Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-07-01 · high
Aramco Ventures and NVIDIA led the round, funding a plan to grow serving capacity 50 times over five years.
Why now: The round lands as capital keeps flowing into the infrastructure that runs open models like DeepSeek and Kimi, not into new frontier labs.
Together AI raised $800 million in a Series C at an $8.3 billion valuation, led by Aramco Ventures with NVIDIA and Vista, per Yahoo Finance. The company plans to grow serving capacity roughly 50 times over five years, per the same report. That capacity runs open models including DeepSeek, Nemotron, MiniMax, and Kimi, the infrastructure layer underneath any team that picks open weights over a closed API.
Aramco Ventures leading the round is worth sitting with. That's serious money moving into AI infrastructure rather than oil and gas diversification, and NVIDIA's presence keeps the round inside the same ecosystem that sells the chips Together AI runs its fleet on.
The 50x capacity target says more than the check size does. Together AI isn't positioning as a niche host for open-weight holdouts. It's betting that demand for running DeepSeek or Kimi in production keeps climbing for years, not quarters.
My read: this is the clearest sign yet that open models don't need to win the frontier benchmark race to win adoption. They need cheap, funded infrastructure to run on, and that infrastructure just got $800 million richer. Watch whether the capacity buildout tracks real demand or gets ahead of it. That's the number that separates a real business from a bet on one.
The round lands while capital keeps flowing into the supply side of the open-weights story, not into new frontier labs chasing the next closed model.
Each link below shares sources, entities, or timing with this story.
A model-routing API covering OpenAI, Anthropic, DeepSeek, Moonshot, Minimax, Nvidia, xAI, and Z.ai, with strategies including preferring flex usage tiers, specifying up to three performance benchmarks for automatic selection, or sending only complex queries to expensive models...
Martin Alderson's essay "The upcoming AI margin collapse, part 1: GLM 5.2" hit 675 points and 462 comments on Hacker News, and it's the rare HN chart-topper that's actually about spreadsheet math instead of vibes. The argument is simple. Z.ai's GLM 5.2 delivers frontier-adjace...
I've spent the last year assuming that if I wanted real agentic coding quality, I paid for a closed model. That assumption took a hit on June 1. MiniMax shipped M3 with a new sparse-attention architecture (they call it MSA) that handles up to 1M tokens at roughly 9x prefill an...
Three moves, two days, no coordination between them. August 10–11: GitHub shipped Ollama as a BYOK provider inside Copilot for JetBrains (GitHub Changelog). Unsloth released Unsloth Desktop with a command literally named unsloth start claude, which points Claude Code and Codex...
Re-architected PhysicsNeMo libraries and updated CUDA-X exposed as agent-ready tools, so engineering agents can invoke AI physics models, accelerated solvers and quantum chemistry directly instead of through bespoke wrappers. NVIDIA Research's ACE-RTL agent with Nemotron 3 Ult...
Reuters, via Tech Startups, reports capital released against deployment milestones with Anthropic deploying up to two gigawatts of Instinct MI450 starting 2027. Same structure as Nvidia/OpenAI: compute vendor capital flowing to the lab that commits to buy the silicon. A two-gi...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.