Businesses are using more AI for the same money, while exposed hacking tools and voter pushback show who pays AI's hidden costs.
The AI Transformation Brief
The price of AI is dropping, and the costs it leaves behind are landing on someone else. Ramp's data shows business AI use rising while spending barely grew over late summer. AI Weekly Hundreds of copies of an AI hacking tool sat open on the internet for anyone to use. TechCrunch Voters rejected data centers in 40 of 41 local ballot measures. Futurum Every company now has to name who pays for those costs.
Card and bill-payment data from Ramp shows business spending on AI usage grew 1.5% across August and September combined, after rising 42% in July alone. October picked up to 8.9%. AI Weekly Ramp's lead economist says price cuts from OpenAI and Anthropic made AI cheaper to use, and almost all AI spending comes from fast-growing tech companies. The median firm spends $12 to $15 per employee a month, and the top 10% spend $600 to $1,000. TechCrunch
Falling prices put buyers in charge. Ramp's economist calls it a buyer's market, with suppliers offering discounts to lock in spending. TechCrunch A discount tied to a long commitment can cost more than it saves when prices keep dropping every quarter. Renegotiate AI contracts every six months, avoid multiyear minimums, and measure the cost of each finished task before approving bigger budgets.
Security firm Censys found more than 800 separate copies of Artex, an open-source tool that uses AI to probe and break into computer systems, reachable online without a password. TechCrunch Censys counted 1,465 exposed control screens as of November 4, with 155 also running other attack software. FourWeekMBA Artex has been linked to attacks on South Korean banks and on IT support firm ConnectWise. Many of those attacks relied on weak passwords and missing two-step sign-in. TechCrunch
The attackers are using AI, and their methods are old. Weak or known passwords and missing two-step sign-in opened the doors. TechCrunch AI makes simple attacks faster and cheaper to run against many targets at once. Before buying new AI security tools, require two-step sign-in on every account, retire known default passwords, and confirm any outside security testers keep their own tools off the open internet.
Your security team answers for basic sign-in rules, and any outside testers you hire answer for keeping their tools locked down. If either slips, attackers can use AI to find the gap quickly.
In the November 3 midterms, voters backed restrictions or rejected data center expansion in 40 of 41 local measures across 10 states. Two days later, OpenAI pledged to pay the full cost of electricity and related grid work at future data centers it owns, publish water and power use, and avoid secrecy agreements with public officials. Its Georgia site lets the utility cut deliveries by up to 1,000 megawatts during peak demand. Futurum
Local approval is becoming the scarcest ingredient in AI. Suppliers that pay their own power bills are likely to pass some of that cost into prices, and sites that cannot win local support will delay new service. Futurum Ask your AI and cloud suppliers where your work runs, what happens to service if a site is delayed, and whether power costs can be added to your contract later.
Palmer looks at OpenAI's deal letting Georgia Power cut up to 1,000 megawatts to its planned campus during peak demand. He wants the obligations behind the promise: how each commitment is verified and enforced, and who pays if the campus arrives late or never opens. Shelly Palmer The same question applies to every AI contract that promises savings.
AI is getting cheaper to use, and its hidden costs are moving to whoever signs the contract.
This quarter, assign an owner for AI spending, basic AI security and supplier commitments on power and location.
Price cuts and open alternatives have made AI a buyer's market.
Suppliers will compete on discounts now and recover costs later through power charges, minimum commitments and premium security tiers.
Track the AI cost of each finished task alongside total spend.
Refuse multiyear minimums while prices fall, and cap any new power or location charges suppliers try to pass through.
Ask management to confirm two-step sign-in covers every account and that outside security testers keep their tools locked down.
Ask which AI suppliers depend on sites facing local opposition.
The shift this edition is AI becoming cheap to use while its costs move elsewhere: to discounts that lock buyers in, to weak passwords attackers exploit, and to communities that pay for power. The assumption it breaks is that a falling price means a lower total cost. The decision it forces is to price AI by the finished task and write the hidden costs into every contract.
If your AI bill fell next quarter, would you know who started paying the difference?
Make AI accountable. Measure the results. Redesign the business.