GT Axis Daily, July 10, 2026

GPT-5.6 comes in three flavors named Sol, Terra, and Luna: Sol is the premium brain for hard problems, Terra is the everyday workhorse at half the cost, and Luna is the cheap-and-fast option. The clever part is the naming itself, the number tells you the generation and the name tells you the "tier," so a business can pick intelligence, speed, or price without guessing. Translation: OpenAI is done selling one expensive model to everyone and is now selling a menu, which is exactly what you do when customers start watching the bill.

Now branded SpaceXAI, the company put out what it calls an "Opus-class" model, meaning it claims to rival the strongest model from Anthropic, but faster and cheaper at $2 per million words in and $6 out. It was trained alongside Cursor, the popular AI coding tool SpaceXAI just bought, so the pitch is squarely at engineers and technical work rather than casual chat. The tell here is timing: launching hours before OpenAI's big day is a shot across the bow, and pricing it low is how you buy your way into a crowded room.

Tens of thousands of AI requests a week in Excel and Outlook are now being handled by Microsoft's own in-house models, called MAI, instead of OpenAI's or Anthropic's. The simple jobs go first, cleaning up a spreadsheet, summarizing an email, while the hard creative work still runs on OpenAI for now. Microsoft's AI chief has said the goal is to eventually stop paying Anthropic entirely. Why it matters: the company that helped fund OpenAI is now building its own exit, one boring Excel prompt at a time, because at Microsoft's scale even a few cents per request adds up to real money.

The two old rivals are lining up together against the AI pure-plays, a sign the "big platform" incumbents see the upstarts as the real threat now. For a non-technical read: the giants have the distribution, the enterprise contracts, and the wallets to outlast a price war, and they'd rather cooperate at the top than let a startup rewrite the rules underneath them.

GitHub shipped a standalone desktop version of Copilot built to run tasks start to finish, plus browser tools that let the AI actually open web pages, click around, and check that an app works. In plain terms, this is the shift from "AI that suggests the next line of code" to "AI you hand a job and walk away from." For any leader, that's the real productivity story: fewer people babysitting software, more of it acting on its own with a human checking the result.

Quantum machines are famously error-prone, which is the main thing keeping them out of real-world use. IBM's new open-source tool, Qiskit Paulice, automatically weaves error-catching checks into quantum programs so the hardware's noise does less damage. The plain-English point: IBM is doing the unglamorous plumbing work that turns quantum from a lab curiosity into something businesses might one day actually run.

IonQ partnered with Archer Materials in a roughly $1.5 million agreement to study on-shore quantum deployment and to sharpen fraud-detection systems using IonQ's hardware. It's a small check, but the direction matters: IonQ keeps stacking real-world partnerships instead of just publishing lab results. The catch for investors is that these are multi-year bets, not this-quarter revenue.

SEEQC submitted paperwork for a Nasdaq listing to fund its work on chip-scale control systems, the specialized hardware that makes quantum processors run, with partnerships tied to NVIDIA and IBM. Watch this one: when the "picks and shovels" suppliers of a young industry start going public, it's a sign the field is maturing from pure research toward an actual supply chain.

Together AI closed an $800 million round led by Saudi oil giant Aramco's venture arm, with NVIDIA joining in, and said it plans to grow its computing capacity fiftyfold over five years. The tell for leaders: the AI gold rush money isn't chasing another chatbot, it's buying the raw computing power and warehouses everything else runs on. When Aramco and NVIDIA write the same check, follow the plumbing, not the apps.

Google's parent said it plans to sell about $80 billion in stock to fund data centers and computing capacity, with Warren Buffett's Berkshire Hathaway agreeing to buy $5 billion of it. The plain read: even the richest tech company on earth is reaching for outside cash to keep up with the build-out, and having Buffett, the ultimate value investor, put money in is a quiet vote of confidence. This is how big the infrastructure bill has gotten.