I. Cheap software, expensive life
Corporate software prices fall 22% to 29% a year by 2030, yet the cost of everyday life barely budges (+2.68% to +2.85%). Households spend roughly 95% of their budgets on housing, healthcare, electricity, food, and physical services, and only about 5% on software. Worse, the AI buildout itself pushes physical prices up: datacenters bid against everyone else for scarce master electricians, high-voltage transformers, and gas turbines, while nurses and skilled trades win steady 5% union raises. That is why the Fed stays stuck at 3.00% in two of the three runs even while software prices are in freefall.
II. Every fix creates the next bottleneck
No single shortage lasts all four years because all nine groups immediately route around whatever is scarce. In late 2026, the constraint is 4-to-6 year utility grid waitlists. Everyone tries to bypass the grid at once by ordering off-grid gas turbines, which jams turbine factories in early 2027. Burned by speculative filings, utilities then require 100% upfront cash deposits in late 2027, wiping out unbacked developers right as power plants started in 2026 finish building. That flips electricity into a brief +3.2 GW surplus in early 2028 and pushes the bottleneck inside companies: 30-year-old undocumented databases, residential ratepayer politics, scarce university wet labs, and human liability sign-offs.
III. Even when agents work, software stocks still drop Branch II · Agent Breakthrough
You might expect tech stocks to boom if agents crack messy corporate IT in 2028 (lifting readiness to 59.8% vs. 42.5% in Baseline). Real usage does jump: big companies book every spare megawatt, pushing 2030 AI power demand up 37% to 45.4 GW, lifting GDP growth to 2.35%, and pulling headline inflation low enough (1.26%) for the Fed to cut to 2.25%. Yet tech equities still fall to 64.5. The catch is that the same agents that clean up old databases also make it cheap to rip out incumbent enterprise software. Switching costs vanish, corporate buyers force 29%/yr price cuts, regional banks liquidate leveraged SaaS buyouts for pennies, and the gains go to the buyers (hospitals, manufacturers, biotech labs, and lean startups) rather than software multiples.
IV. Taking losses early beats muddling through Branch III · Credit Crunch
In Baseline, troubled datacenter and software debt drags out: election-year utility rebates mask rising grid costs in 2028, expire in 2029 (+3.15% physical CPI), and trigger state surcharges on commercial datacenters; cloud giants dodge the fees by relabeling idle servers as defense clouds until states close the loophole in 2030 (1.70% GDP, 62.5 equities). In Credit Crunch, private-credit datacenter loans default in early 2028 and Washington refuses a bailout. Growth dips to 1.4% at first, then two things turn the trajectory around: utilities keep forfeited developer deposits and use them to lower household electric bills (avoiding the 2029 voter backlash), and clearing bad debt early lets banks reopen IPO financing in 2029 for cash-flowing biotech and robotics firms. By 2030, the no-bailout path finishes ahead of Baseline on both GDP (2.50%) and tech equities (73.5).
V. Insurers, unions, and public talk vs. private moves
Across all three scenarios, US entry-level desk hiring drops 28–30%: painful for new graduates, yet far short of total replacement. What puts a floor near 70% is corporate insurers and courts. As AI-generated bugs, synthetic phishing, and automated legal filings flood in, liability insurers refuse coverage unless accredited humans verify and sign off. Junior work shifts from writing first drafts to auditing outputs for legal liability, with routine prep moving offshore and final sign-off kept domestic. Underneath the headline numbers, almost every turn is driven by the gap between what groups announce and what they quietly do: employers praising “worker copilots” while freezing campus hiring, utilities using tech deposits to clear residential transformer backlogs, and universities releasing open models while leasing empty humanities buildings to pharma trials.