{{first_name | Hi}}, your strategic AI update is here.
Oracle's New Mexico AI campus stalled on power and permits, Goldman sees Big Tech's AI capex hitting $1.2 trillion, and Microsoft moved Copilot agents to usage billing. If you're scaling AI, capacity is getting harder to secure, and your bill rises with every task an agent runs. I'd settle where your AI runs during this budget cycle. In two years, the share of IT leaders naming infrastructure as their top AI blocker went from 9% to 40%.
Innovation Snapshot
"The bottleneck is always at the top of the bottle."
— Nikolas Konstantin, Executive Coach
In today's lineup
Robots
Your AI plan is now an infrastructure plan
Real ROI from Gen AI
The gateway between your models and your budget
People
Astrid Atkinson gets data centers onto the grid faster
Love
Match AI training to the decisions each role makes
Competing in the Age of AI by Marco Iansiti and Karim Lakhani
Reading time: 3 min.
ROBOTS 🤖
How are robotics and AI changing industries? We break down the latest news, tools, and innovations for you.
Top Insights
Oracle's New Mexico AI campus stalls on power and permits, so AI capacity now waits on the grid.
Goldman sees hyperscaler AI capex up 50% to $1.2 trillion, with a $300 billion revenue bar to break even.
Microsoft puts coding and agents into one Copilot app, billing the agents by usage instead of per seat.
Japan's regulator tightens scrutiny of AI data center lending, as doubts grow over returns.
Australia weighs AI laws after an OpenAI agent breached its health database, including mandatory breach reporting.
Your AI plan is now an infrastructure plan
A Digital Realty survey of 2,131 IT decision-makers across 19 countries found the constraint on AI has moved. Two years ago, 9% named a lack of specialized infrastructure as their primary blocker. This year 40% do.
Money is still flowing. AI spending is set to rise 32% this year, and only 3% report no measurable return. The trouble is where the data and the workloads sit: 92% now tie data location decisions directly to their AI plans.
What to do about it:
Decide where data lives at the architecture stage, before procurement gets involved
Size capacity against inference volume, because 98% expect real-time AI within twelve months
Price private capacity against token spend before renting becomes the larger bill
Digital Realty's CTO, Chris Sharp, put it plainly: renting AI works at low volume, and at scale it becomes a liability.
How does your organization decide where AI workloads run today?
Real ROI from Gen AI
How human-centric AI delivers business results
Blue Origin runs 2,700 internal agents on Amazon Bedrock, where manufacturing teams resolve non-conformances 70% faster and four-day analyses now finish in four hours.
Dojo put Gemini Enterprise into its payments business, where staff built 680 agents in two months and cut complex chargeback dispute processing by 70%.
Peterborough Regional Health Centre consolidated 18 production systems onto Microsoft Fabric, cutting the wait for an inpatient bed from 56.8 hours to 32.
Tool Spotlight
Your weekly briefing on tools that create competitive leverage
The gateway between your models and your budget
Kong puts one control plane in front of every model you call, routing requests and metering spend. Version 2.0 reached general availability earlier this month, putting text, audio, image and video pricing in one catalog. Best for CIO teams already running an API gateway.
TrueFoundry runs inside your own VPC, on-premise or air-gapped, with hosted and self-hosted models behind one API and per-team token quotas. Best for AI leaders who must prove where each request ran and what each team spent.
Three more gateways worth a look
OpenRouter suits teams that want one API across many model providers, with failover and nothing to host.
Baseten suits enterprises whose biggest bill is serving open-source models at scale.
Cloudflare AI Gateway suits teams already on Cloudflare who want caching, rate limits and per-token cost tracking.
PEOPLE 👥
Meet the innovators turning bold ideas into real-world impact.
Transformation Champion
Astrid Atkinson gets data centers onto the grid faster
AI data centers wait years for a grid connection, because utilities size interconnections to a peak that rarely arrives. Atkinson ran infrastructure teams at Google before founding Camus Energy. Its FlexConnect product applies cloud oversubscription logic to the distribution grid, cutting interconnection timelines by up to five years. Camus made this year's 50 by 2050 list earlier this month.
LOVE ❤️
Practical wisdom, growth tactics, and a must-read book that will challenge the way you think.
Match AI training to the decisions each role makes
"If you treat AI as an innovation, it will remain an innovation," Schneider Electric's Chief AI Officer Philippe Rambach told CIO.inc this month. Schneider gates every use case on business KPIs and trains four groups differently:
Every employee, down to the production line, learns how AI works
Senior managers learn business value
AI specialists keep pace with the technology
Process and product owners learn to build AI into their roadmaps
Rambach calls that last group the hardest to serve.
Transformative Reads
One book, handpicked from my conversations with friends, industry leaders, and tech innovators:
In Competing in the Age of AI, Harvard Business School professors Marco Iansiti and Karim Lakhani show how firms like Ant Group put data and algorithms at the core of the operating model to remove the old limits on scale. A strong case for fixing the operating model before buying more capacity.
Perfect for: CIOs and COOs redesigning how work gets done around AI.
In Culture
In 2024, 9% of IT leaders hit this wall. In 2026, 40% did.
THANK YOU!
Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.
We just ask for one thing: tell one friend about Lighthouse today!
Want daily AI & Innovation insights? Connect with me on LinkedIn or X.
Got suggestions? Just reply to this email!
Much Love,
Matt and the Future Works team
