A few weeks ago, Morgan Stanley published what might be the most important — and most ignored — report of the year so far. The headline: a transformative leap in artificial intelligence is coming in the first half of 2026. The fine print: it may already be underway, and the rest of the world is largely asleep at the wheel.
The Intelligence Curve Is Steepening Fast
The investment bank's analysis centres on a simple but staggering premise: the compute being poured into AI model training is growing at a pace that has no historical parallel. Researchers pointed to a recent interview with Elon Musk, in which he argued that applying 10x the compute to large language model training effectively doubles a model's "intelligence" — and crucially, the scaling laws underpinning that claim are still holding firm.
This isn't just speculation anymore. OpenAI's GPT-5.4 "Thinking" model, released just days ago, scored 83.0% on the GDPVal benchmark — a measure of performance on economically valuable, expert-level tasks. That score places it at or above the capability level of human domain experts across a wide range of fields.
To put that in perspective: a year ago we were still debating whether LLMs could reliably pass professional exams. Now we're tracking whether they outperform the professionals themselves. The curve isn't just steepening — it's starting to look vertical.
The Infrastructure Can't Keep Up
There's a brutal constraint hiding behind all the optimism. Morgan Stanley's "Intelligence Factory" model projects a net U.S. power shortfall of 9 to 18 gigawatts through 2028. That's a 12–25% deficit in the electricity required to run the data centres being planned and built right now.
Developers aren't waiting for the grid to catch up. They're converting decommissioned Bitcoin mining facilities into high-performance compute centres, deploying natural gas turbines on-site, and standing up fuel cells to bridge the gap. A new economic dynamic — shorthand "15-15-15" — is emerging: 15-year data centre leases at 15% yields, generating $15 per watt in net value creation. It's a capital intensity that makes traditional infrastructure investment look modest.
This is the unglamorous side of the AI boom that rarely makes the front page. The intelligence explosion has a power bill, and right now, nobody's entirely sure how it gets paid.
Jobs Are Already Going — Not "Eventually"
Perhaps the most sobering part of the Morgan Stanley report is the labour market section. The bank doesn't frame job displacement as a future risk. It frames it as a current reality. Executives are already executing large-scale workforce reductions because AI tooling has made whole categories of work faster, cheaper, and automatable. The report describes AI as a "powerful deflationary force" — not one that's arriving, but one that's already applying downward pressure.
Sam Altman has been more direct. He's publicly envisioned a near-future where one to five person companies — fully AI-augmented — can out-compete large incumbents that employ thousands. Whether or not you think that's hyperbole, the direction is undeniable: leverage per person is increasing faster than most organisations are prepared to absorb.
What This Means If You're Building
For engineers, architects, and founders in the trenches, a few things stand out from all of this:
Benchmark scores are now business metrics. When a model hits 83% on GDPVal, it's not just an academic milestone — it's a signal about what you can delegate, automate, and build on top of. If you haven't updated your mental model of what AI can do in the last six months, you're working with stale assumptions.
Infrastructure is the bottleneck, not intelligence. The models are moving faster than the power grid. That has real implications for latency, cost, and availability — especially for workloads that need to scale globally.
Organisational inertia is the real risk. The technology is outpacing the ability of most companies to adapt their processes, headcount, and culture. The firms that figure out how to restructure around AI augmentation will pull away from those still debating it in steering committees.
A Different Kind of Readiness
Morgan Stanley's warning isn't doom-and-gloom — it's a wake-up call about the speed of change. The breakthrough isn't some future event to prepare for. It's a process already in motion, measured in benchmark points and power bills and redundancy notices.
The question isn't whether the world is ready. It's which parts of it are paying attention closely enough to move first.
Sources: Morgan Stanley "Intelligence Factory" report (March 2026), Fortune, OpenAI GDPVal benchmark results.
Top comments (0)