AI & Data Centers
Explain the infrastructure the AI era runs on
Data centers decide who leads in AI. That story has two sides: the company building them at unprecedented scale, and the businesses deciding they no longer need their own.
Selected work
The race for compute, told to the people who would build it
Every major technology company was claiming AI leadership. This one could prove it with hardware: new data-center-scale training clusters, each linking tens of thousands of GPUs. The audience that mattered most was the engineers it needed to hire, so the launch was built around a technical blog written for them, cleared with research, legal and hardware partners, and then carried to the press and social as proof the company was building at the frontier.
A data center is a neighbor before it’s a server farm
Communities hear ‘data center’ and think power bills and empty jobs. Openings were told instead as local stories: the construction work, the permanent jobs, the investment in schools and grids, and what the building actually powers. Investors and neighbors got the same account, which is why it held.
Why a software company started making chips
To the market, a social company designing its own processors looked like a distraction. The story reframed it as inevitable: off-the-shelf chips weren’t built for the company’s models, so it designed hardware and software together. A side project became a long-term infrastructure strategy.
One milestone, three audiences
Engineers wanted specifications, investors wanted to know the spending would pay off, and reporters wanted the stakes. The messaging gave each what it needed from the same facts: the size of the investment, the leap from the last generation, and a system designed end to end, from the physical layer to the products people use.
When owning the data center becomes the problem
The insurer covered half its market and had been the plan of choice for 80 years, largely because it said yes to every custom request. That flexibility left it running a heavily customized core system that was out of support, with its data trapped inside. Membership couldn’t grow much further, so the move off its own data center had to pay for itself. The executive story started there, not with the technology.
Ask before you demo
Most technology pitches open with the product. This one opened with the client’s executives ranking what they actually wanted from a migration (lower costs, automation, agility, getting out of the data center business) and naming what stood in the way. The demo came last, framed around their answers.
Banking without the branch, in a country built on branches
Japan still runs on cash, paper forms and personal seals, and online banking has never caught on the way it has elsewhere. For regional banks the math was grim: half were on course for losses within a decade, and their customers in mountain villages and on small islands were aging and hard to reach. One regional bank didn’t try to force customers online. It put an AI- and chat-powered tablet in the hands of its staff. Opening an account fell to about six minutes, 70% to 80% of branch paperwork disappeared, and branches turned from waiting rooms into lounges. Because the tablet travels, the bank can now come to customers: the bank became a person, not a building.
Making sustainability part of the code
Software has a carbon footprint that almost nobody measures. A program to change that trained 30% of the technology workforce in sustainable software practices. The communications turned a niche engineering idea into something clients asked for: the program influenced more than $1 billion in sales and cut more than 3 million tons of emissions.
Case studies


