2026-09-09
By Vadym · Generated with AI, curated by me
Today is a spending day: China commits half a trillion dollars to compute, Mistral raises the largest round a European tech company has ever closed, and Accenture ships a thousand engineers to make Gemini useful inside real companies. Sitting in the middle of all that money is OpenAI's Navier-Stokes claim — a reminder that impressive and verified are not the same thing.
OpenAI reported that an internal model, run as 10,000 agents over 88 hours, produced a proof of finite-time blowup for the forced 3D Navier-Stokes equations, publishing a 165-page write-up and a Lean 4 formalization. But the result covers the forced version of the equations, not the unforced version the Clay Mathematics Institute's $1 million prize actually specifies, OpenAI hasn't sought independent verification, and NYU mathematician Tristan Buckmaster has raised questions about the work — including a dispute over whether de-identified data from other mathematicians' research shaped it.
Google DeepMind released AlphaGenome Atlas, precomputed predictions for the molecular effect of every possible single-letter substitution across the genome's roughly 3 billion positions — about 9 billion variants in total, spanning more than a petabyte of data, some 30 times the size of the AlphaFold database. It's free for noncommercial research through a web portal, the AlphaGenome API, and Google Antigravity, with commercial access planned via Google Cloud.
Mistral AI closed a €3 billion (roughly $3.5B) Series D led by Samsung Electronics, with EQT's Scaleup Europe Fund and PSG Equity co-leading alongside new money from Advent, BlackRock-managed funds, and the Luxembourg government. The post-money valuation tops €21 billion, nearly double last year's €11.7B mark and the largest equity round a European tech company has ever closed. CEO Arthur Mensch says the capital is going toward building and owning data centers, not just renting capacity.
Accenture and Google Cloud formed the Accenture Gemini Enterprise Business Group, planning to embed a 1,000-person forward-deployed engineer workforce directly inside client organizations to scale agentic AI deployments. The unit draws on Accenture's roughly 50,000 Google Cloud-trained staff. As a proof point, the companies point to YouTube's use of a Gemini Enterprise agent during an NFL Sunday Ticket demand surge, which they say lifted customer sentiment 11% and cut average handling time 37%.
Samsung named Yoon Jang-hyun — CTO and president of its Device eXperience division — to lead both the hardware and AI software teams building its first humanoid robot, ahead of a possible CES 2027 debut. Handing a software-side executive authority over the physical design is meant to let Samsung shape the robot's body around its AI stack from the start, instead of bolting AI onto hardware decided separately.
China's Ministry of Industry and Information Technology laid out a 2026–2030 plan to grow national AI computing capacity roughly fourfold, to 9,800 exaflops, up from 2,185 exaflops in June (itself up 177% year-over-year). The plan commits ¥3.8 trillion (about $532B) in cumulative IT infrastructure spending, calls for "orderly deployment" of clusters with 10,000 to 100,000-plus accelerator cards, and pushes harder on domestically produced chips.
Six stories, one thread: everyone's betting big before anyone's fully checked the receipts. China and Mistral are pouring hundreds of billions into compute on the assumption today's scaling laws hold for years. Accenture and Samsung are betting the next unlock is organizational — engineers embedded on-site, hardware and software under one boss — not another model release. And OpenAI's Navier-Stokes claim sits right in the middle of all that spending as the cautionary tale: impressive isn't the same as verified, and the gap between the two is where trust erodes.
— Boba
Curated by Vadym