Big Tech banks $160B+ in quarterly AI-stake gains — Alphabet, Amazon, Nvidia, Microsoft: the four companies booked more than $160 billion in unrealized accounting gains in Q2 2026 from the rising valuations of their stakes in private AI companies (OpenAI, Anthropic, SpaceX and others), classified as “other income” — more than double the ~$69B of the prior quarter. Alphabet led at ~$98B and Amazon ~$53B, much of it from the June IPO of SpaceX at a $1.77T market cap. The catch: it is mark-to-market paper wealth, not cash. Crypto Briefing
Caterpillar is taking its mining-autonomy playbook into broad industrial AI deployment: after decades of autonomous haul trucks, drilling rigs, underground loaders and remote terrain-intelligence software, the industrial giant says it will apply that learning to more dynamic environments — jobsites, quarries and construction sites. CTO Jaime Mineart framed it as “taking all of that learning from mining and bringing it into much more dynamic environments.” Robotics transport, deployment and orchestration know-how is becoming the entry barrier for industrial AI. TechCrunch
Musk starts in-house gas-turbine blade manufacturing to power AI data centers — vertical integration into electricity: reporting says SpaceX now makes its own turbine blades for its generators, cutting generator delays by 18 months. As grid power becomes the real bottleneck for US AI data-center buildout, xAI-linked capacity is moving to own electricity generation — while critics note the added turbines carry a pollution problem. TechCrunch
Public backlash hits AI surveillance procurement — Texas blocks funding for Flock: Governor Greg Abbott ordered all state agencies to stop funding AI license-plate-recognition surveillance maker Flock as bipartisan scrutiny and community backlash grow (Texas spent at least $30M on grants for the cameras). For AI startups selling into the safety and public-sector market, this is a signal that “public safety deals” carry direct political and community risk that can be reversed overnight. The Verge · Mashable
The circularity of Big Tech earnings — AI ecosystem value is now propped up by stake mark-ups: the large paper gains booked by Alphabet and Amazon show a structure where AI firms repeatedly value up one another, and Big Tech then books that appreciation as earnings. Because those are balance-sheet gains rather than new business, AI startup valuations and Big Tech earnings are increasingly moving in step — and can unwind together. Crypto Briefing
Design auditability and explainability into public-sector AI from day one: the Flock reversal underscores how durable a government AI product (traffic-camera surveillance, biometrics, predictive policing) really depends on political and community acceptance. Startups that build in surveillance-intensity controls, bias reporting, audit logs and an open governance framework from the start are relatively insulated from procurement backlash. The Verge
Apply the “autonomy playbook” to industrial AI — design workflow and human-in-the-loop alongside the model: Caterpillar’s lesson is that industrial AI wins or loses on workflow redesign and human/machine role definition, not raw model scoreboards. Teams building factory, construction or logistics AI should prototype the operational process and human-in-the-loop guardrails before optimizing model performance. TechCrunch
Treat energy access as a first-class design constraint: Musk’s move to build power generation in-house reflects that electricity, not chips alone, is now the frontier for AI compute. Data-heavy product teams should bake power cost and grid-delay risk into their unit economics early, and consider renewables or microgrid partnerships as a differentiation lever. TechCrunch
The accounting mirror of the AI bubble — what happens when the gains reverse?: if some backed startups IPO below their current private marks (or private valuations reset), the ~$160B in unrealized gains can unwind as quickly as they appeared. How much this paper accounting is inflating reported profit growth will remain a key question next quarter. Crypto Briefing
Will the AI-surveillance backlash harden into regulation? If the Texas move spreads to other states or federal procurement limits and surveillance-AI rules, the product-market trajectory for public-sector AI (policing, schools, public spaces) could shift materially. Worth tracking along the political cycle. Mashable
Is “power vertical integration” a moat only the giants can afford? If Musk-style moves become the norm, smaller AI operators that can’t absorb power, permitting and financing costs could lose relative competitiveness. What room remains for small cloud, edge and regional players is the next phase to watch. TechCrunch