turning raw GPU power into a better, more reliable Claude experience.

China’s cloud market is heating up fast, and the main push is coming from AI

Cloud infrastructure spending in mainland China reached $14.7 billion in Q4 2025, up 26 percent from the same quarter a year earlier.

It was the third quarter in a row with growth above 20 percent, and Omdia expects the same 26 percent growth rate to carry on through 2026.

AI and agents as the growth engine

Omdia links most of this jump in cloud spend to AI and the fast rise of AI agents.

Chinese firms are paying for more cloud capacity to train and run AI models, support private AI setups, and power classic workloads like compute, storage, and databases.

Focus is shifting from just models to agents that can call tools, complete tasks, and plug into business systems, which pushes deeper and steadier use of cloud platforms.

How Alibaba Cloud, Huawei Cloud, and Tencent Cloud respond

Alibaba Cloud leads the market with about 37 percent share, helped by strong AI product growth, its Qwen3.5 model family, upgraded Model Studio, and the Wukong agent platform.

Huawei Cloud holds around 17 percent share and is pushing industry AI, including for healthcare, along with tools such as the CodeArts coding agent and a new AI ecosystem fund.

Tencent Cloud has close to 10 percent market share and is doubling down on agents too, with upgrades to its agent platform and products like WorkBuddy and QClaw meant for real use in companies.

Partners become key to real AI value

Omdia notes that partners now play a larger role in turning cloud and AI into real business outcomes.

Roughly a quarter of China’s Q4 2025 cloud infrastructure revenue came through partners, and that share is set to grow as more firms rely on integrators and software vendors to bring AI into daily work.

For tech leaders watching China, the clear message is that AI and agents are now central to cloud growth, vendor plans, and partner ecosystems.

In short, China’s cloud story in Q4 2025 is really a story about AI demand turning into hard spend on infrastructure.

This big change helps the company handle more work as its group of artists grows by 250,000 new people every month.

Roemie Hillenaar, who started Creative Fabrica, says this move helps creators turn their ideas into real products without needing special technical skills.

New Tools For Creators

The platform is using a special system called the Gemini Enterprise Agent Platform to run its new creative tools.

Artists can now use Google systems like Veo to make videos and Lyria to create professional music and sounds.

They also use models like Imagen and Nano Banana to create images that are ready to be used or sold right away.

Fair Pay For Artists

A new feature called Edit-to-Earn lets people change existing designs while making sure the original artist still gets paid.

This keeps the human artist in charge even when computer tools are doing some of the heavy lifting.

Joost Smit from Google Cloud Benelux says this setup allows for much more creation while keeping the work original and safe.

Custom Gifts and Memories
For Mother's Day, the company released a tool called Moments that picks through your phone photos to find the best shots.

It uses smart agents to put these photos together into a finished gift with music and special effects.

This partnership proves that using big cloud systems can make professional design tools easy for anyone to use at home.