The AI compute gap: Enterprises are buying infrastructure faster than they can measure what it costs
THE SO WHAT
When 107 enterprises are accelerating AI infra spend faster than they can measure unit economics, AI has entered the 'blank check' phase of capex. This week, finance and engineering should align on a simple metric—cost per successful AI transaction or workflow—before approving the next specialized compute purchase.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIAI-powered software could recover 300GW of hidden capacity for the US power grid, enough to power thousands of AI data centers
If tools like GridCARE can actually unlock even a fraction of 300 GW of latent transmission capacity, the near-term constraint on AI buildout shifts from steel-in-the-ground to software and regulatory approval. Data center operators and large loads should be asking utilities what grid-aware software they’re piloting before committing to expensive on-site generation.
Applied AIThe agent security gap: 54% of enterprises have already had an AI agent incident, and most still let agents share credentials
More than half of surveyed enterprises already seeing AI agent incidents means this is no longer a hypothetical risk — it’s an active attack surface. If your agents don’t have per-agent scoped identities and least-privilege access today, you’re effectively running shared root accounts with an automated clicker attached.
Applied AIHere’s Why Anthropic Is Pushing States to Regulate AI Faster
When a major lab says last year’s California and New York AI transparency laws are already outdated, it’s a signal that state-level rules will iterate faster than most corporate governance programs. If you deploy frontier models across multiple states, you need a living compliance playbook, not a one-time policy PDF.
Applied AIGoogle Vids now lets you star in your own AI videos
Personalized AI avatars in Google Vids turn every knowledge worker into a scalable on-camera asset — training, sales, and internal comms can now be multiplied without studio time. Teams should start drawing hard lines on where synthetic self-representation is acceptable and where a real human still needs to show up.