Sent to 20 readers
Race to AGI Daily Digest - Tuesday, June 30, 2026
Share:
TLDR
Google is turning the GPU crunch into a power move by limiting Meta’s Gemini access, reinforcing compute as a strategic weapon.
Hyperscale Data’s $120m Michigan AI campus reserve shows how AI debt turns into long-term concrete and power contracts.
Oracle’s Fusion agentic AI apps and 8090 Labs’ $135m AI software factory mark a landgrab for enterprise ‘agentic’ stacks.
Chip export rules remain the backdrop, as the fight over who gets H200-class compute shapes both markets and national security.
The Full Story
Building on Monday’s macro warning about AI debt, today shows what that looks like on the ground: the firms that control compute and distribution are starting to pull real power moves.
At the top of the stack, Google is once again limiting Meta’s Gemini access, explicitly blaming the GPU crunch. So compute isn’t just a cost line anymore; it’s a negotiation lever. If you want to see the dynamic in black and white, check the latest coverage -> and then look at who actually owns the cloud knobs at Google ->. This is our “compute crunch as strategic weapon” storyline playing out in public.
Lower down, the concrete is being poured. Hyperscale Data has set a $120 million reserve for a Michigan AI campus, essentially pre-booking future compute in real estate form Michigan AI campus plan ->. Following Monday’s talk about AI debt as a macro risk, this is the sort of long‑dated, capital-heavy bet regulators are worried about.
On the application side, the word of the day is “agentic.” Oracle just rolled out Fusion agentic AI apps aimed at supply chain execution Oracle’s Fusion agentic apps ->, turning multi-step workflows into semi-autonomous agents. At the same time, 8090 Labs raised $135 million for an “AI software factory,” with Salesforce Ventures on the cap table 8090 Labs raise ->. That ties the compute story back into the enterprise AI ecosystem that Salesforce is quietly stitching together Salesforce ->.
Even on the tooling fringe, Base44’s new Base1 LLM is pitched around cutting “vibe coding” costs launch details ->. Everyone is trying to either stretch scarce GPUs further or lock in more of them.
Stocks are echoing the same split: Tesla and Baidu on the demand side, TSMC and ASML on the supply side, and Google sitting in the middle, all up sharply. Overlay that with the ongoing worries around H200 exports to China AI chip exports narrative ->, and you get the week’s through-line: access to compute is becoming both a business moat and a geopolitical fault line.
Get This Delivered Daily
Join thousands of AI professionals who start their day with Race to AGI.