GT Axis Daily, July 15, 2026

OpenAI's newest model, GPT-5.6, is now the default engine behind Microsoft 365 Copilot across Word, Excel, PowerPoint, and Chat. What's notable isn't the upgrade, it's that Microsoft named the model publicly at all. For years the assistant was just "Copilot" and the engine under the hood was nobody's business. Naming it is a small admission that the model choice is now a product feature customers care about, and it lands while Microsoft is openly building its own models to reduce its dependence on OpenAI.

IBM warned that its quarter came in $660 million light on revenue, with its infrastructure business down 7%. The stock fell 25.2% in a single session, wiping out $67 billion and beating its own Black Monday 1987 record for worst day ever. Here's the part that matters: CEO Arvind Krishna said clients shifted their end-of-quarter spending toward hardware that's in short supply, trying to buy before prices rise. Translation for any leader: your customer's AI budget isn't new money, it's money moved off something else, and right now the thing it's being moved off of is software. · [CNBC](https://www.cnbc.com/2026/07/14/ibm-warns-second-quarter-earnings-fell-short-of-expectations.html)

Sam Altman pitched Washington on handing the government a 5% stake, and floated a bigger version where every major US AI company routes 5% of its equity into a public fund modeled on Alaska's oil fund, the one that pays every resident an annual check. The stated logic is that AI will create so much wealth the public should own a slice. The timing tells you more: OpenAI is weeks from an IPO filing, and a government that profits when you profit regulates you differently than one that doesn't. Nothing is agreed, and a deal this size would likely need an act of Congress.

At Google Cloud Next, Google launched the Gemini Enterprise Agent Platform. An "agent" is AI that does the task rather than just answering questions about it, and the pitch here isn't that Google's agents are smarter. It's that IT departments can watch them, audit what they did, set limits on what they're allowed to touch, and shut them off. That's the actual blocker in most companies. Nobody's stalling because the model is dumb, they're stalling because nobody wants to explain to the board how software with nobody's job title leaked the customer list.

The Taiwanese company that manufactures nearly every advanced AI chip on earth, including NVIDIA's and Apple's, reported June revenue up 68% from a year ago and up 6% from May. First half revenue is up 36%. This is the single most useful number in AI right now, because it's the point where announcements turn into physical objects. Every gigawatt data center and every custom chip program eventually becomes an order at these fabs. A record month means the buildout is real, not slideware. It also means the whole AI economy still runs through a handful of buildings on one island.

The Jane Goodall Institute and FormationQ are using IonQ's quantum computers to run a two-year study on why some primate species fight lethally and near-identical cousins live peacefully. They're doing it by simulating thousands of virtual primates foraging and interacting across artificial landscapes, then varying the ecology to see what tips a group toward violence. Why this matters beyond the charm: it's a real research problem where the math genuinely gets too big for normal computers, which is a more honest quantum use case than most of what gets announced.

The Richland Parish data center is being expanded to 5 gigawatts, enough electricity to run a mid-sized city, making it the largest in Meta's fleet and one of the largest AI infrastructure projects anywhere. It'll house Hyperion, Meta's biggest AI training cluster. The local numbers are worth noting because they're where this stops being abstract: 7,500 construction jobs at peak, $1.6 billion contracted to Louisiana businesses, and teachers in the parish getting bonuses up to $50,000 off the tax revenue.

That's more in one half-year than all of 2025 combined. But the headline number hides the real story: OpenAI and Anthropic alone accounted for $217 billion, which is 43% of everything. More than 70% of all startup money in the second quarter went to AI companies, up from just under half a year earlier. This is not a broad boom, it's an extremely concentrated bet.