The Arena
Meta says its own AI silicon is ready — and a Chinese driving-chip maker says the same about its scale. Meta plans to put MTIA 450 — its third-generation custom accelerator, codenamed Arke — into data centers in the first half of 2027, with MTIA 500 (Astrid) finishing design work in about a month and reaching production data centers by the end of 2027, according to Bloomberg.
The Arena
Two stories on the same question this afternoon: what happens to frontier-lab pricing when the enterprise stack stops renting reasoning by the token. Plus a security layer for the robot fleet. Salesforce debuted Koa, a reasoning model it built with Nvidia on top of the open-weight Nemotron base, and says it will handle the multi-step thinking inside Agentforce that used to be sent out to Claude or ChatGPT.
The Arena
The edge is the story today: new silicon built for agents, an OS safety suite that runs its models on the device, and a fresh open model betting on spatial reasoning — all landing within hours of each other. MediaTek launched the Dimensity 9600 Pro, a TSMC 2nm flagship whose headline spec is an agent, not a benchmark.
The Arena
Euclyd, an Eindhoven chip startup founded in 2024, closed a Series A of more than €200 million (about $231 million) co-led by Samsung, Somerset Capital Partners, the EQT-managed Scaleup Europe Fund and Innovation Industries, the company announced on September 15. Rounding out the syndicate are EIFO, imec.xpand, the Brabant Development Agency and Quadri.
The Stack
NVIDIA's research arm published the code for a loop where agents studied other agents' transcripts and cut their own token bill. Meanwhile South Korea's central bank measured who actually gets the AI jobs. NVLabs released SoL-Pi, an MIT-licensed extension for the Pi coding agent that packages four efficiency mechanisms discovered by automated research loops rather than by engineers. It installs on an unmodified Pi release, and every mechanism is opt-in and off by default.
The Arena
The AI slowdown fight is drowning out a quieter story: some of the labs' biggest enterprise customers are starting to walk back. And a former president just told Democrats to make AI safety their opening wedge. Palantir, Nvidia and Booz Allen Hamilton are prepared to restrict or drop frontier AI models unless Anthropic and OpenAI guarantee they will not misuse company intellectual property, according to a report that spread across Reuters and The Information.
The Arena
A rival model is now living inside Microsoft's own productivity suite, consumer GPUs are being vacuumed up by AI builders, and a former Apple researcher raised $50 million to make AI avatars stop staring at you blankly. Microsoft has added Grok as a model option in Copilot for Word, Excel and PowerPoint.
The Arena
The market spent Monday repricing the AI buildout after Dario Amodei's call to pace the frontier. The people who actually write the compute cheques spent the same day saying the number has not moved. Broadcom CEO Hock Tan said the slowdown debate has not touched his AI revenue forecasts. Asked by Jim Cramer on "Mad Money" whether anything in the argument over pacing frontier models had made him reconsider the company's fiscal 2027 and 2028 AI semiconductor targets, Tan answered "No, not in the…
The Arena
Monday's news ran from the inside of the AI rack to the inside of the White House — a networking challenger that says the GPUs are not the bottleneck, a private Altman-Trump meeting that surfaced the same day the two split publicly on slowing down, and a mathematician's attempt to say what the profession should be for. Cornelis Networks announced roughly $205 million in funding alongside Active Compute Fabric, an open networking architecture that puts programmable compute inside the fabric…
The Stack
Hot Chips 2026 wrapped with a memory bet, and the funding calendar is unusually busy for a Monday. d-Matrix says stacking memory directly on the compute die beats HBM by 20x on bandwidth density. The startup's Raptor accelerator puts a TSMC N4 logic die face-to-face on top of custom 3D DRAM and claims roughly 100 TB/s per card at 32 GB — against a practical ceiling of about 20 TB/s for today's HBM4 packages, which d-Matrix attributes to limited package "beachfront" and raw physics: driving 100…
The Arena
Two charts the industry doesn't usually read together — one on API prices, one on GPU shelves — both say the same thing: the AI cost curve is bending, and not in the direction the growth story needs. The market price of a million LLM tokens has fallen below one dollar. Silicon Data's LLM Token Spend Index, which tracks the average paid price per million tokens across the market, dropped 29% in August to $0.97 — the first print under the dollar mark, and roughly a halving from the $2-plus level…
The Stack
A compatibility layer that Nvidia spent two decades making unnecessary now trains real PyTorch models on a consumer Radeon, and it ships as an installer. A Windows stack built on ZLUDA + AMD's ROCm/HIP completed a real LibTorch reinforcement-learning run on an AMD Radeon RX 9060 XT, and the maintainer published the whole thing as a reproducible installer.